diff options
82 files changed, 2676 insertions, 1890 deletions
diff --git a/aterm-gc.supp b/aterm-gc.supp index dcd6371fe5aa..21b9a2372856 100644 --- a/aterm-gc.supp +++ b/aterm-gc.supp @@ -115,3 +115,35 @@ fun:* fun:AT_collect } + +{ + ATerm library conservatively scans for GC roots + Memcheck:Value4 + fun:* + fun:* + fun:mark_phase +} + +{ + ATerm library conservatively scans for GC roots + Memcheck:Cond + fun:* + fun:* + fun:mark_phase +} + +{ + ATerm library conservatively scans for GC roots + Memcheck:Value4 + fun:* + fun:* + fun:mark_phase_young +} + +{ + ATerm library conservatively scans for GC roots + Memcheck:Cond + fun:* + fun:* + fun:mark_phase_young +} diff --git a/bootstrap.sh b/bootstrap.sh index f007c713bf54..2547f5dc88db 100755 --- a/bootstrap.sh +++ b/bootstrap.sh @@ -1,4 +1,5 @@ #! /bin/sh -e +rm -f aclocal.m4 mkdir -p config libtoolize --copy aclocal diff --git a/configure.ac b/configure.ac index f108c53beee4..e34e4ba3acff 100644 --- a/configure.ac +++ b/configure.ac @@ -50,39 +50,24 @@ AC_DEFINE_UNQUOTED(SYSTEM, ["$system"], [platform identifier (`cpu-os')]) test "$localstatedir" = '${prefix}/var' && localstatedir=/nix/var -# Whether to produce a statically linked binary. On Cygwin, this is -# the default: dynamically linking against the ATerm DLL does work, -# except that it requires the ATerm "lib" directory to be in $PATH, as -# Windows doesn't have anything like an RPATH embedded in executable. -# Since this is kind of annoying, we use static libraries for now. - -AC_ARG_ENABLE(static-nix, AC_HELP_STRING([--enable-static-nix], - [produce statically linked binaries]), - static_nix=$enableval, static_nix=no) - -if test "$sys_name" = cygwin; then - static_nix=yes -fi - -if test "$static_nix" = yes; then +# Windows-specific stuff. On Cygwin, dynamically linking against the +# ATerm DLL works, except that it requires the ATerm "lib" directory +# to be in $PATH, as Windows doesn't have anything like an RPATH +# embedded in executable. Since this is kind of annoying, we use +# static libraries for now. +if test "$sys_name" = "cygwin"; then AC_DISABLE_SHARED AC_ENABLE_STATIC fi -# Windows-specific stuff. -if test "$sys_name" = "cygwin"; then - # We cannot delete open files. - AC_DEFINE(CANNOT_DELETE_OPEN_FILES, 1, [Whether it is impossible to delete open files.]) -fi - # Solaris-specific stuff. if test "$sys_name" = "sunos"; then # Solaris requires -lsocket -lnsl for network functions - ADDITIONAL_NETWORK_LIBS="-lsocket -lnsl" - AC_SUBST(ADDITIONAL_NETWORK_LIBS) + LIBS="-lsocket -lnsl $LIBS" fi + AC_PROG_CC AC_PROG_CXX @@ -101,6 +86,13 @@ AC_DISABLE_STATIC AC_ENABLE_SHARED AC_PROG_LIBTOOL +if test "$enable_shared" = yes; then + SUB_CONFIGURE_FLAGS="--enable-shared --disable-static" +else + SUB_CONFIGURE_FLAGS="--enable-static --disable-shared" +fi +AC_SUBST(SUB_CONFIGURE_FLAGS) + # Use 64-bit file system calls so that we can support files > 2 GiB. AC_SYS_LARGEFILE @@ -229,6 +221,8 @@ AC_ARG_WITH(bzip2, AC_HELP_STRING([--with-bzip2=PATH], [prefix of bzip2]), bzip2=$withval, bzip2=) AM_CONDITIONAL(HAVE_BZIP2, test -n "$bzip2") +ATERM_VERSION=2.5 +AC_SUBST(ATERM_VERSION) if test -z "$bzip2"; then # Headers and libraries will be used from the temporary installation # in externals/inst-bzip2. @@ -249,6 +243,24 @@ AC_SUBST(bzip2_include) AC_SUBST(bzip2_bin) AC_SUBST(bzip2_bin_test) +AC_ARG_WITH(sqlite, AC_HELP_STRING([--with-sqlite=PATH], + [prefix of SQLite]), + sqlite=$withval, sqlite=) +AM_CONDITIONAL(HAVE_SQLITE, test -n "$sqlite") +SQLITE_VERSION=3070500 +AC_SUBST(SQLITE_VERSION) +if test -z "$sqlite"; then + sqlite_lib='${top_builddir}/externals/sqlite-autoconf-$(SQLITE_VERSION)/libsqlite3.la' + sqlite_include='-I${top_builddir}/externals/sqlite-autoconf-$(SQLITE_VERSION)' + sqlite_bin='${top_builddir}/externals/sqlite-autoconf-$(SQLITE_VERSION)' +else + sqlite_lib="-L$sqlite/lib -lsqlite3" + sqlite_include="-I$sqlite/include" + sqlite_bin="$sqlite/bin" +fi +AC_SUBST(sqlite_lib) +AC_SUBST(sqlite_include) +AC_SUBST(sqlite_bin) # Whether to use the Boehm garbage collector. AC_ARG_ENABLE(gc, AC_HELP_STRING([--enable-gc], @@ -274,8 +286,7 @@ AC_CHECK_FUNCS([setresuid setreuid lchown]) # Nice to have, but not essential. -AC_CHECK_FUNCS([strsignal]) -AC_CHECK_FUNCS([posix_fallocate]) +AC_CHECK_FUNCS([strsignal posix_fallocate nanosleep]) # This is needed if ATerm or bzip2 are static libraries, @@ -285,14 +296,6 @@ if test "$(uname)" = "Darwin"; then fi -if test "$static_nix" = yes; then - # `-all-static' has to be added at the end of configure, because - # the C compiler doesn't know about -all-static (it's filtered out - # by libtool, but configure doesn't use libtool). - LDFLAGS="-all-static $LDFLAGS" -fi - - AM_CONFIG_HEADER([config.h]) AC_CONFIG_FILES([Makefile externals/Makefile diff --git a/corepkgs/buildenv/default.nix b/corepkgs/buildenv/default.nix index 36dd9d0c6a0a..d76f5274099a 100644 --- a/corepkgs/buildenv/default.nix +++ b/corepkgs/buildenv/default.nix @@ -11,4 +11,8 @@ derivation { paths = derivations; active = map (x: if x ? meta && x.meta ? active then x.meta.active else "true") derivations; priority = map (x: if x ? meta && x.meta ? priority then x.meta.priority else "5") derivations; + + # Building user environments remotely just causes huge amounts of + # network traffic, so don't do that. + preferLocalBuild = true; } diff --git a/corepkgs/nar/nar.sh.in b/corepkgs/nar/nar.sh.in index 67933ac67f0b..1369d3a21fb2 100644 --- a/corepkgs/nar/nar.sh.in +++ b/corepkgs/nar/nar.sh.in @@ -7,8 +7,6 @@ dst=$out/tmp.nar.bz2 @bzip2@ < tmp > $dst -@bindir@/nix-hash -vvvvv --flat --type $hashAlgo --base32 tmp > $out/nar-hash - @bindir@/nix-hash --flat --type $hashAlgo --base32 $dst > $out/narbz2-hash @coreutils@/mv $out/tmp.nar.bz2 $out/$(@coreutils@/cat $out/narbz2-hash).nar.bz2 diff --git a/doc/manual/conf-file.xml b/doc/manual/conf-file.xml index ec64e8b114f7..cb47b99418b6 100644 --- a/doc/manual/conf-file.xml +++ b/doc/manual/conf-file.xml @@ -260,7 +260,7 @@ build-use-chroot = /dev /proc /bin</programlisting> Nix store metadata (in <filename>/nix/var/nix/db</filename>) are synchronously flushed to disk. This improves robustness in case of system crashes, but reduces performance. The default is - <literal>false</literal>.</para></listitem> + <literal>true</literal>.</para></listitem> </varlistentry> diff --git a/doc/manual/nix-store.xml b/doc/manual/nix-store.xml index 10bb3eda5774..a32559c033f4 100644 --- a/doc/manual/nix-store.xml +++ b/doc/manual/nix-store.xml @@ -404,6 +404,7 @@ error: cannot delete path `/nix/store/zq0h41l75vlb4z45kzgjjmsjxvcv1qk7-mesa-6.4' <arg choice='plain'><option>--tree</option></arg> <arg choice='plain'><option>--binding</option> <replaceable>name</replaceable></arg> <arg choice='plain'><option>--hash</option></arg> + <arg choice='plain'><option>--size</option></arg> <arg choice='plain'><option>--roots</option></arg> </group> <arg><option>--use-output</option></arg> @@ -587,9 +588,21 @@ query is applied to the target of the symlink.</para> <varlistentry><term><option>--hash</option></term> <listitem><para>Prints the SHA-256 hash of the contents of the - store paths <replaceable>paths</replaceable>. Since the hash is - stored in the Nix database, this is a fast - operation.</para></listitem> + store paths <replaceable>paths</replaceable> (that is, the hash of + the output of <command>nix-store --dump</command> on the given + paths). Since the hash is stored in the Nix database, this is a + fast operation.</para></listitem> + + </varlistentry> + + <varlistentry><term><option>--size</option></term> + + <listitem><para>Prints the size in bytes of the contents of the + store paths <replaceable>paths</replaceable> โ to be precise, the + size of the output of <command>nix-store --dump</command> on the + given paths. Note that the actual disk space required by the + store paths may be higher, especially on filesystems with large + cluster sizes.</para></listitem> </varlistentry> diff --git a/doc/manual/quick-start.xml b/doc/manual/quick-start.xml index 6d96cb5ac0bc..d2431151f360 100644 --- a/doc/manual/quick-start.xml +++ b/doc/manual/quick-start.xml @@ -60,7 +60,7 @@ available remotely.</para></listitem> in the channel: <screen> -$ nix-env -qa โ*โ <lineannotation>(mind the quotes!)</lineannotation> +$ nix-env -qa \* docbook-xml-4.2 firefox-1.0pre-PR-0.10.1 hello-2.1.1 diff --git a/externals/Makefile.am b/externals/Makefile.am index 63150b1f5e71..884d87bf17c9 100644 --- a/externals/Makefile.am +++ b/externals/Makefile.am @@ -12,30 +12,56 @@ $(BZIP2).tar.gz: $(BZIP2): $(BZIP2).tar.gz gunzip < $(srcdir)/$(BZIP2).tar.gz | tar xvf - -have-bzip2: - $(MAKE) $(BZIP2) - touch have-bzip2 - if HAVE_BZIP2 build-bzip2: else -build-bzip2: have-bzip2 - (pfx=`pwd` && \ - cd $(BZIP2) && \ - $(MAKE) && \ - $(MAKE) install PREFIX=$$pfx/inst-bzip2) +build-bzip2: $(BZIP2) + (cd $(BZIP2) && \ + $(MAKE) CC="$(CC)" && \ + $(MAKE) install PREFIX=$(abs_builddir)/inst-bzip2) touch build-bzip2 -install: +install-exec-local:: build-bzip2 mkdir -p $(DESTDIR)${bzip2_bin} $(INSTALL_PROGRAM) $(bzip2_bin_test)/bzip2 $(bzip2_bin_test)/bunzip2 $(DESTDIR)${bzip2_bin} endif -all: build-bzip2 +# SQLite + +SQLITE = sqlite-autoconf-$(SQLITE_VERSION) +SQLITE_TAR = sqlite-autoconf-$(SQLITE_VERSION).tar.gz + +$(SQLITE_TAR): + @echo "Nix requires the SQLite library to build." + @echo "Please download version $(SQLITE_VERSION) from" + @echo " http://www.sqlite.org/$(SQLITE_TAR)" + @echo "and place it in the externals/ directory." + false + +$(SQLITE): $(SQLITE_TAR) + gzip -d < $(srcdir)/$(SQLITE_TAR) | tar xvf - + +if HAVE_SQLITE +build-sqlite: +else +build-sqlite: $(SQLITE) + (cd $(SQLITE) && \ + CC="$(CC)" CFLAGS="-DSQLITE_ENABLE_COLUMN_METADATA=1" ./configure --disable-static --prefix=$(pkglibdir)/dummy --libdir=${pkglibdir} $(SUB_CONFIGURE_FLAGS) && \ + $(MAKE) ) + touch build-sqlite + +install-exec-local:: build-sqlite + cd $(SQLITE) && $(MAKE) install + rm -rf "$(DESTDIR)/$(pkglibdir)/dummy" +endif + + +all: build-bzip2 build-sqlite -EXTRA_DIST = $(BZIP2).tar.gz +EXTRA_DIST = $(BZIP2).tar.gz $(SQLITE_TAR) -ext-clean: - $(RM) -f have-bzip2 build-bzip2 - $(RM) -rf $(BZIP2) +clean: + $(RM) -f build-bzip2 build-sqlite + $(RM) -rf $(BZIP2) $(SQLITE) + $(RM) -rf inst-bzip2 diff --git a/release.nix b/release.nix index c89d79a7d4c9..1fc9405bdb28 100644 --- a/release.nix +++ b/release.nix @@ -33,6 +33,9 @@ let stripHash ${bzip2.src} cp -pv ${bzip2.src} externals/$strippedName + stripHash ${sqlite.src} + cp -pv ${sqlite.src} externals/$strippedName + # TeX needs a writable font cache. export VARTEXFONTS=$TMPDIR/texfonts ''; @@ -71,7 +74,7 @@ let configureFlags = '' --disable-init-state - --with-bzip2=${bzip2} + --with-bzip2=${bzip2} --with-sqlite=${sqlite} --enable-gc ''; }; @@ -92,10 +95,10 @@ let configureFlags = '' --disable-init-state --disable-shared - --with-bzip2=${bzip2} + --with-bzip2=${bzip2} --with-sqlite=${sqlite} ''; - lcovFilter = ["*/boost/*" "*-tab.*"]; + lcovFilter = [ "*/boost/*" "*-tab.*" ]; # We call `dot', and even though we just use it to # syntax-check generated dot files, it still requires some @@ -144,11 +147,11 @@ let with import nixpkgs { inherit system; }; releaseTools.rpmBuild rec { - name = "nix-rpm"; + name = "nix-rpm-${diskImage.name}"; src = jobs.tarball; diskImage = diskImageFun vmTools.diskImages; memSize = 1024; - meta = { schedulingPriority = prio; }; + meta.schedulingPriority = prio; }; @@ -165,7 +168,7 @@ let src = jobs.tarball; diskImage = diskImageFun vmTools.diskImages; memSize = 1024; - meta = { schedulingPriority = prio; }; + meta.schedulingPriority = prio; configureFlags = "--sysconfdir=/etc"; debRequires = [ "curl" ]; }; diff --git a/scripts/GeneratePatches.pm.in b/scripts/GeneratePatches.pm.in new file mode 100755 index 000000000000..2d2653255e54 --- /dev/null +++ b/scripts/GeneratePatches.pm.in @@ -0,0 +1,334 @@ +#! @perl@ -w -I@libexecdir@/nix + +use strict; +use File::Temp qw(tempdir); + + +# Some patch generations options. + +# Max size of NAR archives to generate patches for. +my $maxNarSize = $ENV{"NIX_MAX_NAR_SIZE"}; +$maxNarSize = 160 * 1024 * 1024 if !defined $maxNarSize; + +# If patch is bigger than this fraction of full archive, reject. +my $maxPatchFraction = $ENV{"NIX_PATCH_FRACTION"}; +$maxPatchFraction = 0.60 if !defined $maxPatchFraction; + +my $timeLimit = $ENV{"NIX_BSDIFF_TIME_LIMIT"}; +$timeLimit = 180 if !defined $timeLimit; + +my $hashAlgo = "sha256"; + + +sub findOutputPaths { + my $narFiles = shift; + + my %outPaths; + + foreach my $p (keys %{$narFiles}) { + + # Ignore derivations. + next if ($p =~ /\.drv$/); + + # Ignore builders (too much ambiguity -- they're all called + # `builder.sh'). + next if ($p =~ /\.sh$/); + next if ($p =~ /\.patch$/); + + # Don't bother including tar files etc. + next if ($p =~ /\.tar$/ || $p =~ /\.tar\.(gz|bz2|Z|lzma|xz)$/ || $p =~ /\.zip$/ || $p =~ /\.bin$/ || $p =~ /\.tgz$/ || $p =~ /\.rpm$/ || $p =~ /cvs-export$/ || $p =~ /fetchhg$/); + + $outPaths{$p} = 1; + } + + return %outPaths; +} + + +sub getNameVersion { + my $p = shift; + $p =~ /\/[0-9a-z]+((?:-[a-zA-Z][^\/-]*)+)([^\/]*)$/; + my $name = $1; + my $version = $2; + return undef unless defined $name && defined $version; + $name =~ s/^-//; + $version =~ s/^-//; + return ($name, $version); +} + + +# A quick hack to get a measure of the `distance' between two +# versions: it's just the position of the first character that differs +# (or 999 if they are the same). +sub versionDiff { + my $s = shift; + my $t = shift; + my $i; + return 999 if $s eq $t; + for ($i = 0; $i < length $s; $i++) { + return $i if $i >= length $t or + substr($s, $i, 1) ne substr($t, $i, 1); + } + return $i; +} + + +sub getNarBz2 { + my $narPath = shift; + my $narFiles = shift; + my $storePath = shift; + + my $narFileList = $$narFiles{$storePath}; + die "missing path $storePath" unless defined $narFileList; + + my $narFile = @{$narFileList}[0]; + die unless defined $narFile; + + $narFile->{url} =~ /\/([^\/]+)$/; + die unless defined $1; + return "$narPath/$1"; +} + + +sub containsPatch { + my $patches = shift; + my $storePath = shift; + my $basePath = shift; + my $patchList = $$patches{$storePath}; + return 0 if !defined $patchList; + my $found = 0; + foreach my $patch (@{$patchList}) { + # !!! baseHash might differ + return 1 if $patch->{basePath} eq $basePath; + } + return 0; +} + + +sub generatePatches { + my ($srcNarFiles, $dstNarFiles, $srcPatches, $dstPatches, $narPath, $patchesPath, $patchesURL, $tmpDir) = @_; + + my %srcOutPaths = findOutputPaths $srcNarFiles; + my %dstOutPaths = findOutputPaths $dstNarFiles; + + # For each output path in the destination, see if we need to / can + # create a patch. + + print STDERR "creating patches...\n"; + + foreach my $p (keys %dstOutPaths) { + + # If exactly the same path already exists in the source, skip it. + next if defined $srcOutPaths{$p}; + + print " $p\n"; + + # If not, then we should find the paths in the source that are + # `most' likely to be present on a system that wants to + # install this path. + + (my $name, my $version) = getNameVersion $p; + next unless defined $name && defined $version; + + my @closest = (); + my $closestVersion; + my $minDist = -1; # actually, larger means closer + + # Find all source paths with the same name. + + foreach my $q (keys %srcOutPaths) { + (my $name2, my $version2) = getNameVersion $q; + next unless defined $name2 && defined $version2; + + if ($name eq $name2) { + + my $srcSystem = @{$$dstNarFiles{$p}}[0]->{system}; + my $dstSystem = @{$$srcNarFiles{$q}}[0]->{system}; + if (defined $srcSystem && defined $dstSystem && $srcSystem ne $dstSystem) { + print " SKIPPING $q due to different systems ($srcSystem vs. $dstSystem)\n"; + next; + } + + # If the sizes differ too much, then skip. This + # disambiguates between, e.g., a real component and a + # wrapper component (cf. Firefox in Nixpkgs). + my $srcSize = @{$$srcNarFiles{$q}}[0]->{size}; + my $dstSize = @{$$dstNarFiles{$p}}[0]->{size}; + my $ratio = $srcSize / $dstSize; + $ratio = 1 / $ratio if $ratio < 1; + # print " SIZE $srcSize $dstSize $ratio $q\n"; + + if ($ratio >= 3) { + print " SKIPPING $q due to size ratio $ratio ($srcSize vs. $dstSize)\n"; + next; + } + + # If there are multiple matching names, include the + # ones with the closest version numbers. + my $dist = versionDiff $version, $version2; + if ($dist > $minDist) { + $minDist = $dist; + @closest = ($q); + $closestVersion = $version2; + } elsif ($dist == $minDist) { + push @closest, $q; + } + } + } + + if (scalar(@closest) == 0) { + print " NO BASE: $p\n"; + next; + } + + foreach my $closest (@closest) { + + # Generate a patch between $closest and $p. + print STDERR " $p <- $closest\n"; + + # If the patch already exists, skip it. + if (containsPatch($srcPatches, $p, $closest) || + containsPatch($dstPatches, $p, $closest)) + { + print " skipping, already exists\n"; + next; + } + + my $srcNarBz2 = getNarBz2 $narPath, $srcNarFiles, $closest; + my $dstNarBz2 = getNarBz2 $narPath, $dstNarFiles, $p; + + if (! -f $srcNarBz2) { + warn "patch source archive $srcNarBz2 is missing\n"; + next; + } + + system("@bunzip2@ < $srcNarBz2 > $tmpDir/A") == 0 + or die "cannot unpack $srcNarBz2"; + + if ((stat "$tmpDir/A")[7] >= $maxNarSize) { + print " skipping, source is too large\n"; + next; + } + + system("@bunzip2@ < $dstNarBz2 > $tmpDir/B") == 0 + or die "cannot unpack $dstNarBz2"; + + if ((stat "$tmpDir/B")[7] >= $maxNarSize) { + print " skipping, destination is too large\n"; + next; + } + + my $time1 = time(); + my $res = system("ulimit -t $timeLimit; @libexecdir@/bsdiff $tmpDir/A $tmpDir/B $tmpDir/DIFF"); + my $time2 = time(); + if ($res) { + warn "binary diff computation aborted after ", $time2 - $time1, " seconds\n"; + next; + } + + my $baseHash = `@bindir@/nix-hash --flat --type $hashAlgo --base32 $tmpDir/A` or die; + chomp $baseHash; + + my $narHash = `@bindir@/nix-hash --flat --type $hashAlgo --base32 $tmpDir/B` or die; + chomp $narHash; + + my $narDiffHash = `@bindir@/nix-hash --flat --type $hashAlgo --base32 $tmpDir/DIFF` or die; + chomp $narDiffHash; + + my $narDiffSize = (stat "$tmpDir/DIFF")[7]; + my $dstNarBz2Size = (stat $dstNarBz2)[7]; + + print " size $narDiffSize; full size $dstNarBz2Size; ", $time2 - $time1, " seconds\n"; + + if ($narDiffSize >= $dstNarBz2Size) { + print " rejecting; patch bigger than full archive\n"; + next; + } + + if ($narDiffSize / $dstNarBz2Size >= $maxPatchFraction) { + print " rejecting; patch too large relative to full archive\n"; + next; + } + + my $finalName = "$narDiffHash.nar-bsdiff"; + + if (-e "$patchesPath/$finalName") { + print " not copying, already exists\n"; + } + + else { + system("cp '$tmpDir/DIFF' '$patchesPath/$finalName.tmp'") == 0 + or die "cannot copy diff"; + rename("$patchesPath/$finalName.tmp", "$patchesPath/$finalName") + or die "cannot rename $patchesPath/$finalName.tmp"; + } + + # Add the patch to the manifest. + addPatch $dstPatches, $p, + { url => "$patchesURL/$finalName", hash => "$hashAlgo:$narDiffHash" + , size => $narDiffSize, basePath => $closest, baseHash => "$hashAlgo:$baseHash" + , narHash => "$hashAlgo:$narHash", patchType => "nar-bsdiff" + }; + } + } +} + + +# Propagate useful patches from $srcPatches to $dstPatches. A patch +# is useful if it produces either paths in the $dstNarFiles or paths +# that can be used as the base for other useful patches. +sub propagatePatches { + my ($srcPatches, $dstNarFiles, $dstPatches) = @_; + + print STDERR "propagating patches...\n"; + + my $changed; + do { + # !!! we repeat this to reach the transitive closure; inefficient + $changed = 0; + + print STDERR "loop\n"; + + my %dstBasePaths; + foreach my $q (keys %{$dstPatches}) { + foreach my $patch (@{$$dstPatches{$q}}) { + $dstBasePaths{$patch->{basePath}} = 1; + } + } + + foreach my $p (keys %{$srcPatches}) { + my $patchList = $$srcPatches{$p}; + + my $include = 0; + + # Is path $p included in the destination? If so, include + # patches that produce it. + $include = 1 if defined $$dstNarFiles{$p}; + + # Is path $p a path that serves as a base for paths in the + # destination? If so, include patches that produce it. + # !!! check baseHash + $include = 1 if defined $dstBasePaths{$p}; + + if ($include) { + foreach my $patch (@{$patchList}) { + $changed = 1 if addPatch $dstPatches, $p, $patch; + } + } + + } + + } while $changed; +} + + +# Add all new patches in $srcPatches to $dstPatches. +sub copyPatches { + my ($srcPatches, $dstPatches) = @_; + foreach my $p (keys %{$srcPatches}) { + addPatch $dstPatches, $p, $_ foreach @{$$srcPatches{$p}}; + } +} + + +return 1; diff --git a/scripts/Makefile.am b/scripts/Makefile.am index aa5d6f78c18a..60bb0a9b8105 100644 --- a/scripts/Makefile.am +++ b/scripts/Makefile.am @@ -1,23 +1,23 @@ bin_SCRIPTS = nix-collect-garbage \ nix-pull nix-push nix-prefetch-url \ nix-install-package nix-channel nix-build \ - nix-copy-closure + nix-copy-closure nix-generate-patches -noinst_SCRIPTS = nix-profile.sh generate-patches.pl \ +noinst_SCRIPTS = nix-profile.sh GeneratePatches.pm \ find-runtime-roots.pl build-remote.pl nix-reduce-build \ copy-from-other-stores.pl nix-http-export.cgi -nix-pull nix-push: readmanifest.pm readconfig.pm download-using-manifests.pl +nix-pull nix-push: NixManifest.pm NixConfig.pm download-using-manifests.pl -install-exec-local: readmanifest.pm download-using-manifests.pl copy-from-other-stores.pl find-runtime-roots.pl +install-exec-local: NixManifest.pm GeneratePatches.pm download-using-manifests.pl copy-from-other-stores.pl find-runtime-roots.pl $(INSTALL) -d $(DESTDIR)$(sysconfdir)/profile.d $(INSTALL_PROGRAM) nix-profile.sh $(DESTDIR)$(sysconfdir)/profile.d/nix.sh $(INSTALL) -d $(DESTDIR)$(libexecdir)/nix - $(INSTALL_DATA) readmanifest.pm $(DESTDIR)$(libexecdir)/nix - $(INSTALL_DATA) readconfig.pm $(DESTDIR)$(libexecdir)/nix - $(INSTALL_DATA) ssh.pm $(DESTDIR)$(libexecdir)/nix + $(INSTALL_DATA) NixManifest.pm $(DESTDIR)$(libexecdir)/nix + $(INSTALL_DATA) NixConfig.pm $(DESTDIR)$(libexecdir)/nix + $(INSTALL_DATA) SSH.pm $(DESTDIR)$(libexecdir)/nix + $(INSTALL_DATA) GeneratePatches.pm $(DESTDIR)$(libexecdir)/nix $(INSTALL_PROGRAM) find-runtime-roots.pl $(DESTDIR)$(libexecdir)/nix - $(INSTALL_PROGRAM) generate-patches.pl $(DESTDIR)$(libexecdir)/nix $(INSTALL_PROGRAM) build-remote.pl $(DESTDIR)$(libexecdir)/nix $(INSTALL) -d $(DESTDIR)$(libexecdir)/nix/substituters $(INSTALL_PROGRAM) download-using-manifests.pl $(DESTDIR)$(libexecdir)/nix/substituters @@ -30,15 +30,16 @@ EXTRA_DIST = nix-collect-garbage.in \ nix-pull.in nix-push.in nix-profile.sh.in \ nix-prefetch-url.in nix-install-package.in \ nix-channel.in \ - readmanifest.pm.in \ - readconfig.pm.in \ - ssh.pm \ + NixManifest.pm.in \ + NixConfig.pm.in \ + SSH.pm \ + GeneratePatches.pm.in \ nix-build.in \ download-using-manifests.pl.in \ copy-from-other-stores.pl.in \ - generate-patches.pl.in \ nix-copy-closure.in \ find-runtime-roots.pl.in \ build-remote.pl.in \ nix-reduce-build.in \ - nix-http-export.cgi.in + nix-http-export.cgi.in \ + nix-generate-patches.in diff --git a/scripts/readconfig.pm.in b/scripts/NixConfig.pm.in index aeb443aeed3a..aeb443aeed3a 100644 --- a/scripts/readconfig.pm.in +++ b/scripts/NixConfig.pm.in diff --git a/scripts/readmanifest.pm.in b/scripts/NixManifest.pm.in index 7244984ead6a..be0dda616922 100644 --- a/scripts/readmanifest.pm.in +++ b/scripts/NixManifest.pm.in @@ -33,18 +33,8 @@ sub readManifest { my $manifestVersion = 2; - my $storePath; - my $url; - my $hash; - my $size; - my $basePath; - my $baseHash; - my $patchType; - my $narHash; - my $references; - my $deriver; - my $hashAlgo; - my $copyFrom; + my ($storePath, $url, $hash, $size, $basePath, $baseHash, $patchType); + my ($narHash, $narSize, $references, $deriver, $hashAlgo, $copyFrom, $system); while (<MANIFEST>) { chomp; @@ -62,9 +52,11 @@ sub readManifest { undef $hash; undef $size; undef $narHash; + undef $narSize; undef $basePath; undef $baseHash; undef $patchType; + undef $system; $references = ""; $deriver = ""; $hashAlgo = "md5"; @@ -89,8 +81,10 @@ sub readManifest { if (!$found) { push @{$narFileList}, { url => $url, hash => $hash, size => $size - , narHash => $narHash, references => $references + , narHash => $narHash, narSize => $narSize + , references => $references , deriver => $deriver, hashAlgo => $hashAlgo + , system => $system }; } @@ -100,8 +94,8 @@ sub readManifest { addPatch $patches, $storePath, { url => $url, hash => $hash, size => $size , basePath => $basePath, baseHash => $baseHash - , narHash => $narHash, patchType => $patchType - , hashAlgo => $hashAlgo + , narHash => $narHash, narSize => $narSize + , patchType => $patchType, hashAlgo => $hashAlgo }; } @@ -132,9 +126,11 @@ sub readManifest { elsif (/^\s*BaseHash:\s*(\S+)\s*$/) { $baseHash = $1; } elsif (/^\s*Type:\s*(\S+)\s*$/) { $patchType = $1; } elsif (/^\s*NarHash:\s*(\S+)\s*$/) { $narHash = $1; } + elsif (/^\s*NarSize:\s*(\d+)\s*$/) { $narSize = $1; } elsif (/^\s*References:\s*(.*)\s*$/) { $references = $1; } elsif (/^\s*Deriver:\s*(\S+)\s*$/) { $deriver = $1; } elsif (/^\s*ManifestVersion:\s*(\d+)\s*$/) { $manifestVersion = $1; } + elsif (/^\s*System:\s*(\S+)\s*$/) { $system = $1; } # Compatibility; elsif (/^\s*NarURL:\s*(\S+)\s*$/) { $url = $1; } @@ -150,7 +146,7 @@ sub readManifest { sub writeManifest { - my ($manifest, $narFiles, $patches) = @_; + my ($manifest, $narFiles, $patches, $noCompress) = @_; open MANIFEST, ">$manifest.tmp"; # !!! check exclusive @@ -165,12 +161,14 @@ sub writeManifest { print MANIFEST " StorePath: $storePath\n"; print MANIFEST " NarURL: $narFile->{url}\n"; print MANIFEST " Hash: $narFile->{hash}\n" if defined $narFile->{hash}; - print MANIFEST " NarHash: $narFile->{narHash}\n"; print MANIFEST " Size: $narFile->{size}\n" if defined $narFile->{size}; + print MANIFEST " NarHash: $narFile->{narHash}\n"; + print MANIFEST " NarSize: $narFile->{narSize}\n" if $narFile->{narSize}; print MANIFEST " References: $narFile->{references}\n" if defined $narFile->{references} && $narFile->{references} ne ""; print MANIFEST " Deriver: $narFile->{deriver}\n" if defined $narFile->{deriver} && $narFile->{deriver} ne ""; + print MANIFEST " System: $narFile->{system}\n" if defined $narFile->{system}; print MANIFEST "}\n"; } } @@ -182,8 +180,9 @@ sub writeManifest { print MANIFEST " StorePath: $storePath\n"; print MANIFEST " NarURL: $patch->{url}\n"; print MANIFEST " Hash: $patch->{hash}\n"; - print MANIFEST " NarHash: $patch->{narHash}\n"; print MANIFEST " Size: $patch->{size}\n"; + print MANIFEST " NarHash: $patch->{narHash}\n"; + print MANIFEST " NarSize: $patch->{narSize}\n" if $patch->{narSize}; print MANIFEST " BasePath: $patch->{basePath}\n"; print MANIFEST " BaseHash: $patch->{baseHash}\n"; print MANIFEST " Type: $patch->{patchType}\n"; @@ -199,11 +198,13 @@ sub writeManifest { # Create a bzipped manifest. - system("@bzip2@ < $manifest > $manifest.bz2.tmp") == 0 - or die "cannot compress manifest"; + unless (defined $noCompress) { + system("@bzip2@ < $manifest > $manifest.bz2.tmp") == 0 + or die "cannot compress manifest"; - rename("$manifest.bz2.tmp", "$manifest.bz2") - or die "cannot rename $manifest.bz2.tmp: $!"; + rename("$manifest.bz2.tmp", "$manifest.bz2") + or die "cannot rename $manifest.bz2.tmp: $!"; + } } diff --git a/scripts/ssh.pm b/scripts/SSH.pm index c6d667a65d90..68f4a628b072 100644 --- a/scripts/ssh.pm +++ b/scripts/SSH.pm @@ -3,6 +3,8 @@ use File::Temp qw(tempdir); our @sshOpts = split ' ', ($ENV{"NIX_SSHOPTS"} or ""); +push @sshOpts, "-x"; + my $sshStarted = 0; my $sshHost; @@ -24,14 +26,17 @@ sub openSSHConnection { # child continues to run if we are killed. So instead make SSH # print "started" when it has established the connection, and wait # until we see that. - open SSH, "ssh $sshHost @sshOpts -M -N -o LocalCommand='echo started' -o PermitLocalCommand=yes |" or die; - while (<SSH>) { + open SSHPIPE, "ssh $sshHost @sshOpts -M -N -o LocalCommand='echo started' -o PermitLocalCommand=yes |" or die; + + while (<SSHPIPE>) { chomp; - last if /started/; + if ($_ eq "started") { + $sshStarted = 1; + return 1; + } } - - $sshStarted = 1; - return 1; + + return 0; } # Tell the master SSH client to exit. diff --git a/scripts/build-remote.pl.in b/scripts/build-remote.pl.in index c440b6a0f142..e943b0d9e304 100755 --- a/scripts/build-remote.pl.in +++ b/scripts/build-remote.pl.in @@ -3,7 +3,8 @@ use Fcntl ':flock'; use English '-no_match_vars'; use IO::Handle; -use ssh qw/sshOpts openSSHConnection/; +use SSH qw/sshOpts openSSHConnection/; +no warnings('once'); # General operation: @@ -31,57 +32,22 @@ $ENV{"DISPLAY"} = ""; $ENV{"SSH_ASKPASS"} = ""; -my $loadIncreased = 0; - -my ($amWilling, $localSystem, $neededSystem, $drvPath, $maxSilentTime) = @ARGV; -$maxSilentTime = 0 unless defined $maxSilentTime; - sub sendReply { my $reply = shift; print STDERR "# $reply\n"; } -sub decline { - sendReply "decline"; - exit 0; -} - -my $currentLoad = $ENV{"NIX_CURRENT_LOAD"}; -decline unless defined $currentLoad; -mkdir $currentLoad, 0777 or die unless -d $currentLoad; - -my $conf = $ENV{"NIX_REMOTE_SYSTEMS"}; -decline if !defined $conf || ! -e $conf; - -my $canBuildLocally = $amWilling && ($localSystem eq $neededSystem); - +sub all { $_ || return 0 for @_; 1 } -# Read the list of machines. -my @machines; -open CONF, "< $conf" or die; - -while (<CONF>) { - chomp; - s/\#.*$//g; - next if /^\s*$/; - /^\s*(\S+)\s+(\S+)\s+(\S+)\s+(\d+)(\s+([0-9\.]+))?\s*$/ or die; - push @machines, - { hostName => $1 - , systemTypes => [split(/,/, $2)] - , sshKeys => $3 - , maxJobs => $4 - , speedFactor => 1.0 * ($6 || 1) - , enabled => 1 - }; -} -close CONF; +# Initialisation. +my $loadIncreased = 0; +my ($localSystem, $maxSilentTime, $printBuildTrace) = @ARGV; +$maxSilentTime = 0 unless defined $maxSilentTime; -# Acquire the exclusive lock on $currentLoad/main-lock. -my $mainLock = "$currentLoad/main-lock"; -open MAINLOCK, ">>$mainLock" or die; -flock(MAINLOCK, LOCK_EX) or die; +my $currentLoad = $ENV{"NIX_CURRENT_LOAD"}; +my $conf = $ENV{"NIX_REMOTE_SYSTEMS"}; sub openSlotLock { @@ -91,150 +57,213 @@ sub openSlotLock { open $slotLock, ">>$slotLockFn" or die; return $slotLock; } - -my $hostName; -my $slotLock; -while (1) { +# Read the list of machines. +my @machines; +if (defined $conf && -e $conf) { + open CONF, "< $conf" or die; + while (<CONF>) { + chomp; + s/\#.*$//g; + next if /^\s*$/; + my @tokens = split /\s/, $_; + push @machines, + { hostName => $tokens[0] + , systemTypes => [ split(/,/, $tokens[1]) ] + , sshKeys => $tokens[2] + , maxJobs => int($tokens[3]) + , speedFactor => 1.0 * (defined $tokens[4] ? int($tokens[4]) : 1) + , features => [ split(/,/, $tokens[5] || "") ] + , enabled => 1 + }; + } + close CONF; +} + + + +# Wait for the calling process to ask us whether we can build some derivation. +my ($drvPath, $hostName, $slotLock); + +REQ: while (1) { + $_ = <STDIN> || exit 0; + my ($amWilling, $neededSystem); + ($amWilling, $neededSystem, $drvPath, $requiredFeatures) = split; + my @requiredFeatures = split /,/, $requiredFeatures; + + my $canBuildLocally = $amWilling && ($localSystem eq $neededSystem); + + if (!defined $currentLoad) { + sendReply "decline"; + next; + } + + # Acquire the exclusive lock on $currentLoad/main-lock. + mkdir $currentLoad, 0777 or die unless -d $currentLoad; + my $mainLock = "$currentLoad/main-lock"; + open MAINLOCK, ">>$mainLock" or die; + flock(MAINLOCK, LOCK_EX) or die; - # Find all machine that can execute this build, i.e., that support - # builds for the given platform and are not at their job limit. - my $rightType = 0; - my @available = (); - LOOP: foreach my $cur (@machines) { - if ($cur->{enabled} && grep { $neededSystem eq $_ } @{$cur->{systemTypes}}) { - $rightType = 1; - - # We have a machine of the right type. Determine the load on - # the machine. - my $slot = 0; - my $load = 0; - my $free; - while ($slot < $cur->{maxJobs}) { - my $slotLock = openSlotLock($cur, $slot); - if (flock($slotLock, LOCK_EX | LOCK_NB)) { - $free = $slot unless defined $free; - flock($slotLock, LOCK_UN) or die; - } else { - $load++; + + while (1) { + # Find all machine that can execute this build, i.e., that + # support builds for the given platform and features, and are + # not at their job limit. + my $rightType = 0; + my @available = (); + LOOP: foreach my $cur (@machines) { + if ($cur->{enabled} + && (grep { $neededSystem eq $_ } @{$cur->{systemTypes}}) + && all(map { my $f = $_; 0 != grep { $f eq $_ } @{$cur->{features}} } @requiredFeatures)) + { + $rightType = 1; + + # We have a machine of the right type. Determine the load on + # the machine. + my $slot = 0; + my $load = 0; + my $free; + while ($slot < $cur->{maxJobs}) { + my $slotLock = openSlotLock($cur, $slot); + if (flock($slotLock, LOCK_EX | LOCK_NB)) { + $free = $slot unless defined $free; + flock($slotLock, LOCK_UN) or die; + } else { + $load++; + } + close $slotLock; + $slot++; } - close $slotLock; - $slot++; + + push @available, { machine => $cur, load => $load, free => $free } + if $load < $cur->{maxJobs}; } - - push @available, { machine => $cur, load => $load, free => $free } - if $load < $cur->{maxJobs}; } - } - if (defined $ENV{NIX_DEBUG_HOOK}) { - print STDERR "load on " . $_->{machine}->{hostName} . " = " . $_->{load} . "\n" - foreach @available; - } + if (defined $ENV{NIX_DEBUG_HOOK}) { + print STDERR "load on " . $_->{machine}->{hostName} . " = " . $_->{load} . "\n" + foreach @available; + } - # Didn't find any available machine? Then decline or postpone. - if (scalar @available == 0) { - # Postpone if we have a machine of the right type, except if the - # local system can and wants to do the build. - if ($rightType && !$canBuildLocally) { - sendReply "postpone"; - exit 0; - } else { - decline; + # Didn't find any available machine? Then decline or postpone. + if (scalar @available == 0) { + # Postpone if we have a machine of the right type, except + # if the local system can and wants to do the build. + if ($rightType && !$canBuildLocally) { + sendReply "postpone"; + } else { + sendReply "decline"; + } + close MAINLOCK; + next REQ; } - } - - # Prioritise the available machines as follows: - # - First by load divided by speed factor, rounded to the nearest - # integer. This causes fast machines to be preferred over slow - # machines with similar loads. - # - Then by speed factor. - # - Finally by load. - sub lf { my $x = shift; return int($x->{load} / $x->{machine}->{speedFactor} + 0.4999); } - @available = sort - { lf($a) <=> lf($b) - || $b->{machine}->{speedFactor} <=> $a->{machine}->{speedFactor} - || $a->{load} <=> $b->{load} - } @available; + # Prioritise the available machines as follows: + # - First by load divided by speed factor, rounded to the nearest + # integer. This causes fast machines to be preferred over slow + # machines with similar loads. + # - Then by speed factor. + # - Finally by load. + sub lf { my $x = shift; return int($x->{load} / $x->{machine}->{speedFactor} + 0.4999); } + @available = sort + { lf($a) <=> lf($b) + || $b->{machine}->{speedFactor} <=> $a->{machine}->{speedFactor} + || $a->{load} <=> $b->{load} + } @available; - # Select the best available machine and lock a free slot. - my $selected = $available[0]; - my $machine = $selected->{machine}; - $slotLock = openSlotLock($machine, $selected->{free}); - flock($slotLock, LOCK_EX | LOCK_NB) or die; - utime undef, undef, $slotLock; + # Select the best available machine and lock a free slot. + my $selected = $available[0]; + my $machine = $selected->{machine}; + + $slotLock = openSlotLock($machine, $selected->{free}); + flock($slotLock, LOCK_EX | LOCK_NB) or die; + utime undef, undef, $slotLock; - close MAINLOCK; + close MAINLOCK; - # Connect to the selected machine. - @sshOpts = ("-i", $machine->{sshKeys}, "-x"); - $hostName = $machine->{hostName}; - last if openSSHConnection $hostName; + # Connect to the selected machine. + @sshOpts = ("-i", $machine->{sshKeys}, "-x"); + $hostName = $machine->{hostName}; + last REQ if openSSHConnection $hostName; - warn "unable to open SSH connection to $hostName, trying other available machines...\n"; - $machine->{enabled} = 0; + warn "unable to open SSH connection to $hostName, trying other available machines...\n"; + $machine->{enabled} = 0; + } } # Tell Nix we've accepted the build. sendReply "accept"; -my $x = <STDIN>; -chomp $x; - -if ($x ne "okay") { - exit 0; -} +my @inputs = split /\s/, readline(STDIN); +my @outputs = split /\s/, readline(STDIN); -# Do the actual build. print STDERR "building `$drvPath' on `$hostName'\n"; +print STDERR "@ build-remote $drvPath $hostName\n" if $printBuildTrace; -my $inputs = `cat inputs`; die if ($? != 0); -$inputs =~ s/\n/ /g; - -my $outputs = `cat outputs`; die if ($? != 0); -$outputs =~ s/\n/ /g; - -print "copying inputs...\n"; my $maybeSign = ""; $maybeSign = "--sign" if -e "/nix/etc/nix/signing-key.sec"; -system("NIX_SSHOPTS=\"@sshOpts\" @bindir@/nix-copy-closure $hostName $maybeSign $drvPath $inputs") == 0 + +# Register the derivation as a temporary GC root. Note that $PPID is +# the PID of the remote SSH process, which, due to the use of a +# persistant SSH connection, should be the same across all remote +# command invocations for this session. +my $rootsDir = "@localstatedir@/nix/gcroots/tmp"; +system("ssh $hostName @sshOpts 'mkdir -m 1777 -p $rootsDir; ln -sfn $drvPath $rootsDir/\$PPID.drv'"); + +sub removeRoots { + system("ssh $hostName @sshOpts 'rm -f $rootsDir/\$PPID.drv $rootsDir/\$PPID.out'"); +} + + +# Copy the derivation and its dependencies to the build machine. +system("NIX_SSHOPTS=\"@sshOpts\" @bindir@/nix-copy-closure $hostName $maybeSign $drvPath @inputs") == 0 or die "cannot copy inputs to $hostName: $?"; -print "building...\n"; - -my $buildFlags = "--max-silent-time $maxSilentTime --fallback"; - -# `-tt' forces allocation of a pseudo-terminal. This is required to -# make the remote nix-store process receive a signal when the -# connection dies. Without it, the remote process might continue to -# run indefinitely (that is, until it next tries to write to -# stdout/stderr). -if (system("ssh $hostName @sshOpts -tt 'nix-store -r $drvPath $buildFlags > /dev/null'") != 0) { - # If we couldn't run ssh or there was an ssh problem (indicated by - # exit code 255), then we return exit code 1; otherwise we assume - # that the builder failed, which we indicate to Nix using exit - # code 100. It's important to distinguish between the two because - # the first is a transient failure and the latter is permanent. - my $res = $? == -1 || ($? >> 8) == 255 ? 1 : 100; - print STDERR "build of `$drvPath' on `$hostName' failed with exit code $?\n"; + +# Perform the build. +my $buildFlags = "--max-silent-time $maxSilentTime --fallback --add-root $rootsDir/\$PPID.out --option verbosity 0"; + +# We let the remote side kill its process group when the connection is +# closed unexpectedly. This is necessary to ensure that no processes +# are left running on the remote system if the local Nix process is +# killed. (SSH itself doesn't kill child processes if the connection +# is interrupted unless the `-tt' flag is used to force a pseudo-tty, +# in which case every child receives SIGHUP; however, `-tt' doesn't +# work on some platforms when connection sharing is used.) +pipe STDIN, DUMMY; # make sure we have a readable STDIN +if (system("ssh $hostName @sshOpts '(read; kill -INT -\$\$) <&0 & nix-store -r $drvPath $buildFlags > /dev/null' 2>&4") != 0) { + # Note that if we get exit code 100 from `nix-store -r', it + # denotes a permanent build failure (as opposed to an SSH problem + # or a temporary Nix problem). We propagate this to the caller to + # allow it to distinguish between transient and permanent + # failures. + my $res = $? >> 8; + print STDERR "build of `$drvPath' on `$hostName' failed with exit code $res\n"; + removeRoots; exit $res; } -print "build of `$drvPath' on `$hostName' succeeded\n"; +#print "build of `$drvPath' on `$hostName' succeeded\n"; -foreach my $output (split '\n', $outputs) { + +# Copy the output from the build machine. +foreach my $output (@outputs) { my $maybeSignRemote = ""; $maybeSignRemote = "--sign" if $UID != 0; - system("ssh $hostName @sshOpts 'nix-store --export $maybeSignRemote $output' | @bindir@/nix-store --import > /dev/null") == 0 + system("ssh $hostName @sshOpts 'nix-store --export $maybeSignRemote $output'" . + "| NIX_HELD_LOCKS=$output @bindir@/nix-store --import > /dev/null") == 0 or die "cannot copy $output from $hostName: $?"; } + + +# Get rid of the temporary GC roots. +removeRoots; diff --git a/scripts/copy-from-other-stores.pl.in b/scripts/copy-from-other-stores.pl.in index 8f0ff4ca8df8..10130c0893ea 100644 --- a/scripts/copy-from-other-stores.pl.in +++ b/scripts/copy-from-other-stores.pl.in @@ -17,25 +17,19 @@ foreach my $dir (@remoteStoresAll) { } +$ENV{"NIX_REMOTE"} = ""; + + sub findStorePath { my $storePath = shift; - - my $storePathName = basename $storePath; - foreach my $store (@remoteStores) { - # Determine whether $storePath exists by looking for the - # existence of the info file, and if so, get store path info - # from that file. This rather breaks abstraction: we should - # be using `nix-store' for that. But right now there is no - # good way to tell nix-store to access a store mounted under a - # different location (there's $NIX_STORE, but that only works - # if the remote store is mounted under its "real" location). - my $infoFile = "$store/var/nix/db/info/$storePathName"; - my $storePath2 = "$store/store/$storePathName"; - if (-f $infoFile && -e $storePath2) { - return ($infoFile, $storePath2); - } + my $sourcePath = "$store/store/" . basename $storePath; + next unless -e $sourcePath || -l $sourcePath; + $ENV{"NIX_DB_DIR"} = "$store/var/nix/db"; + return ($store, $sourcePath) if + system("@bindir@/nix-store --check-validity $storePath") == 0; } + return undef; } @@ -46,37 +40,38 @@ if ($ARGV[0] eq "--query") { if ($cmd eq "have") { my $storePath = <STDIN>; chomp $storePath; - (my $infoFile) = findStorePath $storePath; - print STDOUT ($infoFile ? "1\n" : "0\n"); + print STDOUT (defined findStorePath($storePath) ? "1\n" : "0\n"); } elsif ($cmd eq "info") { my $storePath = <STDIN>; chomp $storePath; - (my $infoFile) = findStorePath $storePath; - if (!$infoFile) { + my ($store, $sourcePath) = findStorePath($storePath); + if (!defined $store) { print "0\n"; next; # not an error } print "1\n"; - my $deriver = ""; - my @references = (); - - open INFO, "<$infoFile" or die "cannot read info file $infoFile\n"; - while (<INFO>) { - chomp; - /^([\w-]+): (.*)$/ or die "bad info file"; - my $key = $1; - my $value = $2; - if ($key eq "Deriver") { $deriver = $value; } - elsif ($key eq "References") { @references = split ' ', $value; } - } - close INFO; + $ENV{"NIX_DB_DIR"} = "$store/var/nix/db"; + + my $deriver = `@bindir@/nix-store --query --deriver $storePath`; + die "cannot query deriver of `$storePath'" if $? != 0; + chomp $deriver; + $deriver = "" if $deriver eq "unknown-deriver"; + + my @references = split "\n", + `@bindir@/nix-store --query --references $storePath`; + die "cannot query references of `$storePath'" if $? != 0; + + my $narSize = `@bindir@/nix-store --query --size $storePath`; + die "cannot query size of `$storePath'" if $? != 0; + chomp $narSize; print "$deriver\n"; print scalar @references, "\n"; print "$_\n" foreach @references; - print "0\n"; # !!! showing size not supported (yet) + print "$narSize\n"; + print "$narSize\n"; } else { die "unknown command `$cmd'"; } @@ -87,8 +82,8 @@ if ($ARGV[0] eq "--query") { elsif ($ARGV[0] eq "--substitute") { die unless scalar @ARGV == 2; my $storePath = $ARGV[1]; - (my $infoFile, my $sourcePath) = findStorePath $storePath; - die unless $infoFile; + my ($store, $sourcePath) = findStorePath $storePath; + die unless $store; print "\n*** Copying `$storePath' from `$sourcePath'\n\n"; system("$binDir/nix-store --dump $sourcePath | $binDir/nix-store --restore $storePath") == 0 or die "cannot copy `$sourcePath' to `$storePath'"; diff --git a/scripts/download-using-manifests.pl.in b/scripts/download-using-manifests.pl.in index d48c7dd4bc3f..fe80bc11f3a7 100644 --- a/scripts/download-using-manifests.pl.in +++ b/scripts/download-using-manifests.pl.in @@ -1,7 +1,7 @@ #! @perl@ -w -I@libexecdir@/nix use strict; -use readmanifest; +use NixManifest; use POSIX qw(strftime); use File::Temp qw(tempdir); @@ -12,6 +12,10 @@ STDOUT->autoflush(1); my $manifestDir = ($ENV{"NIX_MANIFESTS_DIR"} or "@localstatedir@/nix/manifests"); my $logFile = "@localstatedir@/log/nix/downloads"; +# For queries, skip expensive calls to nix-hash etc. We're just +# estimating the expected download size. +my $fast = 1; + # Load all manifests. my %narFiles; @@ -31,6 +35,151 @@ for my $manifest (glob "$manifestDir/*.nixmanifest") { } +sub isValidPath { + my $p = shift; + if ($fast) { + return -e $p; + } else { + return system("$binDir/nix-store --check-validity '$p' 2> /dev/null") == 0; + } +} + + +sub parseHash { + my $hash = shift; + if ($hash =~ /^(.+):(.+)$/) { + return ($1, $2); + } else { + return ("md5", $hash); + } +} + + +# Compute the most efficient sequence of downloads to produce the +# given path. +sub computeSmallestDownload { + my $targetPath = shift; + + # Build a graph of all store paths that might contribute to the + # construction of $targetPath, and the special node "start". The + # edges are either patch operations, or downloads of full NAR + # files. The latter edges only occur between "start" and a store + # path. + my %graph; + + $graph{"start"} = {d => 0, pred => undef, edges => []}; + + my @queue = (); + my $queueFront = 0; + my %done; + + sub addNode { + my $graph = shift; + my $u = shift; + $$graph{$u} = {d => 999999999999, pred => undef, edges => []} + unless defined $$graph{$u}; + } + + sub addEdge { + my $graph = shift; + my $u = shift; + my $v = shift; + my $w = shift; + my $type = shift; + my $info = shift; + addNode $graph, $u; + push @{$$graph{$u}->{edges}}, + {weight => $w, start => $u, end => $v, type => $type, info => $info}; + my $n = scalar @{$$graph{$u}->{edges}}; + } + + push @queue, $targetPath; + + while ($queueFront < scalar @queue) { + my $u = $queue[$queueFront++]; + next if defined $done{$u}; + $done{$u} = 1; + + addNode \%graph, $u; + + # If the path already exists, it has distance 0 from the + # "start" node. + if (isValidPath($u)) { + addEdge \%graph, "start", $u, 0, "present", undef; + } + + else { + + # Add patch edges. + my $patchList = $patches{$u}; + foreach my $patch (@{$patchList}) { + if (isValidPath($patch->{basePath})) { + # !!! this should be cached + my ($baseHashAlgo, $baseHash) = parseHash $patch->{baseHash}; + my $format = "--base32"; + $format = "" if $baseHashAlgo eq "md5"; + my $hash = $fast && $baseHashAlgo eq "sha256" + ? `$binDir/nix-store -q --hash "$patch->{basePath}"` + : `$binDir/nix-hash --type '$baseHashAlgo' $format "$patch->{basePath}"`; + chomp $hash; + $hash =~ s/.*://; + next if $hash ne $baseHash; + } + push @queue, $patch->{basePath}; + addEdge \%graph, $patch->{basePath}, $u, $patch->{size}, "patch", $patch; + } + + # Add NAR file edges to the start node. + my $narFileList = $narFiles{$u}; + foreach my $narFile (@{$narFileList}) { + # !!! how to handle files whose size is not known in advance? + # For now, assume some arbitrary size (1 MB). + addEdge \%graph, "start", $u, ($narFile->{size} || 1000000), "narfile", $narFile; + } + } + } + + + # Run Dijkstra's shortest path algorithm to determine the shortest + # sequence of download and/or patch actions that will produce + # $targetPath. + + my @todo = keys %graph; + + while (scalar @todo > 0) { + + # Remove the closest element from the todo list. + # !!! inefficient, use a priority queue + @todo = sort { -($graph{$a}->{d} <=> $graph{$b}->{d}) } @todo; + my $u = pop @todo; + + my $u_ = $graph{$u}; + + foreach my $edge (@{$u_->{edges}}) { + my $v_ = $graph{$edge->{end}}; + if ($v_->{d} > $u_->{d} + $edge->{weight}) { + $v_->{d} = $u_->{d} + $edge->{weight}; + # Store the edge; to edge->start is actually the + # predecessor. + $v_->{pred} = $edge; + } + } + } + + + # Retrieve the shortest path from "start" to $targetPath. + my @path = (); + my $cur = $targetPath; + return () unless defined $graph{$targetPath}->{pred}; + while ($cur ne "start") { + push @path, $graph{$cur}->{pred}; + $cur = $graph{$cur}->{pred}->{start}; + } + + return @path; +} + + # Parse the arguments. if ($ARGV[0] eq "--query") { @@ -46,6 +195,7 @@ if ($ARGV[0] eq "--query") { elsif ($cmd eq "info") { my $storePath = <STDIN>; chomp $storePath; + my $info; if (defined $narFiles{$storePath}) { $info = @{$narFiles{$storePath}}[0]; @@ -57,13 +207,32 @@ if ($ARGV[0] eq "--query") { print "0\n"; next; # not an error } + print "1\n"; print "$info->{deriver}\n"; my @references = split " ", $info->{references}; print scalar @references, "\n"; print "$_\n" foreach @references; - my $size = $info->{size} || 0; - print "$size\n"; + + my @path = computeSmallestDownload $storePath; + + my $downloadSize = 0; + while (scalar @path > 0) { + my $edge = pop @path; + my $u = $edge->{start}; + my $v = $edge->{end}; + if ($edge->{type} eq "patch") { + $downloadSize += $edge->{info}->{size} || 0; + } + elsif ($edge->{type} eq "narfile") { + $downloadSize += $edge->{info}->{size} || 0; + } + } + + print "$downloadSize\n"; + + my $narSize = $info->{narSize} || 0; + print "$narSize\n"; } else { die "unknown command `$cmd'"; } @@ -79,6 +248,7 @@ elsif ($ARGV[0] ne "--substitute") { die unless scalar @ARGV == 2; my $targetPath = $ARGV[1]; +$fast = 0; # Create a temporary directory. @@ -110,148 +280,9 @@ foreach my $localPath (@{$localPathList}) { } -# Build a graph of all store paths that might contribute to the -# construction of $targetPath, and the special node "start". The -# edges are either patch operations, or downloads of full NAR files. -# The latter edges only occur between "start" and a store path. - -my %graph; - -$graph{"start"} = {d => 0, pred => undef, edges => []}; - -my @queue = (); -my $queueFront = 0; -my %done; - -sub addToQueue { - my $v = shift; - return if defined $done{$v}; - $done{$v} = 1; - push @queue, $v; -} - -sub addNode { - my $u = shift; - $graph{$u} = {d => 999999999999, pred => undef, edges => []} - unless defined $graph{$u}; -} - -sub addEdge { - my $u = shift; - my $v = shift; - my $w = shift; - my $type = shift; - my $info = shift; - addNode $u; - push @{$graph{$u}->{edges}}, - {weight => $w, start => $u, end => $v, type => $type, info => $info}; - my $n = scalar @{$graph{$u}->{edges}}; -} - -addToQueue $targetPath; - -sub isValidPath { - my $p = shift; - return system("$binDir/nix-store --check-validity '$p' 2> /dev/null") == 0; -} - -sub parseHash { - my $hash = shift; - if ($hash =~ /^(.+):(.+)$/) { - return ($1, $2); - } else { - return ("md5", $hash); - } -} - -while ($queueFront < scalar @queue) { - my $u = $queue[$queueFront++]; -# print "$u\n"; - - addNode $u; - - # If the path already exists, it has distance 0 from the "start" - # node. - if (isValidPath($u)) { - addEdge "start", $u, 0, "present", undef; - } - - else { - - # Add patch edges. - my $patchList = $patches{$u}; - foreach my $patch (@{$patchList}) { - if (isValidPath($patch->{basePath})) { - # !!! this should be cached - my ($baseHashAlgo, $baseHash) = parseHash $patch->{baseHash}; - my $format = "--base32"; - $format = "" if $baseHashAlgo eq "md5"; - my $hash = `$binDir/nix-hash --type '$baseHashAlgo' $format "$patch->{basePath}"`; - chomp $hash; - if ($hash ne $baseHash) { - print LOGFILE "$$ rejecting $patch->{basePath}\n"; - next; - } - } - addToQueue $patch->{basePath}; - addEdge $patch->{basePath}, $u, $patch->{size}, "patch", $patch; - } - - # Add NAR file edges to the start node. - my $narFileList = $narFiles{$u}; - foreach my $narFile (@{$narFileList}) { - # !!! how to handle files whose size is not known in advance? - # For now, assume some arbitrary size (1 MB). - addEdge "start", $u, ($narFile->{size} || 1000000), "narfile", $narFile; - if ($u eq $targetPath) { - my $size = $narFile->{size} || -1; - print LOGFILE "$$ full-download-would-be $size\n"; - } - } - - } -} - - -# Run Dijkstra's shortest path algorithm to determine the shortest -# sequence of download and/or patch actions that will produce -# $targetPath. - -sub byDistance { # sort by distance, reversed - return -($graph{$a}->{d} <=> $graph{$b}->{d}); -} - -my @todo = keys %graph; - -while (scalar @todo > 0) { - - # Remove the closest element from the todo list. - @todo = sort byDistance @todo; - my $u = pop @todo; - - my $u_ = $graph{$u}; - - foreach my $edge (@{$u_->{edges}}) { - my $v_ = $graph{$edge->{end}}; - if ($v_->{d} > $u_->{d} + $edge->{weight}) { - $v_->{d} = $u_->{d} + $edge->{weight}; - # Store the edge; to edge->start is actually the - # predecessor. - $v_->{pred} = $edge; - } - } -} - - -# Retrieve the shortest path from "start" to $targetPath. -my @path = (); -my $cur = $targetPath; -die "don't know how to produce $targetPath\n" - unless defined $graph{$targetPath}->{pred}; -while ($cur ne "start") { - push @path, $graph{$cur}->{pred}; - $cur = $graph{$cur}->{pred}->{start}; -} +# Compute the shortest path. +my @path = computeSmallestDownload $targetPath; +die "don't know how to produce $targetPath\n" if scalar @path == 0; # Traverse the shortest path, perform the actions described by the diff --git a/scripts/generate-patches.pl.in b/scripts/generate-patches.pl.in deleted file mode 100755 index dba6473508d3..000000000000 --- a/scripts/generate-patches.pl.in +++ /dev/null @@ -1,416 +0,0 @@ -#! @perl@ -w -I@libexecdir@/nix - -use strict; -use File::Temp qw(tempdir); -use readmanifest; - - -# Some patch generations options. - -# Max size of NAR archives to generate patches for. -my $maxNarSize = $ENV{"NIX_MAX_NAR_SIZE"}; -$maxNarSize = 100 * 1024 * 1024 if !defined $maxNarSize; - -# If patch is bigger than this fraction of full archive, reject. -my $maxPatchFraction = $ENV{"NIX_PATCH_FRACTION"}; -$maxPatchFraction = 0.60 if !defined $maxPatchFraction; - - -die unless scalar @ARGV == 5; - -my $hashAlgo = "sha256"; - -my $narDir = $ARGV[0]; -my $patchesDir = $ARGV[1]; -my $patchesURL = $ARGV[2]; -my $srcManifest = $ARGV[3]; -my $dstManifest = $ARGV[4]; - -my $tmpDir = tempdir("nix-generate-patches.XXXXXX", CLEANUP => 1, TMPDIR => 1) - or die "cannot create a temporary directory"; - -print "TEMP = $tmpDir\n"; - -#END { rmdir $tmpDir; } - -my %srcNarFiles; -my %srcLocalPaths; -my %srcPatches; - -my %dstNarFiles; -my %dstLocalPaths; -my %dstPatches; - -readManifest "$srcManifest", - \%srcNarFiles, \%srcLocalPaths, \%srcPatches; - -readManifest "$dstManifest", - \%dstNarFiles, \%dstLocalPaths, \%dstPatches; - - -sub findOutputPaths { - my $narFiles = shift; - - my %outPaths; - - foreach my $p (keys %{$narFiles}) { - - # Ignore derivations. - next if ($p =~ /\.drv$/); - - # Ignore builders (too much ambiguity -- they're all called - # `builder.sh'). - next if ($p =~ /\.sh$/); - next if ($p =~ /\.patch$/); - - # Don't bother including tar files etc. - next if ($p =~ /\.tar$/ || $p =~ /\.tar\.(gz|bz2|Z|lzma|xz)$/ || $p =~ /\.zip$/ || $p =~ /\.bin$/ || $p =~ /\.tgz$/ || $p =~ /\.rpm$/ || $p =~ /cvs-export$/ || $p =~ /fetchhg$/); - - $outPaths{$p} = 1; - } - - return %outPaths; -} - -print "finding src output paths...\n"; -my %srcOutPaths = findOutputPaths \%srcNarFiles; - -print "finding dst output paths...\n"; -my %dstOutPaths = findOutputPaths \%dstNarFiles; - - -sub getNameVersion { - my $p = shift; - $p =~ /\/[0-9a-z]+((?:-[a-zA-Z][^\/-]*)+)([^\/]*)$/; - my $name = $1; - my $version = $2; - return undef unless defined $name && defined $version; - $name =~ s/^-//; - $version =~ s/^-//; - return ($name, $version); -} - - -# A quick hack to get a measure of the `distance' between two -# versions: it's just the position of the first character that differs -# (or 999 if they are the same). -sub versionDiff { - my $s = shift; - my $t = shift; - my $i; - return 999 if $s eq $t; - for ($i = 0; $i < length $s; $i++) { - return $i if $i >= length $t or - substr($s, $i, 1) ne substr($t, $i, 1); - } - return $i; -} - - -sub getNarBz2 { - my $narFiles = shift; - my $storePath = shift; - - my $narFileList = $$narFiles{$storePath}; - die "missing path $storePath" unless defined $narFileList; - - my $narFile = @{$narFileList}[0]; - die unless defined $narFile; - - $narFile->{url} =~ /\/([^\/]+)$/; - die unless defined $1; - return "$narDir/$1"; -} - - -sub containsPatch { - my $patches = shift; - my $storePath = shift; - my $basePath = shift; - my $patchList = $$patches{$storePath}; - return 0 if !defined $patchList; - my $found = 0; - foreach my $patch (@{$patchList}) { - # !!! baseHash might differ - return 1 if $patch->{basePath} eq $basePath; - } - return 0; -} - - -# Compute the "weighted" number of uses of a path in the build graph. -sub computeUses { - my $narFiles = shift; - my $path = shift; - - # Find the deriver of $path. - return 1 unless defined $$narFiles{$path}; - my $deriver = @{$$narFiles{$path}}[0]->{deriver}; - return 1 unless defined $deriver && $deriver ne ""; - -# print " DERIVER $deriver\n"; - - # Optimisation: build the referrers graph from the references - # graph. - my %referrers; - foreach my $q (keys %{$narFiles}) { - my @refs = split " ", @{$$narFiles{$q}}[0]->{references}; - foreach my $r (@refs) { - $referrers{$r} = [] unless defined $referrers{$r}; - push @{$referrers{$r}}, $q; - } - } - - # Determine the shortest path from $deriver to all other reachable - # paths in the `referrers' graph. - - my %dist; - $dist{$deriver} = 0; - - my @queue = ($deriver); - my $pos = 0; - - while ($pos < scalar @queue) { - my $p = $queue[$pos]; - $pos++; - - foreach my $q (@{$referrers{$p}}) { - if (!defined $dist{$q}) { - $dist{$q} = $dist{$p} + 1; -# print " $q $dist{$q}\n"; - push @queue, $q; - } - } - } - - my $wuse = 1.0; - foreach my $user (keys %dist) { - next if $user eq $deriver; -# print " $user $dist{$user}\n"; - $wuse += 1.0 / 2.0**$dist{$user}; - } - -# print " XXX $path $wuse\n"; - - return $wuse; -} - - -# For each output path in the destination, see if we need to / can -# create a patch. - -print "creating patches...\n"; - -foreach my $p (keys %dstOutPaths) { - - # If exactly the same path already exists in the source, skip it. - next if defined $srcOutPaths{$p}; - - print " $p\n"; - - # If not, then we should find the paths in the source that are - # `most' likely to be present on a system that wants to install - # this path. - - (my $name, my $version) = getNameVersion $p; - next unless defined $name && defined $version; - - my @closest = (); - my $closestVersion; - my $minDist = -1; # actually, larger means closer - - # Find all source paths with the same name. - - foreach my $q (keys %srcOutPaths) { - (my $name2, my $version2) = getNameVersion $q; - next unless defined $name2 && defined $version2; - - if ($name eq $name2) { - - # If the sizes differ too much, then skip. This - # disambiguates between, e.g., a real component and a - # wrapper component (cf. Firefox in Nixpkgs). - my $srcSize = @{$srcNarFiles{$q}}[0]->{size}; - my $dstSize = @{$dstNarFiles{$p}}[0]->{size}; - my $ratio = $srcSize / $dstSize; - $ratio = 1 / $ratio if $ratio < 1; -# print " SIZE $srcSize $dstSize $ratio $q\n"; - - if ($ratio >= 3) { - print " SKIPPING $q due to size ratio $ratio ($srcSize $dstSize)\n"; - next; - } - - # If the numbers of weighted uses differ too much, then - # skip. This disambiguates between, e.g., the bootstrap - # GCC and the final GCC in Nixpkgs. -# my $srcUses = computeUses \%srcNarFiles, $q; -# my $dstUses = computeUses \%dstNarFiles, $p; -# $ratio = $srcUses / $dstUses; -# $ratio = 1 / $ratio if $ratio < 1; -# print " USE $srcUses $dstUses $ratio $q\n"; - -# if ($ratio >= 2) { -# print " SKIPPING $q due to use ratio $ratio ($srcUses $dstUses)\n"; -# next; -# } - - # If there are multiple matching names, include the ones - # with the closest version numbers. - my $dist = versionDiff $version, $version2; - if ($dist > $minDist) { - $minDist = $dist; - @closest = ($q); - $closestVersion = $version2; - } elsif ($dist == $minDist) { - push @closest, $q; - } - } - } - - if (scalar(@closest) == 0) { - print " NO BASE: $p\n"; - next; - } - - foreach my $closest (@closest) { - - # Generate a patch between $closest and $p. - print " $p <- $closest\n"; - - # If the patch already exists, skip it. - if (containsPatch(\%srcPatches, $p, $closest) || - containsPatch(\%dstPatches, $p, $closest)) - { - print " skipping, already exists\n"; - next; - } - -# next; - - my $srcNarBz2 = getNarBz2 \%srcNarFiles, $closest; - my $dstNarBz2 = getNarBz2 \%dstNarFiles, $p; - - if (! -f $srcNarBz2) { - warn "patch source archive $srcNarBz2 is missing\n"; - next; - } - - system("@bunzip2@ < $srcNarBz2 > $tmpDir/A") == 0 - or die "cannot unpack $srcNarBz2"; - - if ((stat "$tmpDir/A")[7] >= $maxNarSize) { - print " skipping, source is too large\n"; - next; - } - - system("@bunzip2@ < $dstNarBz2 > $tmpDir/B") == 0 - or die "cannot unpack $dstNarBz2"; - - if ((stat "$tmpDir/B")[7] >= $maxNarSize) { - print " skipping, destination is too large\n"; - next; - } - - system("@libexecdir@/bsdiff $tmpDir/A $tmpDir/B $tmpDir/DIFF") == 0 - or die "cannot compute binary diff"; - - my $baseHash = `@bindir@/nix-hash --flat --type $hashAlgo --base32 $tmpDir/A` or die; - chomp $baseHash; - - my $narHash = `@bindir@/nix-hash --flat --type $hashAlgo --base32 $tmpDir/B` or die; - chomp $narHash; - - my $narDiffHash = `@bindir@/nix-hash --flat --type $hashAlgo --base32 $tmpDir/DIFF` or die; - chomp $narDiffHash; - - my $narDiffSize = (stat "$tmpDir/DIFF")[7]; - my $dstNarBz2Size = (stat $dstNarBz2)[7]; - - print " size $narDiffSize; full size $dstNarBz2Size\n"; - - if ($narDiffSize >= $dstNarBz2Size) { - print " rejecting; patch bigger than full archive\n"; - next; - } - - if ($narDiffSize / $dstNarBz2Size >= $maxPatchFraction) { - print " rejecting; patch too large relative to full archive\n"; - next; - } - - my $finalName = - "$narDiffHash.nar-bsdiff"; - - if (-e "$patchesDir/$finalName") { - print " not copying, already exists\n"; - } - - else { - - system("cp '$tmpDir/DIFF' '$patchesDir/$finalName.tmp'") == 0 - or die "cannot copy diff"; - - rename("$patchesDir/$finalName.tmp", "$patchesDir/$finalName") - or die "cannot rename $patchesDir/$finalName.tmp"; - - } - - # Add the patch to the manifest. - addPatch \%dstPatches, $p, - { url => "$patchesURL/$finalName", hash => "$hashAlgo:$narDiffHash" - , size => $narDiffSize, basePath => $closest, baseHash => "$hashAlgo:$baseHash" - , narHash => "$hashAlgo:$narHash", patchType => "nar-bsdiff" - }, 0; - } -} - - -# Add in any potentially useful patches in the source (namely, those -# patches that produce either paths in the destination or paths that -# can be used as the base for other useful patches). - -print "propagating patches...\n"; - -my $changed; -do { - # !!! we repeat this to reach the transitive closure; inefficient - $changed = 0; - - print "loop\n"; - - my %dstBasePaths; - foreach my $q (keys %dstPatches) { - foreach my $patch (@{$dstPatches{$q}}) { - $dstBasePaths{$patch->{basePath}} = 1; - } - } - - foreach my $p (keys %srcPatches) { - my $patchList = $srcPatches{$p}; - - my $include = 0; - - # Is path $p included in the destination? If so, include - # patches that produce it. - $include = 1 if defined $dstNarFiles{$p}; - - # Is path $p a path that serves as a base for paths in the - # destination? If so, include patches that produce it. - # !!! check baseHash - $include = 1 if defined $dstBasePaths{$p}; - - if ($include) { - foreach my $patch (@{$patchList}) { - $changed = 1 if addPatch \%dstPatches, $p, $patch; - } - } - - } - -} while $changed; - - -# Rewrite the manifest of the destination (with the new patches). -writeManifest "${dstManifest}", - \%dstNarFiles, \%dstPatches; diff --git a/scripts/nix-build.in b/scripts/nix-build.in index ed85d5712128..f9d81b36c7a0 100644 --- a/scripts/nix-build.in +++ b/scripts/nix-build.in @@ -123,6 +123,11 @@ EOF $verbose = 1; } + elsif ($arg eq "--quiet") { + push @buildArgs, $arg; + push @instArgs, $arg; + } + elsif (substr($arg, 0, 1) eq "-") { push @buildArgs, $arg; } @@ -165,7 +170,7 @@ foreach my $expr (@exprs) { # Build. my @outPaths; - $pid = open(OUTPATHS, "-|") || exec "$binDir/nix-store", "--add-root", $outLink, "--indirect", "-rv", + $pid = open(OUTPATHS, "-|") || exec "$binDir/nix-store", "--add-root", $outLink, "--indirect", "-r", @buildArgs, @drvPaths; while (<OUTPATHS>) {chomp; push @outPaths, $_;} if (!close OUTPATHS) { diff --git a/scripts/nix-copy-closure.in b/scripts/nix-copy-closure.in index a477cc01a4cf..c037f003f0d0 100644 --- a/scripts/nix-copy-closure.in +++ b/scripts/nix-copy-closure.in @@ -1,6 +1,6 @@ #! @perl@ -w -I@libexecdir@/nix -use ssh; +use SSH; my $binDir = $ENV{"NIX_BIN_DIR"} || "@bindir@"; @@ -61,7 +61,7 @@ if ($toMode) { # Copy TO the remote machine. my @allStorePaths; # Get the closure of this path. - my $pid = open(READ, "$binDir/nix-store --query --requisites @storePaths|") or die; + my $pid = open(READ, "set -f; $binDir/nix-store --query --requisites @storePaths|") or die; while (<READ>) { chomp; @@ -73,7 +73,7 @@ if ($toMode) { # Copy TO the remote machine. # Ask the remote host which paths are invalid. - open(READ, "ssh $sshHost @sshOpts nix-store --check-validity --print-invalid @allStorePaths|"); + open(READ, "set -f; ssh $sshHost @sshOpts nix-store --check-validity --print-invalid @allStorePaths|"); my @missing = (); while (<READ>) { chomp; @@ -88,7 +88,7 @@ if ($toMode) { # Copy TO the remote machine. print STDERR " $_\n" foreach @missing; my $extraOpts = ""; $extraOpts .= "--sign" if $sign == 1; - system("nix-store --export $extraOpts @missing $compressor | ssh $sshHost @sshOpts '$decompressor nix-store --import'") == 0 + system("set -f; nix-store --export $extraOpts @missing $compressor | ssh $sshHost @sshOpts '$decompressor nix-store --import'") == 0 or die "copying store paths to remote machine `$sshHost' failed: $?"; } @@ -101,7 +101,7 @@ else { # Copy FROM the remote machine. # machine. Paths are assumed to be store paths; there is no # resolution (following of symlinks). my $pid = open(READ, - "ssh @sshOpts $sshHost nix-store --query --requisites @storePaths|") or die; + "set -f; ssh @sshOpts $sshHost nix-store --query --requisites @storePaths|") or die; my @allStorePaths; @@ -115,7 +115,7 @@ else { # Copy FROM the remote machine. # What paths are already valid locally? - open(READ, "@bindir@/nix-store --check-validity --print-invalid @allStorePaths|"); + open(READ, "set -f; @bindir@/nix-store --check-validity --print-invalid @allStorePaths|"); my @missing = (); while (<READ>) { chomp; @@ -130,7 +130,7 @@ else { # Copy FROM the remote machine. print STDERR " $_\n" foreach @missing; my $extraOpts = ""; $extraOpts .= "--sign" if $sign == 1; - system("ssh $sshHost @sshOpts 'nix-store --export $extraOpts @missing $compressor' | $decompressor @bindir@/nix-store --import") == 0 + system("set -f; ssh $sshHost @sshOpts 'nix-store --export $extraOpts @missing $compressor' | $decompressor @bindir@/nix-store --import") == 0 or die "copying store paths from remote machine `$sshHost' failed: $?"; } diff --git a/scripts/nix-generate-patches.in b/scripts/nix-generate-patches.in new file mode 100644 index 000000000000..c96cc704a1e8 --- /dev/null +++ b/scripts/nix-generate-patches.in @@ -0,0 +1,42 @@ +#! @perl@ -w -I@libexecdir@/nix + +use strict; +use File::Temp qw(tempdir); +use NixManifest; +use GeneratePatches; + +if (scalar @ARGV != 5) { + print STDERR <<EOF; +Usage: nix-generate-patches NAR-DIR PATCH-DIR PATCH-URI OLD-MANIFEST NEW-MANIFEST + +This command generates binary patches between NAR files listed in +OLD-MANIFEST and NEW-MANIFEST. The patches are written to the +directory PATCH-DIR, and the prefix PATCH-URI is used to generate URIs +for the patches. The patches are added to NEW-MANIFEST. All NARs are +required to exist in NAR-DIR. Patches are generated between +succeeding versions of packages with the same name. +EOF + exit 1; +} + +my $narPath = $ARGV[0]; +my $patchesPath = $ARGV[1]; +my $patchesURL = $ARGV[2]; +my $srcManifest = $ARGV[3]; +my $dstManifest = $ARGV[4]; + +my (%srcNarFiles, %srcLocalPaths, %srcPatches); +readManifest $srcManifest, \%srcNarFiles, \%srcLocalPaths, \%srcPatches; + +my (%dstNarFiles, %dstLocalPaths, %dstPatches); +readManifest $dstManifest, \%dstNarFiles, \%dstLocalPaths, \%dstPatches; + +my $tmpDir = tempdir("nix-generate-patches.XXXXXX", CLEANUP => 1, TMPDIR => 1) + or die "cannot create a temporary directory"; + +generatePatches \%srcNarFiles, \%dstNarFiles, \%srcPatches, \%dstPatches, + $narPath, $patchesPath, $patchesURL, $tmpDir; + +propagatePatches \%srcPatches, \%dstNarFiles, \%dstPatches; + +writeManifest $dstManifest, \%dstNarFiles, \%dstPatches; diff --git a/scripts/nix-pull.in b/scripts/nix-pull.in index e2a0cc1fafdf..f3b533ff7a4c 100644 --- a/scripts/nix-pull.in +++ b/scripts/nix-pull.in @@ -2,7 +2,7 @@ use strict; use File::Temp qw(tempdir); -use readmanifest; +use NixManifest; my $tmpDir = tempdir("nix-pull.XXXXXX", CLEANUP => 1, TMPDIR => 1) or die "cannot create a temporary directory"; diff --git a/scripts/nix-push.in b/scripts/nix-push.in index 38097f740162..1d8ba86a8d0c 100644 --- a/scripts/nix-push.in +++ b/scripts/nix-push.in @@ -2,7 +2,7 @@ use strict; use File::Temp qw(tempdir); -use readmanifest; +use NixManifest; my $hashAlgo = "sha256"; @@ -172,12 +172,6 @@ for (my $n = 0; $n < scalar @storePaths; $n++) { $narbz2Hash =~ /^[0-9a-z]+$/ or die "invalid hash"; close HASH; - open HASH, "$narDir/nar-hash" or die "cannot open nar-hash"; - my $narHash = <HASH>; - chomp $narHash; - $narHash =~ /^[0-9a-z]+$/ or die "invalid hash"; - close HASH; - my $narName = "$narbz2Hash.nar.bz2"; my $narFile = "$narDir/$narName"; @@ -195,6 +189,14 @@ for (my $n = 0; $n < scalar @storePaths; $n++) { chomp $deriver; $deriver = "" if $deriver eq "unknown-deriver"; + my $narHash = `$binDir/nix-store --query --hash '$storePath'`; + die "cannot query hash for `$storePath'" if $? != 0; + chomp $narHash; + + my $narSize = `$binDir/nix-store --query --size '$storePath'`; + die "cannot query size for `$storePath'" if $? != 0; + chomp $narSize; + my $url; if ($localCopy) { $url = "$targetArchivesUrl/$narName"; @@ -205,7 +207,8 @@ for (my $n = 0; $n < scalar @storePaths; $n++) { { url => $url , hash => "$hashAlgo:$narbz2Hash" , size => $narbz2Size - , narHash => "$hashAlgo:$narHash" + , narHash => "$narHash" + , narSize => $narSize , references => $references , deriver => $deriver } diff --git a/src/bin2c/bin2c.c b/src/bin2c/bin2c.c index 18bf81d69e25..5ed8a57082fd 100644 --- a/src/bin2c/bin2c.c +++ b/src/bin2c/bin2c.c @@ -14,10 +14,10 @@ int main(int argc, char * * argv) { int c; if (argc != 2) abort(); - print("static unsigned char %s[] = {", argv[1]); + print("static unsigned char %s[] = { ", argv[1]); while ((c = getchar()) != EOF) { print("0x%02x, ", (unsigned char) c); } - print("};\n"); + print("0 };\n"); return 0; } diff --git a/src/bsdiff-4.3/bsdiff.c b/src/bsdiff-4.3/bsdiff.c index 150a7f79c488..374ed038fa1f 100644 --- a/src/bsdiff-4.3/bsdiff.c +++ b/src/bsdiff-4.3/bsdiff.c @@ -277,6 +277,7 @@ int main(int argc,char *argv[]) for(scsc=scan+=len;scan<newsize;scan++) { len=search(I,old,oldsize,new+scan,newsize-scan, 0,oldsize,&pos); + if (len > 64 * 1024) break; for(;scsc<scan+len;scsc++) if((scsc+lastoffset<oldsize) && diff --git a/src/libexpr/primops.cc b/src/libexpr/primops.cc index a4812de06519..3e955ea3fe09 100644 --- a/src/libexpr/primops.cc +++ b/src/libexpr/primops.cc @@ -965,7 +965,7 @@ static void prim_substring(EvalState & state, Value * * args, Value & v) if (start < 0) throw EvalError("negative start position in `substring'"); - mkString(v, string(s, start, len), context); + mkString(v, start >= s.size() ? "" : string(s, start, len), context); } diff --git a/src/libmain/shared.cc b/src/libmain/shared.cc index 29fc13e33627..68f145820361 100644 --- a/src/libmain/shared.cc +++ b/src/libmain/shared.cc @@ -54,25 +54,26 @@ void printGCWarning() void printMissing(const PathSet & paths) { - unsigned long long downloadSize; + unsigned long long downloadSize, narSize; PathSet willBuild, willSubstitute, unknown; - queryMissing(paths, willBuild, willSubstitute, unknown, downloadSize); + queryMissing(paths, willBuild, willSubstitute, unknown, downloadSize, narSize); if (!willBuild.empty()) { - printMsg(lvlInfo, format("the following derivations will be built:")); + printMsg(lvlInfo, format("these derivations will be built:")); foreach (PathSet::iterator, i, willBuild) printMsg(lvlInfo, format(" %1%") % *i); } if (!willSubstitute.empty()) { - printMsg(lvlInfo, format("the following paths will be downloaded/copied (%.2f MiB):") % - (downloadSize / (1024.0 * 1024.0))); + printMsg(lvlInfo, format("these paths will be downloaded/copied (%.2f MiB download, %.2f MiB unpacked):") + % (downloadSize / (1024.0 * 1024.0)) + % (narSize / (1024.0 * 1024.0))); foreach (PathSet::iterator, i, willSubstitute) printMsg(lvlInfo, format(" %1%") % *i); } if (!unknown.empty()) { - printMsg(lvlInfo, format("don't know how to build the following paths%1%:") + printMsg(lvlInfo, format("don't know how to build these paths%1%:") % (readOnlyMode ? " (may be caused by read-only store access)" : "")); foreach (PathSet::iterator, i, unknown) printMsg(lvlInfo, format(" %1%") % *i); @@ -200,17 +201,16 @@ static void initAndRun(int argc, char * * argv) remaining.clear(); /* Process default options. */ + int verbosityDelta = 0; for (Strings::iterator i = args.begin(); i != args.end(); ++i) { string arg = *i; - if (arg == "--verbose" || arg == "-v") - verbosity = (Verbosity) ((int) verbosity + 1); + if (arg == "--verbose" || arg == "-v") verbosityDelta++; + else if (arg == "--quiet") verbosityDelta--; else if (arg == "--log-type") { ++i; if (i == args.end()) throw UsageError("`--log-type' requires an argument"); setLogType(*i); } - else if (arg == "--build-output" || arg == "-B") - ; /* !!! obsolete - remove eventually */ else if (arg == "--no-build-output" || arg == "-Q") buildVerbosity = lvlVomit; else if (arg == "--print-build-trace") @@ -251,6 +251,9 @@ static void initAndRun(int argc, char * * argv) else remaining.push_back(arg); } + verbosityDelta += queryIntSetting("verbosity", lvlInfo); + verbosity = (Verbosity) (verbosityDelta < 0 ? 0 : verbosityDelta); + /* Automatically clean up the temporary roots file when we exit. */ RemoveTempRoots removeTempRoots __attribute__((unused)); @@ -390,7 +393,7 @@ int main(int argc, char * * argv) printMsg(lvlError, format("error: %1%%2%") % (showTrace ? e.prefix() : "") % e.msg()); if (e.prefix() != "" && !showTrace) printMsg(lvlError, "(use `--show-trace' to show detailed location information)"); - return 1; + return e.status; } catch (std::exception & e) { printMsg(lvlError, format("error: %1%") % e.what()); return 1; diff --git a/src/libmain/shared.hh b/src/libmain/shared.hh index f70f6893b4bd..c99810c78a20 100644 --- a/src/libmain/shared.hh +++ b/src/libmain/shared.hh @@ -1,7 +1,7 @@ #ifndef __SHARED_H #define __SHARED_H -#include "types.hh" +#include "util.hh" #include <signal.h> diff --git a/src/libstore/Makefile.am b/src/libstore/Makefile.am index 9accc3005fc3..e19256b925ea 100644 --- a/src/libstore/Makefile.am +++ b/src/libstore/Makefile.am @@ -10,7 +10,14 @@ pkginclude_HEADERS = \ globals.hh references.hh pathlocks.hh \ worker-protocol.hh -libstore_la_LIBADD = ../libutil/libutil.la ../boost/format/libformat.la @ADDITIONAL_NETWORK_LIBS@ +libstore_la_LIBADD = ../libutil/libutil.la ../boost/format/libformat.la ${aterm_lib} ${sqlite_lib} + +EXTRA_DIST = schema.sql AM_CXXFLAGS = -Wall \ - -I$(srcdir)/.. -I$(srcdir)/../libutil + ${sqlite_include} -I$(srcdir)/.. -I$(srcdir)/../libutil + +local-store.lo: schema.sql.hh + +%.sql.hh: %.sql + ../bin2c/bin2c schema < $< > $@ || (rm $@ && exit 1) diff --git a/src/libstore/build.cc b/src/libstore/build.cc index ef2f7adf3194..83bd6754a61a 100644 --- a/src/libstore/build.cc +++ b/src/libstore/build.cc @@ -64,6 +64,7 @@ static const uid_t rootUserId = 0; /* Forward definition. */ class Worker; +class HookInstance; /* A pointer to a goal. */ @@ -213,8 +214,14 @@ public: bool cacheFailure; + /* Set if at least one derivation had a BuildError (i.e. permanent + failure). */ + bool permanentFailure; + LocalStore & store; + boost::shared_ptr<HookInstance> hook; + Worker(LocalStore & store); ~Worker(); @@ -263,7 +270,8 @@ public: /* Wait for input to become available. */ void waitForInput(); - + + unsigned int exitStatus(); }; @@ -615,6 +623,107 @@ void deletePathWrapped(const Path & path) ////////////////////////////////////////////////////////////////////// +struct HookInstance +{ + /* Pipes for talking to the build hook. */ + Pipe toHook; + + /* Pipe for the hook's standard output/error. */ + Pipe fromHook; + + /* Pipe for the builder's standard output/error. */ + Pipe builderOut; + + /* The process ID of the hook. */ + Pid pid; + + HookInstance(); + + ~HookInstance(); +}; + + +HookInstance::HookInstance() +{ + debug("starting build hook"); + + Path buildHook = absPath(getEnv("NIX_BUILD_HOOK")); + + /* Create a pipe to get the output of the child. */ + fromHook.create(); + + /* Create the communication pipes. */ + toHook.create(); + + /* Create a pipe to get the output of the builder. */ + builderOut.create(); + + /* Fork the hook. */ + pid = fork(); + switch (pid) { + + case -1: + throw SysError("unable to fork"); + + case 0: + try { /* child */ + + commonChildInit(fromHook); + + if (chdir("/") == -1) throw SysError("changing into `/"); + + /* Dup the communication pipes. */ + toHook.writeSide.close(); + if (dup2(toHook.readSide, STDIN_FILENO) == -1) + throw SysError("dupping to-hook read side"); + + /* Use fd 4 for the builder's stdout/stderr. */ + builderOut.readSide.close(); + if (dup2(builderOut.writeSide, 4) == -1) + throw SysError("dupping builder's stdout/stderr"); + + execl(buildHook.c_str(), buildHook.c_str(), thisSystem.c_str(), + (format("%1%") % maxSilentTime).str().c_str(), + (format("%1%") % printBuildTrace).str().c_str(), + NULL); + + throw SysError(format("executing `%1%'") % buildHook); + + } catch (std::exception & e) { + std::cerr << format("build hook error: %1%") % e.what() << std::endl; + } + quickExit(1); + } + + /* parent */ + pid.setSeparatePG(true); + pid.setKillSignal(SIGTERM); + fromHook.writeSide.close(); + toHook.readSide.close(); +} + + +HookInstance::~HookInstance() +{ + try { + /* Cleanly shut down the hook by closing its stdin if it's not + already building. Otherwise pid's destructor will kill + it. */ + if (pid != -1 && toHook.writeSide != -1) { + toHook.writeSide.close(); + pid.wait(true); + } + } catch (...) { + ignoreException(); + } +} + + +////////////////////////////////////////////////////////////////////// + + +typedef enum {rpAccept, rpDecline, rpPostpone} HookReply; + class DerivationGoal : public Goal { private: @@ -649,14 +758,11 @@ private: AutoCloseFD fdLogFile; /* Pipe for the builder's standard output/error. */ - Pipe logPipe; - - /* Whether we're building using a build hook. */ - bool usingBuildHook; - - /* Pipes for talking to the build hook (if any). */ - Pipe toHook; + Pipe builderOut; + /* The build hook. */ + boost::shared_ptr<HookInstance> hook; + /* Whether we're currently doing a chroot build. */ bool useChroot; @@ -694,12 +800,8 @@ private: void buildDone(); /* Is the build hook willing to perform the build? */ - typedef enum {rpAccept, rpDecline, rpPostpone} HookReply; HookReply tryBuildHook(); - /* Synchronously wait for a build hook to finish. */ - void terminateBuildHook(bool kill = false); - /* Start building a derivation. */ void startBuilder(); @@ -711,10 +813,6 @@ private: /* Open a log file and a pipe to it. */ Path openLogFile(); - /* Common initialisation to be performed in child processes (i.e., - both in builders and in build hooks). */ - void initChild(); - /* Delete the temporary directory, if we have one. */ void deleteTmpDir(bool force); @@ -742,6 +840,7 @@ DerivationGoal::DerivationGoal(const Path & drvPath, Worker & worker) trace("created"); } + DerivationGoal::~DerivationGoal() { /* Careful: we should never ever throw an exception from a @@ -754,6 +853,7 @@ DerivationGoal::~DerivationGoal() } } + void DerivationGoal::killChild() { if (pid != -1) { @@ -778,6 +878,8 @@ void DerivationGoal::killChild() assert(pid == -1); } + + hook.reset(); } @@ -887,7 +989,10 @@ void DerivationGoal::outputsSubstituted() foreach (PathSet::iterator, i, drv.inputSrcs) addWaitee(worker.makeSubstitutionGoal(*i)); - state = &DerivationGoal::inputsRealised; + if (waitees.empty()) /* to prevent hang (no wake-up event) */ + inputsRealised(); + else + state = &DerivationGoal::inputsRealised; } @@ -961,6 +1066,16 @@ PathSet outputPaths(const DerivationOutputs & outputs) } +static bool canBuildLocally(const string & platform) +{ + return platform == thisSystem +#ifdef CAN_DO_LINUX32_BUILDS + || (platform == "i686-linux" && thisSystem == "x86_64-linux") +#endif + ; +} + + void DerivationGoal::tryToBuild() { trace("trying to build"); @@ -1028,28 +1143,36 @@ void DerivationGoal::tryToBuild() foreach (DerivationOutputs::iterator, i, drv.outputs) if (pathFailed(i->second.path)) return; + /* Don't do a remote build if the derivation has the attribute + `preferLocalBuild' set. */ + bool preferLocalBuild = + drv.env["preferLocalBuild"] == "1" && canBuildLocally(drv.platform); + /* Is the build hook willing to accept this job? */ - usingBuildHook = true; - switch (tryBuildHook()) { - case rpAccept: - /* Yes, it has started doing so. Wait until we get EOF - from the hook. */ - state = &DerivationGoal::buildDone; - return; - case rpPostpone: - /* Not now; wait until at least one child finishes. */ - worker.waitForAWhile(shared_from_this()); - outputLocks.unlock(); - return; - case rpDecline: - /* We should do it ourselves. */ - break; + if (!preferLocalBuild) { + switch (tryBuildHook()) { + case rpAccept: + /* Yes, it has started doing so. Wait until we get + EOF from the hook. */ + state = &DerivationGoal::buildDone; + return; + case rpPostpone: + /* Not now; wait until at least one child finishes or + the wake-up timeout expires. */ + worker.waitForAWhile(shared_from_this()); + outputLocks.unlock(); + return; + case rpDecline: + /* We should do it ourselves. */ + break; + } } - - usingBuildHook = false; - - /* Make sure that we are allowed to start a build. */ - if (worker.getNrLocalBuilds() >= maxBuildJobs) { + + /* Make sure that we are allowed to start a build. If this + derivation prefers to be done locally, do it even if + maxBuildJobs is 0. */ + unsigned int curBuilds = worker.getNrLocalBuilds(); + if (curBuilds >= maxBuildJobs && !(preferLocalBuild && curBuilds == 0)) { worker.waitForBuildSlot(shared_from_this()); outputLocks.unlock(); return; @@ -1067,6 +1190,7 @@ void DerivationGoal::tryToBuild() if (printBuildTrace) printMsg(lvlError, format("@ build-failed %1% %2% %3% %4%") % drvPath % drv.outputs["out"].path % 0 % e.msg()); + worker.permanentFailure = true; amDone(ecFailed); return; } @@ -1085,18 +1209,29 @@ void DerivationGoal::buildDone() to have terminated. In fact, the builder could also have simply have closed its end of the pipe --- just don't do that :-) */ - /* !!! this could block! security problem! solution: kill the - child */ - pid_t savedPid = pid; - int status = pid.wait(true); + int status; + pid_t savedPid; + if (hook) { + savedPid = hook->pid; + status = hook->pid.wait(true); + } else { + /* !!! this could block! security problem! solution: kill the + child */ + savedPid = pid; + status = pid.wait(true); + } debug(format("builder process for `%1%' finished") % drvPath); /* So the child is gone now. */ worker.childTerminated(savedPid); - + /* Close the read side of the logger pipe. */ - logPipe.readSide.close(); + if (hook) { + hook->builderOut.readSide.close(); + hook->fromHook.readSide.close(); + } + else builderOut.readSide.close(); /* Close the log file. */ fdLogFile.close(); @@ -1169,11 +1304,11 @@ void DerivationGoal::buildDone() /* When using a build hook, the hook will return a remote build failure using exit code 100. Anything else is a hook problem. */ - bool hookError = usingBuildHook && + bool hookError = hook && (!WIFEXITED(status) || WEXITSTATUS(status) != 100); if (printBuildTrace) { - if (usingBuildHook && hookError) + if (hook && hookError) printMsg(lvlError, format("@ hook-failed %1% %2% %3% %4%") % drvPath % drv.outputs["out"].path % status % e.msg()); else @@ -1192,6 +1327,7 @@ void DerivationGoal::buildDone() foreach (DerivationOutputs::iterator, i, drv.outputs) worker.store.registerFailedPath(i->second.path); + worker.permanentFailure = !hookError && !fixedOutput; amDone(ecFailed); return; } @@ -1208,162 +1344,85 @@ void DerivationGoal::buildDone() } -DerivationGoal::HookReply DerivationGoal::tryBuildHook() +HookReply DerivationGoal::tryBuildHook() { - if (!useBuildHook) return rpDecline; - Path buildHook = getEnv("NIX_BUILD_HOOK"); - if (buildHook == "") return rpDecline; - buildHook = absPath(buildHook); + if (!useBuildHook || getEnv("NIX_BUILD_HOOK") == "") return rpDecline; - /* Create a directory where we will store files used for - communication between us and the build hook. */ - tmpDir = createTempDir(); - - /* Create the log file and pipe. */ - Path logFile = openLogFile(); + if (!worker.hook) + worker.hook = boost::shared_ptr<HookInstance>(new HookInstance); - /* Create the communication pipes. */ - toHook.create(); - - /* Fork the hook. */ - pid = fork(); - switch (pid) { - - case -1: - throw SysError("unable to fork"); - - case 0: - try { /* child */ - - initChild(); - - string s; - foreach (DerivationOutputs::const_iterator, i, drv.outputs) - s += i->second.path + " "; - if (setenv("NIX_HELD_LOCKS", s.c_str(), 1)) - throw SysError("setting an environment variable"); - - execl(buildHook.c_str(), buildHook.c_str(), - (worker.getNrLocalBuilds() < maxBuildJobs ? (string) "1" : "0").c_str(), - thisSystem.c_str(), - drv.platform.c_str(), - drvPath.c_str(), - (format("%1%") % maxSilentTime).str().c_str(), - NULL); - - throw SysError(format("executing `%1%'") % buildHook); - - } catch (std::exception & e) { - std::cerr << format("build hook error: %1%") % e.what() << std::endl; - } - quickExit(1); - } - - /* parent */ - pid.setSeparatePG(true); - pid.setKillSignal(SIGTERM); - logPipe.writeSide.close(); - worker.childStarted(shared_from_this(), - pid, singleton<set<int> >(logPipe.readSide), false, false); + /* Tell the hook about system features (beyond the system type) + required from the build machine. (The hook could parse the + drv file itself, but this is easier.) */ + Strings features = tokenizeString(drv.env["requiredSystemFeatures"]); + foreach (Strings::iterator, i, features) checkStoreName(*i); /* !!! abuse */ - toHook.readSide.close(); + /* Send the request to the hook. */ + writeLine(worker.hook->toHook.writeSide, (format("%1% %2% %3% %4%") + % (worker.getNrLocalBuilds() < maxBuildJobs ? "1" : "0") + % drv.platform % drvPath % concatStringsSep(",", features)).str()); /* Read the first line of input, which should be a word indicating whether the hook wishes to perform the build. */ string reply; - try { - while (true) { - string s = readLine(logPipe.readSide); - if (string(s, 0, 2) == "# ") { - reply = string(s, 2); - break; - } - handleChildOutput(logPipe.readSide, s + "\n"); + while (true) { + string s = readLine(worker.hook->fromHook.readSide); + if (string(s, 0, 2) == "# ") { + reply = string(s, 2); + break; } - } catch (Error & e) { - terminateBuildHook(true); - throw; + s += "\n"; + writeToStderr((unsigned char *) s.c_str(), s.size()); } debug(format("hook reply is `%1%'") % reply); - if (reply == "decline" || reply == "postpone") { - /* Clean up the child. !!! hacky / should verify */ - terminateBuildHook(); + if (reply == "decline" || reply == "postpone") return reply == "decline" ? rpDecline : rpPostpone; - } + else if (reply != "accept") + throw Error(format("bad hook reply `%1%'") % reply); - else if (reply == "accept") { + printMsg(lvlTalkative, format("using hook to build path(s) %1%") + % showPaths(outputPaths(drv.outputs))); - printMsg(lvlInfo, format("using hook to build path(s) %1%") - % showPaths(outputPaths(drv.outputs))); - - /* Write the information that the hook needs to perform the - build, i.e., the set of input paths, the set of output - paths, and the references (pointer graph) in the input - paths. */ + hook = worker.hook; + worker.hook.reset(); - Path inputListFN = tmpDir + "/inputs"; - Path outputListFN = tmpDir + "/outputs"; - Path referencesFN = tmpDir + "/references"; - - /* The `inputs' file lists all inputs that have to be copied - to the remote system. This unfortunately has to contain - the entire derivation closure to ensure that the validity - invariant holds on the remote system. (I.e., it's - unfortunate that we have to list it since the remote system - *probably* already has it.) */ - PathSet allInputs; - allInputs.insert(inputPaths.begin(), inputPaths.end()); - computeFSClosure(drvPath, allInputs); + /* Tell the hook all the inputs that have to be copied to the + remote system. This unfortunately has to contain the entire + derivation closure to ensure that the validity invariant holds + on the remote system. (I.e., it's unfortunate that we have to + list it since the remote system *probably* already has it.) */ + PathSet allInputs; + allInputs.insert(inputPaths.begin(), inputPaths.end()); + computeFSClosure(drvPath, allInputs); - string s; - foreach (PathSet::iterator, i, allInputs) s += *i + "\n"; + string s; + foreach (PathSet::iterator, i, allInputs) s += *i + " "; + writeLine(hook->toHook.writeSide, s); - writeFile(inputListFN, s); - - /* The `outputs' file lists all outputs that have to be copied - from the remote system. */ - s = ""; - foreach (DerivationOutputs::iterator, i, drv.outputs) - s += i->second.path + "\n"; - writeFile(outputListFN, s); - - /* The `references' file has exactly the format accepted by - `nix-store --register-validity'. */ - writeFile(referencesFN, - makeValidityRegistration(allInputs, true, false)); + /* Tell the hooks the outputs that have to be copied back from the + remote system. */ + s = ""; + foreach (DerivationOutputs::iterator, i, drv.outputs) + s += i->second.path + " "; + writeLine(hook->toHook.writeSide, s); + + hook->toHook.writeSide.close(); - /* Tell the hook to proceed. */ - writeLine(toHook.writeSide, "okay"); - toHook.writeSide.close(); + /* Create the log file and pipe. */ + Path logFile = openLogFile(); - if (printBuildTrace) - printMsg(lvlError, format("@ build-started %1% %2% %3% %4%") - % drvPath % drv.outputs["out"].path % drv.platform % logFile); + set<int> fds; + fds.insert(hook->fromHook.readSide); + fds.insert(hook->builderOut.readSide); + worker.childStarted(shared_from_this(), hook->pid, fds, false, false); + + if (printBuildTrace) + printMsg(lvlError, format("@ build-started %1% %2% %3% %4%") + % drvPath % drv.outputs["out"].path % drv.platform % logFile); - return rpAccept; - } - - else throw Error(format("bad hook reply `%1%'") % reply); -} - - -void DerivationGoal::terminateBuildHook(bool kill) -{ - debug("terminating build hook"); - pid_t savedPid = pid; - if (kill) - pid.kill(); - else - pid.wait(true); - /* `false' means don't wake up waiting goals, since we want to - keep this build slot ourselves. */ - worker.childTerminated(savedPid, false); - toHook.writeSide.close(); - fdLogFile.close(); - logPipe.readSide.close(); - deleteTmpDir(true); /* get rid of the hook's temporary directory */ + return rpAccept; } @@ -1380,11 +1439,7 @@ void DerivationGoal::startBuilder() format("building path(s) %1%") % showPaths(outputPaths(drv.outputs))) /* Right platform? */ - if (drv.platform != thisSystem -#ifdef CAN_DO_LINUX32_BUILDS - && !(drv.platform == "i686-linux" && thisSystem == "x86_64-linux") -#endif - ) + if (!canBuildLocally(drv.platform)) throw Error( format("a `%1%' is required to build `%3%', but I am a `%2%'") % drv.platform % thisSystem % drvPath); @@ -1499,7 +1554,7 @@ void DerivationGoal::startBuilder() /* Write closure info to `fileName'. */ writeFile(tmpDir + "/" + fileName, - makeValidityRegistration(paths, false, false)); + worker.store.makeValidityRegistration(paths, false, false)); } @@ -1549,6 +1604,9 @@ void DerivationGoal::startBuilder() if (fixedOutput) useChroot = false; + /* Hack to allow derivations to disable chroot builds. */ + if (drv.env["__noChroot"] == "1") useChroot = false; + if (useChroot) { #if CHROOT_ENABLED /* Create a temporary directory in which we set up the chroot @@ -1572,7 +1630,7 @@ void DerivationGoal::startBuilder() /* Create a /etc/passwd with entries for the build user and the nobody account. The latter is kind of a hack to support - Samba-in-QEMU. */ + Samba-in-QEMU. */ createDirs(chrootRootDir + "/etc"); writeFile(chrootRootDir + "/etc/passwd", @@ -1580,13 +1638,13 @@ void DerivationGoal::startBuilder() "nixbld:x:%1%:%2%:Nix build user:/:/noshell\n" "nobody:x:65534:65534:Nobody:/:/noshell\n") % (buildUser.enabled() ? buildUser.getUID() : getuid()) - % (buildUser.enabled() ? buildUser.getGID() : getgid())).str()); + % (buildUser.enabled() ? buildUser.getGID() : getgid())).str()); /* Declare the build user's group so that programs get a consistent - view of the system (e.g., "id -gn"). */ - writeFile(chrootRootDir + "/etc/group", - (format("nixbld:!:%1%:\n") - % (buildUser.enabled() ? buildUser.getGID() : getgid())).str()); + view of the system (e.g., "id -gn"). */ + writeFile(chrootRootDir + "/etc/group", + (format("nixbld:!:%1%:\n") + % (buildUser.enabled() ? buildUser.getGID() : getgid())).str()); /* Bind-mount a user-configurable set of directories from the host file system. The `/dev/pts' directory must be mounted @@ -1645,9 +1703,12 @@ void DerivationGoal::startBuilder() printMsg(lvlChatty, format("executing builder `%1%'") % drv.builder); - /* Create the log file and pipe. */ + /* Create the log file. */ Path logFile = openLogFile(); + /* Create a pipe to get the output of the builder. */ + builderOut.create(); + /* Fork a child to build the package. Note that while we currently use forks to run and wait for the children, it shouldn't be hard to use threads for this on systems where @@ -1661,7 +1722,7 @@ void DerivationGoal::startBuilder() case 0: /* Warning: in the child we should absolutely not make any - Berkeley DB calls! */ + SQLite calls! */ try { /* child */ @@ -1688,18 +1749,23 @@ void DerivationGoal::startBuilder() throw SysError(format("bind mount from `%1%' to `%2%' failed") % source % target); } - /* Do the chroot(). initChild() will do a chdir() to - the temporary build directory to make sure the - current directory is in the chroot. (Actually the - order doesn't matter, since due to the bind mount - tmpDir and tmpRootDit/tmpDir are the same - directories.) */ + /* Do the chroot(). Below we do a chdir() to the + temporary build directory to make sure the current + directory is in the chroot. (Actually the order + doesn't matter, since due to the bind mount tmpDir + and tmpRootDit/tmpDir are the same directories.) */ if (chroot(chrootRootDir.c_str()) == -1) throw SysError(format("cannot change root directory to `%1%'") % chrootRootDir); } #endif - initChild(); + commonChildInit(builderOut); + + if (chdir(tmpDir.c_str()) == -1) + throw SysError(format("changing into `%1%'") % tmpDir); + + /* Close all other file descriptors. */ + closeMostFDs(set<int>()); #ifdef CAN_DO_LINUX32_BUILDS if (drv.platform == "i686-linux" && thisSystem == "x86_64-linux") { @@ -1720,10 +1786,10 @@ void DerivationGoal::startBuilder() /* If we are running in `build-users' mode, then switch to the user we allocated above. Make sure that we drop - all root privileges. Note that initChild() above has - closed all file descriptors except std*, so that's - safe. Also note that setuid() when run as root sets - the real, effective and saved UIDs. */ + all root privileges. Note that above we have closed + all file descriptors except std*, so that's safe. Also + note that setuid() when run as root sets the real, + effective and saved UIDs. */ if (buildUser.enabled()) { printMsg(lvlChatty, format("switching to user `%1%'") % buildUser.getUser()); @@ -1777,9 +1843,9 @@ void DerivationGoal::startBuilder() /* parent */ pid.setSeparatePG(true); - logPipe.writeSide.close(); + builderOut.writeSide.close(); worker.childStarted(shared_from_this(), pid, - singleton<set<int> >(logPipe.readSide), true, true); + singleton<set<int> >(builderOut.readSide), true, true); if (printBuildTrace) { printMsg(lvlError, format("@ build-started %1% %2% %3% %4%") @@ -1811,12 +1877,12 @@ PathSet parseReferenceSpecifiers(const Derivation & drv, string attr) void DerivationGoal::computeClosure() { map<Path, PathSet> allReferences; - map<Path, Hash> contentHashes; + map<Path, HashResult> contentHashes; /* When using a build hook, the build hook can register the output as valid (by doing `nix-store --import'). If so we don't have to do anything here. */ - if (usingBuildHook) { + if (hook) { bool allValid = true; foreach (DerivationOutputs::iterator, i, drv.outputs) if (!worker.store.isValidPath(i->second.path)) allValid = false; @@ -1868,7 +1934,7 @@ void DerivationGoal::computeClosure() if (ht == htUnknown) throw BuildError(format("unknown hash algorithm `%1%'") % algo); Hash h = parseHash(ht, i->second.hash); - Hash h2 = recursive ? hashPath(ht, path) : hashFile(ht, path); + Hash h2 = recursive ? hashPath(ht, path).first : hashFile(ht, path); if (h != h2) throw BuildError( format("output path `%1%' should have %2% hash `%3%', instead has `%4%'") @@ -1882,7 +1948,7 @@ void DerivationGoal::computeClosure() contained in it. Compute the SHA-256 NAR hash at the same time. The hash is stored in the database so that we can verify later on whether nobody has messed with the store. */ - Hash hash; + HashResult hash; PathSet references = scanForReferences(path, allPaths, hash); contentHashes[path] = hash; @@ -1911,14 +1977,18 @@ void DerivationGoal::computeClosure() } /* Register each output path as valid, and register the sets of - paths referenced by each of them. !!! this should be - atomic so that either all paths are registered as valid, or - none are. */ - foreach (DerivationOutputs::iterator, i, drv.outputs) - worker.store.registerValidPath(i->second.path, - contentHashes[i->second.path], - allReferences[i->second.path], - drvPath); + paths referenced by each of them. */ + ValidPathInfos infos; + foreach (DerivationOutputs::iterator, i, drv.outputs) { + ValidPathInfo info; + info.path = i->second.path; + info.hash = contentHashes[i->second.path].first; + info.narSize = contentHashes[i->second.path].second; + info.references = allReferences[i->second.path]; + info.deriver = drvPath; + infos.push_back(info); + } + worker.store.registerValidPaths(infos); /* It is now safe to delete the lock files, since all future lockers will see that the output paths are valid; they will not @@ -1944,32 +2014,10 @@ Path DerivationGoal::openLogFile() if (fdLogFile == -1) throw SysError(format("creating log file `%1%'") % logFileName); - /* Create a pipe to get the output of the child. */ - logPipe.create(); - return logFileName; } -void DerivationGoal::initChild() -{ - commonChildInit(logPipe); - - if (chdir(tmpDir.c_str()) == -1) - throw SysError(format("changing into `%1%'") % tmpDir); - - /* When running a hook, dup the communication pipes. */ - if (usingBuildHook) { - toHook.writeSide.close(); - if (dup2(toHook.readSide, STDIN_FILENO) == -1) - throw SysError("dupping to-hook read side"); - } - - /* Close all other file descriptors. */ - closeMostFDs(set<int>()); -} - - void DerivationGoal::deleteTmpDir(bool force) { if (tmpDir != "") { @@ -1989,19 +2037,22 @@ void DerivationGoal::deleteTmpDir(bool force) void DerivationGoal::handleChildOutput(int fd, const string & data) { - if (fd == logPipe.readSide) { + if ((hook && fd == hook->builderOut.readSide) || + (!hook && fd == builderOut.readSide)) + { if (verbosity >= buildVerbosity) writeToStderr((unsigned char *) data.c_str(), data.size()); writeFull(fdLogFile, (unsigned char *) data.c_str(), data.size()); } - else abort(); + if (hook && fd == hook->fromHook.readSide) + writeToStderr((unsigned char *) data.c_str(), data.size()); } void DerivationGoal::handleEOF(int fd) { - if (fd == logPipe.readSide) worker.wakeUp(shared_from_this()); + worker.wakeUp(shared_from_this()); } @@ -2345,10 +2396,15 @@ void SubstitutionGoal::finished() canonicalisePathMetaData(storePath); - Hash contentHash = hashPath(htSHA256, storePath); - - worker.store.registerValidPath(storePath, contentHash, - info.references, info.deriver); + HashResult hash = hashPath(htSHA256, storePath); + + ValidPathInfo info2; + info2.path = storePath; + info2.hash = hash.first; + info2.narSize = hash.second; + info2.references = info.references; + info2.deriver = info.deriver; + worker.store.registerValidPath(info2); outputLock->setDeletion(true); @@ -2395,6 +2451,7 @@ Worker::Worker(LocalStore & store) nrLocalBuilds = 0; lastWokenUp = 0; cacheFailure = queryBoolSetting("build-cache-failure", false); + permanentFailure = false; } @@ -2721,6 +2778,11 @@ void Worker::waitForInput() } +unsigned int Worker::exitStatus() +{ + return permanentFailure ? 100 : 1; +} + ////////////////////////////////////////////////////////////////////// @@ -2747,7 +2809,7 @@ void LocalStore::buildDerivations(const PathSet & drvPaths) } if (!failed.empty()) - throw Error(format("build of %1% failed") % showPaths(failed)); + throw Error(format("build of %1% failed") % showPaths(failed), worker.exitStatus()); } @@ -2763,7 +2825,7 @@ void LocalStore::ensurePath(const Path & path) worker.run(goals); if (goal->getExitCode() != Goal::ecSuccess) - throw Error(format("path `%1%' does not exist and cannot be created") % path); + throw Error(format("path `%1%' does not exist and cannot be created") % path, worker.exitStatus()); } diff --git a/src/libstore/derivations.hh b/src/libstore/derivations.hh index 95e49d42c9ea..c14be48afb8e 100644 --- a/src/libstore/derivations.hh +++ b/src/libstore/derivations.hh @@ -1,10 +1,10 @@ #ifndef __DERIVATIONS_H #define __DERIVATIONS_H -#include "hash.hh" - #include <map> +#include "types.hh" + namespace nix { diff --git a/src/libstore/gc.cc b/src/libstore/gc.cc index f58f691c99dd..b8395bfc4352 100644 --- a/src/libstore/gc.cc +++ b/src/libstore/gc.cc @@ -1,6 +1,5 @@ #include "globals.hh" #include "misc.hh" -#include "pathlocks.hh" #include "local-store.hh" #include <boost/shared_ptr.hpp> @@ -31,7 +30,7 @@ static const int defaultGcLevel = 1000; read. To be precise: when they try to create a new temporary root file, they will block until the garbage collector has finished / yielded the GC lock. */ -static int openGCLock(LockType lockType) +int LocalStore::openGCLock(LockType lockType) { Path fnGCLock = (format("%1%/%2%") % nixStateDir % gcLockName).str(); @@ -127,7 +126,7 @@ Path addPermRoot(const Path & _storePath, const Path & _gcRoot, Instead of reading all the roots, it would be more efficient to check if the root is in a directory in or linked from the gcroots directory. */ - if (queryBoolSetting("gc-check-reachability", true)) { + if (queryBoolSetting("gc-check-reachability", false)) { Roots roots = store->findRoots(); if (roots.find(gcRoot) == roots.end()) printMsg(lvlError, @@ -136,7 +135,7 @@ Path addPermRoot(const Path & _storePath, const Path & _gcRoot, "therefore, `%2%' might be removed by the garbage collector") % gcRoot % storePath); } - + /* Grab the global GC root, causing us to block while a GC is in progress. This prevents the set of permanent roots from increasing while a GC is in progress. */ @@ -416,18 +415,13 @@ struct LocalStore::GCState PathSet busy; bool gcKeepOutputs; bool gcKeepDerivations; - - bool drvsIndexed; - typedef std::multimap<string, Path> DrvsByName; - DrvsByName drvsByName; // derivation paths hashed by name attribute - - GCState(GCResults & results_) : results(results_), drvsIndexed(false) + GCState(GCResults & results_) : results(results_) { } }; -static bool doDelete(GCOptions::GCAction action) +static bool shouldDelete(GCOptions::GCAction action) { return action == GCOptions::gcDeleteDead || action == GCOptions::gcDeleteSpecific; @@ -441,45 +435,11 @@ bool LocalStore::isActiveTempFile(const GCState & state, && state.tempRoots.find(string(path, 0, path.size() - suffix.size())) != state.tempRoots.end(); } - -/* Return all the derivations in the Nix store that have `path' as an - output. This function assumes that derivations have the same name - as their outputs. */ -PathSet LocalStore::findDerivers(GCState & state, const Path & path) -{ - PathSet derivers; - - Path deriver = queryDeriver(path); - if (deriver != "") derivers.insert(deriver); - - if (!state.drvsIndexed) { - Paths entries = readDirectory(nixStore); - foreach (Paths::iterator, i, entries) - if (isDerivation(*i)) - state.drvsByName.insert(std::pair<string, Path>( - getNameOfStorePath(*i), nixStore + "/" + *i)); - state.drvsIndexed = true; - } - - string name = getNameOfStorePath(path); - - // Urgh, I should have used Haskell... - std::pair<GCState::DrvsByName::iterator, GCState::DrvsByName::iterator> range = - state.drvsByName.equal_range(name); - - for (GCState::DrvsByName::iterator i = range.first; i != range.second; ++i) - if (isValidPath(i->second)) { - Derivation drv = derivationFromPath(i->second); - foreach (DerivationOutputs::iterator, j, drv.outputs) - if (j->second.path == path) derivers.insert(i->second); - } - - return derivers; -} - bool LocalStore::tryToDelete(GCState & state, const Path & path) { + checkInterrupt(); + if (!pathExists(path)) return true; if (state.deleted.find(path) != state.deleted.end()) return true; if (state.live.find(path) != state.live.end()) return false; @@ -508,10 +468,10 @@ bool LocalStore::tryToDelete(GCState & state, const Path & path) then don't delete the derivation if any of the outputs are live. */ if (state.gcKeepDerivations && isDerivation(path)) { - Derivation drv = derivationFromPath(path); - foreach (DerivationOutputs::iterator, i, drv.outputs) - if (!tryToDelete(state, i->second.path)) { - printMsg(lvlDebug, format("cannot delete derivation `%1%' because its output is alive") % path); + PathSet outputs = queryDerivationOutputs(path); + foreach (PathSet::iterator, i, outputs) + if (!tryToDelete(state, *i)) { + printMsg(lvlDebug, format("cannot delete derivation `%1%' because its output `%2%' is alive") % path % *i); goto isLive; } } @@ -522,18 +482,9 @@ bool LocalStore::tryToDelete(GCState & state, const Path & path) if (!pathExists(path)) return true; /* If gc-keep-outputs is set, then don't delete this path if - its deriver is not garbage. !!! Nix does not reliably - store derivers, so we have to look at all derivations to - determine which of them derive `path'. Since this makes - the garbage collector very slow to start on large Nix - stores, here we just look for all derivations that have the - same name as `path' (where the name is the part of the - filename after the hash, i.e. the `name' attribute of the - derivation). This is somewhat hacky: currently, the - deriver of a path always has the same name as the output, - but this might change in the future. */ + there are derivers of this path that are not garbage. */ if (state.gcKeepOutputs) { - PathSet derivers = findDerivers(state, path); + PathSet derivers = queryValidDerivers(path); foreach (PathSet::iterator, deriver, derivers) { /* Break an infinite recursion if gc-keep-derivations and gc-keep-outputs are both set by tentatively @@ -567,7 +518,7 @@ bool LocalStore::tryToDelete(GCState & state, const Path & path) } /* The path is garbage, so delete it. */ - if (doDelete(state.options.action)) { + if (shouldDelete(state.options.action)) { printMsg(lvlInfo, format("deleting `%1%'") % path); unsigned long long bytesFreed, blocksFreed; @@ -613,6 +564,15 @@ void LocalStore::collectGarbage(const GCOptions & options, GCResults & results) state.gcKeepOutputs = queryBoolSetting("gc-keep-outputs", false); state.gcKeepDerivations = queryBoolSetting("gc-keep-derivations", true); + + /* Using `--ignore-liveness' with `--delete' can have unintended + consequences if `gc-keep-outputs' or `gc-keep-derivations' are + true (the garbage collector will recurse into deleting the + outputs or derivers, respectively). So disable them. */ + if (options.action == GCOptions::gcDeleteSpecific && options.ignoreLiveness) { + state.gcKeepOutputs = false; + state.gcKeepDerivations = false; + } /* Acquire the global GC root. This prevents a) New roots from being added. @@ -667,7 +627,7 @@ void LocalStore::collectGarbage(const GCOptions & options, GCResults & results) vector<Path> entries_(entries.begin(), entries.end()); random_shuffle(entries_.begin(), entries_.end()); - if (doDelete(state.options.action)) + if (shouldDelete(state.options.action)) printMsg(lvlError, format("deleting garbage...")); else printMsg(lvlError, format("determining live/dead paths...")); diff --git a/src/libstore/globals.cc b/src/libstore/globals.cc index 75d2f69c2b72..7069d104aae4 100644 --- a/src/libstore/globals.cc +++ b/src/libstore/globals.cc @@ -20,7 +20,7 @@ string nixBinDir = "/UNINIT"; bool keepFailed = false; bool keepGoing = false; bool tryFallback = false; -Verbosity buildVerbosity = lvlInfo; +Verbosity buildVerbosity = lvlError; unsigned int maxBuildJobs = 1; unsigned int buildCores = 1; bool readOnlyMode = false; diff --git a/src/libstore/local-store.cc b/src/libstore/local-store.cc index f430492fd407..6af34cc77814 100644 --- a/src/libstore/local-store.cc +++ b/src/libstore/local-store.cc @@ -4,6 +4,7 @@ #include "archive.hh" #include "pathlocks.hh" #include "worker-protocol.hh" +#include "derivations.hh" #include <iostream> #include <algorithm> @@ -15,11 +16,166 @@ #include <fcntl.h> #include <errno.h> #include <stdio.h> +#include <time.h> + +#include <sqlite3.h> namespace nix { + +MakeError(SQLiteError, Error); +MakeError(SQLiteBusy, SQLiteError); + + +static void throwSQLiteError(sqlite3 * db, const format & f) + __attribute__ ((noreturn)); + +static void throwSQLiteError(sqlite3 * db, const format & f) +{ + int err = sqlite3_errcode(db); + if (err == SQLITE_BUSY) { + printMsg(lvlError, "warning: SQLite database is busy"); + /* Sleep for a while since retrying the transaction right away + is likely to fail again. */ +#if HAVE_NANOSLEEP + struct timespec t; + t.tv_sec = 0; + t.tv_nsec = 100 * 1000 * 1000; /* 0.1s */ + nanosleep(&t, 0); +#else + sleep(1); +#endif + throw SQLiteBusy(format("%1%: %2%") % f.str() % sqlite3_errmsg(db)); + } + else + throw SQLiteError(format("%1%: %2%") % f.str() % sqlite3_errmsg(db)); +} + + +SQLite::~SQLite() +{ + try { + if (db && sqlite3_close(db) != SQLITE_OK) + throwSQLiteError(db, "closing database"); + } catch (...) { + ignoreException(); + } +} + + +void SQLiteStmt::create(sqlite3 * db, const string & s) +{ + checkInterrupt(); + assert(!stmt); + if (sqlite3_prepare_v2(db, s.c_str(), -1, &stmt, 0) != SQLITE_OK) + throwSQLiteError(db, "creating statement"); + this->db = db; +} + + +void SQLiteStmt::reset() +{ + assert(stmt); + /* Note: sqlite3_reset() returns the error code for the most + recent call to sqlite3_step(). So ignore it. */ + sqlite3_reset(stmt); + curArg = 1; +} + + +SQLiteStmt::~SQLiteStmt() +{ + try { + if (stmt && sqlite3_finalize(stmt) != SQLITE_OK) + throwSQLiteError(db, "finalizing statement"); + } catch (...) { + ignoreException(); + } +} + + +void SQLiteStmt::bind(const string & value) +{ + if (sqlite3_bind_text(stmt, curArg++, value.c_str(), -1, SQLITE_TRANSIENT) != SQLITE_OK) + throwSQLiteError(db, "binding argument"); +} + + +void SQLiteStmt::bind(int value) +{ + if (sqlite3_bind_int(stmt, curArg++, value) != SQLITE_OK) + throwSQLiteError(db, "binding argument"); +} + + +void SQLiteStmt::bind64(long long value) +{ + if (sqlite3_bind_int64(stmt, curArg++, value) != SQLITE_OK) + throwSQLiteError(db, "binding argument"); +} + + +void SQLiteStmt::bind() +{ + if (sqlite3_bind_null(stmt, curArg++) != SQLITE_OK) + throwSQLiteError(db, "binding argument"); +} + + +/* Helper class to ensure that prepared statements are reset when + leaving the scope that uses them. Unfinished prepared statements + prevent transactions from being aborted, and can cause locks to be + kept when they should be released. */ +struct SQLiteStmtUse +{ + SQLiteStmt & stmt; + SQLiteStmtUse(SQLiteStmt & stmt) : stmt(stmt) + { + stmt.reset(); + } + ~SQLiteStmtUse() + { + try { + stmt.reset(); + } catch (...) { + ignoreException(); + } + } +}; + + +struct SQLiteTxn +{ + bool active; + sqlite3 * db; + + SQLiteTxn(sqlite3 * db) : active(false) { + this->db = db; + if (sqlite3_exec(db, "begin;", 0, 0, 0) != SQLITE_OK) + throwSQLiteError(db, "starting transaction"); + active = true; + } + + void commit() + { + if (sqlite3_exec(db, "commit;", 0, 0, 0) != SQLITE_OK) + throwSQLiteError(db, "committing transaction"); + active = false; + } + ~SQLiteTxn() + { + try { + if (active && sqlite3_exec(db, "rollback;", 0, 0, 0) != SQLITE_OK) + throwSQLiteError(db, "aborting transaction"); + } catch (...) { + ignoreException(); + } + } +}; + + void checkStoreNotSymlink() { if (getEnv("NIX_IGNORE_SYMLINK_STORE") == "1") return; @@ -44,16 +200,17 @@ LocalStore::LocalStore() schemaPath = nixDBPath + "/schema"; - if (readOnlyMode) return; + if (readOnlyMode) { + openDB(false); + return; + } /* Create missing state directories if they don't already exist. */ createDirs(nixStore); - createDirs(nixDBPath + "/info"); - createDirs(nixDBPath + "/referrer"); - createDirs(nixDBPath + "/failed"); Path profilesDir = nixStateDir + "/profiles"; createDirs(nixStateDir + "/profiles"); createDirs(nixStateDir + "/temproots"); + createDirs(nixDBPath); Path gcRootsDir = nixStateDir + "/gcroots"; if (!pathExists(gcRootsDir)) { createDirs(gcRootsDir); @@ -63,12 +220,15 @@ LocalStore::LocalStore() checkStoreNotSymlink(); + /* Acquire the big fat lock in shared mode to make sure that no + schema upgrade is in progress. */ try { Path globalLockPath = nixDBPath + "/big-lock"; globalLock = openLockFile(globalLockPath.c_str(), true); } catch (SysError & e) { if (e.errNo != EACCES) throw; readOnlyMode = true; + openDB(false); return; } @@ -76,33 +236,55 @@ LocalStore::LocalStore() printMsg(lvlError, "waiting for the big Nix store lock..."); lockFile(globalLock, ltRead, true); } - + + /* Check the current database schema and if necessary do an + upgrade. */ int curSchema = getSchema(); if (curSchema > nixSchemaVersion) throw Error(format("current Nix store schema is version %1%, but I only support %2%") % curSchema % nixSchemaVersion); - if (curSchema == 0) { /* new store */ - curSchema = nixSchemaVersion; + + else if (curSchema == 0) { /* new store */ + curSchema = nixSchemaVersion; + openDB(true); writeFile(schemaPath, (format("%1%") % nixSchemaVersion).str()); } - if (curSchema == 1) throw Error("your Nix store is no longer supported"); - if (curSchema < nixSchemaVersion) upgradeStore12(); + + else if (curSchema < nixSchemaVersion) { + if (curSchema < 5) + throw Error( + "Your Nix store has a database in Berkeley DB format,\n" + "which is no longer supported. To convert to the new format,\n" + "please upgrade Nix to version 0.12 first."); + + if (!lockFile(globalLock, ltWrite, false)) { + printMsg(lvlError, "waiting for exclusive access to the Nix store..."); + lockFile(globalLock, ltWrite, true); + } + + /* Get the schema version again, because another process may + have performed the upgrade already. */ + curSchema = getSchema(); + + if (curSchema < 6) upgradeStore6(); - doFsync = queryBoolSetting("fsync-metadata", false); + writeFile(schemaPath, (format("%1%") % nixSchemaVersion).str()); + + lockFile(globalLock, ltRead, true); + } + + else openDB(false); } LocalStore::~LocalStore() { try { - flushDelayedUpdates(); - foreach (RunningSubstituters::iterator, i, runningSubstituters) { i->second.to.close(); i->second.from.close(); i->second.pid.wait(true); } - } catch (...) { ignoreException(); } @@ -121,6 +303,98 @@ int LocalStore::getSchema() } +void LocalStore::openDB(bool create) +{ + /* Open the Nix database. */ + if (sqlite3_open_v2((nixDBPath + "/db.sqlite").c_str(), &db.db, + SQLITE_OPEN_READWRITE | (create ? SQLITE_OPEN_CREATE : 0), 0) != SQLITE_OK) + throw Error("cannot open SQLite database"); + + if (sqlite3_busy_timeout(db, 60 * 60 * 1000) != SQLITE_OK) + throwSQLiteError(db, "setting timeout"); + + if (sqlite3_exec(db, "pragma foreign_keys = 1;", 0, 0, 0) != SQLITE_OK) + throwSQLiteError(db, "enabling foreign keys"); + + /* !!! check whether sqlite has been built with foreign key + support */ + + /* Whether SQLite should fsync(). "Normal" synchronous mode + should be safe enough. If the user asks for it, don't sync at + all. This can cause database corruption if the system + crashes. */ + string syncMode = queryBoolSetting("fsync-metadata", true) ? "normal" : "off"; + if (sqlite3_exec(db, ("pragma synchronous = " + syncMode + ";").c_str(), 0, 0, 0) != SQLITE_OK) + throwSQLiteError(db, "setting synchronous mode"); + + /* Set the SQLite journal mode. WAL mode is fastest, but doesn't + seem entirely stable at the moment (Oct. 2010). Thus, use + truncate mode by default. */ + string mode = queryBoolSetting("use-sqlite-wal", false) ? "wal" : "truncate"; + string prevMode; + { + SQLiteStmt stmt; + stmt.create(db, "pragma main.journal_mode;"); + if (sqlite3_step(stmt) != SQLITE_ROW) + throwSQLiteError(db, "querying journal mode"); + prevMode = string((const char *) sqlite3_column_text(stmt, 0)); + } + if (prevMode != mode && + sqlite3_exec(db, ("pragma main.journal_mode = " + mode + ";").c_str(), 0, 0, 0) != SQLITE_OK) + throwSQLiteError(db, "setting journal mode"); + + /* Increase the auto-checkpoint interval to 8192 pages. This + seems enough to ensure that instantiating the NixOS system + derivation is done in a single fsync(). */ + if (mode == "wal" && sqlite3_exec(db, "pragma wal_autocheckpoint = 8192;", 0, 0, 0) != SQLITE_OK) + throwSQLiteError(db, "setting autocheckpoint interval"); + + /* Initialise the database schema, if necessary. */ + if (create) { +#include "schema.sql.hh" + if (sqlite3_exec(db, (const char *) schema, 0, 0, 0) != SQLITE_OK) + throwSQLiteError(db, "initialising database schema"); + } + + /* Backwards compatibility with old (pre-release) databases. Can + remove this eventually. */ + if (sqlite3_table_column_metadata(db, 0, "ValidPaths", "narSize", 0, 0, 0, 0, 0) != SQLITE_OK) { + if (sqlite3_exec(db, "alter table ValidPaths add column narSize integer" , 0, 0, 0) != SQLITE_OK) + throwSQLiteError(db, "adding column narSize"); + } + + /* Prepare SQL statements. */ + stmtRegisterValidPath.create(db, + "insert into ValidPaths (path, hash, registrationTime, deriver, narSize) values (?, ?, ?, ?, ?);"); + stmtUpdatePathInfo.create(db, + "update ValidPaths set narSize = ? where path = ?;"); + stmtAddReference.create(db, + "insert or replace into Refs (referrer, reference) values (?, ?);"); + stmtQueryPathInfo.create(db, + "select id, hash, registrationTime, deriver, narSize from ValidPaths where path = ?;"); + stmtQueryReferences.create(db, + "select path from Refs join ValidPaths on reference = id where referrer = ?;"); + stmtQueryReferrers.create(db, + "select path from Refs join ValidPaths on referrer = id where reference = (select id from ValidPaths where path = ?);"); + stmtInvalidatePath.create(db, + "delete from ValidPaths where path = ?;"); + stmtRegisterFailedPath.create(db, + "insert into FailedPaths (path, time) values (?, ?);"); + stmtHasPathFailed.create(db, + "select time from FailedPaths where path = ?;"); + stmtQueryFailedPaths.create(db, + "select path from FailedPaths;"); + stmtClearFailedPath.create(db, + "delete from FailedPaths where ?1 = '*' or path = ?1;"); + stmtAddDerivationOutput.create(db, + "insert or replace into DerivationOutputs (drv, id, path) values (?, ?, ?);"); + stmtQueryValidDerivers.create(db, + "select v.id, v.path from DerivationOutputs d join ValidPaths v on d.drv = v.id where d.path = ?;"); + stmtQueryDerivationOutputs.create(db, + "select id, path from DerivationOutputs where drv = ?;"); +} + + void canonicalisePathMetaData(const Path & path, bool recurse) { checkInterrupt(); @@ -195,181 +469,107 @@ void canonicalisePathMetaData(const Path & path) } -static Path infoFileFor(const Path & path) -{ - string baseName = baseNameOf(path); - return (format("%1%/info/%2%") % nixDBPath % baseName).str(); -} - - -static Path referrersFileFor(const Path & path) -{ - string baseName = baseNameOf(path); - return (format("%1%/referrer/%2%") % nixDBPath % baseName).str(); -} - - -static Path failedFileFor(const Path & path) -{ - string baseName = baseNameOf(path); - return (format("%1%/failed/%2%") % nixDBPath % baseName).str(); -} - - -static Path tmpFileForAtomicUpdate(const Path & path) +unsigned long long LocalStore::addValidPath(const ValidPathInfo & info) { - return (format("%1%/.%2%.%3%") % dirOf(path) % getpid() % baseNameOf(path)).str(); -} - - -void LocalStore::appendReferrer(const Path & from, const Path & to, bool lock) -{ - Path referrersFile = referrersFileFor(from); - - PathLocks referrersLock; - if (lock) { - referrersLock.lockPaths(singleton<PathSet, Path>(referrersFile)); - referrersLock.setDeletion(true); + SQLiteStmtUse use(stmtRegisterValidPath); + stmtRegisterValidPath.bind(info.path); + stmtRegisterValidPath.bind("sha256:" + printHash(info.hash)); + stmtRegisterValidPath.bind(info.registrationTime == 0 ? time(0) : info.registrationTime); + if (info.deriver != "") + stmtRegisterValidPath.bind(info.deriver); + else + stmtRegisterValidPath.bind(); // null + if (info.narSize != 0) + stmtRegisterValidPath.bind64(info.narSize); + else + stmtRegisterValidPath.bind(); // null + if (sqlite3_step(stmtRegisterValidPath) != SQLITE_DONE) + throwSQLiteError(db, format("registering valid path `%1%' in database") % info.path); + unsigned long long id = sqlite3_last_insert_rowid(db); + + /* If this is a derivation, then store the derivation outputs in + the database. This is useful for the garbage collector: it can + efficiently query whether a path is an output of some + derivation. */ + if (isDerivation(info.path)) { + Derivation drv = parseDerivation(readFile(info.path)); + foreach (DerivationOutputs::iterator, i, drv.outputs) { + SQLiteStmtUse use(stmtAddDerivationOutput); + stmtAddDerivationOutput.bind(id); + stmtAddDerivationOutput.bind(i->first); + stmtAddDerivationOutput.bind(i->second.path); + if (sqlite3_step(stmtAddDerivationOutput) != SQLITE_DONE) + throwSQLiteError(db, format("adding derivation output for `%1%' in database") % info.path); + } } - AutoCloseFD fd = open(referrersFile.c_str(), O_WRONLY | O_APPEND | O_CREAT, 0666); - if (fd == -1) throw SysError(format("opening file `%1%'") % referrersFile); - - string s = " " + to; - writeFull(fd, (const unsigned char *) s.c_str(), s.size()); - - if (doFsync) fsync(fd); + return id; } -/* Atomically update the referrers file. If `purge' is true, the set - of referrers is set to `referrers'. Otherwise, the current set of - referrers is purged of invalid paths. */ -void LocalStore::rewriteReferrers(const Path & path, bool purge, PathSet referrers) +void LocalStore::addReference(unsigned long long referrer, unsigned long long reference) { - Path referrersFile = referrersFileFor(path); - - PathLocks referrersLock(singleton<PathSet, Path>(referrersFile)); - referrersLock.setDeletion(true); - - if (purge) - /* queryReferrers() purges invalid paths, so that's all we - need. */ - queryReferrers(path, referrers); - - Path tmpFile = tmpFileForAtomicUpdate(referrersFile); - - AutoCloseFD fd = open(tmpFile.c_str(), O_WRONLY | O_TRUNC | O_CREAT, 0666); - if (fd == -1) throw SysError(format("opening file `%1%'") % referrersFile); - - string s; - foreach (PathSet::const_iterator, i, referrers) { - s += " "; s += *i; - } - - writeFull(fd, (const unsigned char *) s.c_str(), s.size()); - - if (doFsync) fsync(fd); - - fd.close(); /* for Windows; can't rename open file */ - - if (rename(tmpFile.c_str(), referrersFile.c_str()) == -1) - throw SysError(format("cannot rename `%1%' to `%2%'") % tmpFile % referrersFile); + SQLiteStmtUse use(stmtAddReference); + stmtAddReference.bind(referrer); + stmtAddReference.bind(reference); + if (sqlite3_step(stmtAddReference) != SQLITE_DONE) + throwSQLiteError(db, "adding reference to database"); } -void LocalStore::flushDelayedUpdates() +void LocalStore::registerFailedPath(const Path & path) { - foreach (PathSet::iterator, i, delayedUpdates) { - rewriteReferrers(*i, true, PathSet()); - } - delayedUpdates.clear(); + if (hasPathFailed(path)) return; + SQLiteStmtUse use(stmtRegisterFailedPath); + stmtRegisterFailedPath.bind(path); + stmtRegisterFailedPath.bind(time(0)); + if (sqlite3_step(stmtRegisterFailedPath) != SQLITE_DONE) + throwSQLiteError(db, format("registering failed path `%1%'") % path); } -void LocalStore::registerValidPath(const Path & path, - const Hash & hash, const PathSet & references, - const Path & deriver) +bool LocalStore::hasPathFailed(const Path & path) { - ValidPathInfo info; - info.path = path; - info.hash = hash; - info.references = references; - info.deriver = deriver; - registerValidPath(info); + SQLiteStmtUse use(stmtHasPathFailed); + stmtHasPathFailed.bind(path); + int res = sqlite3_step(stmtHasPathFailed); + if (res != SQLITE_DONE && res != SQLITE_ROW) + throwSQLiteError(db, "querying whether path failed"); + return res == SQLITE_ROW; } -void LocalStore::registerValidPath(const ValidPathInfo & info, bool ignoreValidity) +PathSet LocalStore::queryFailedPaths() { - Path infoFile = infoFileFor(info.path); - - ValidPathInfo oldInfo; - if (pathExists(infoFile)) oldInfo = queryPathInfo(info.path); - - /* Note that it's possible for infoFile to already exist. */ - - /* Acquire a lock on each referrer file. This prevents those - paths from being invalidated. (It would be a violation of the - store invariants if we registered info.path as valid while some - of its references are invalid.) NB: there can be no deadlock - here since we're acquiring the locks in sorted order. */ - PathSet lockNames; - foreach (PathSet::const_iterator, i, info.references) - if (*i != info.path) lockNames.insert(referrersFileFor(*i)); - PathLocks referrerLocks(lockNames); - referrerLocks.setDeletion(true); - - string refs; - foreach (PathSet::const_iterator, i, info.references) { - if (!refs.empty()) refs += " "; - refs += *i; - - if (!ignoreValidity && *i != info.path && !isValidPath(*i)) - throw Error(format("cannot register `%1%' as valid, because its reference `%2%' isn't valid") - % info.path % *i); - - /* Update the referrer mapping for *i. This must be done - before the info file is written to maintain the invariant - that if `path' is a valid path, then all its references - have referrer mappings back to `path'. A " " is prefixed - to separate it from the previous entry. It's not suffixed - to deal with interrupted partial writes to this file. */ - if (oldInfo.references.find(*i) == oldInfo.references.end()) - appendReferrer(*i, info.path, false); + SQLiteStmtUse use(stmtQueryFailedPaths); + + PathSet res; + int r; + while ((r = sqlite3_step(stmtQueryFailedPaths)) == SQLITE_ROW) { + const char * s = (const char *) sqlite3_column_text(stmtQueryFailedPaths, 0); + assert(s); + res.insert(s); } - assert(info.hash.type == htSHA256); + if (r != SQLITE_DONE) + throwSQLiteError(db, "error querying failed paths"); - string s = (format( - "Hash: sha256:%1%\n" - "References: %2%\n" - "Deriver: %3%\n" - "Registered-At: %4%\n") - % printHash(info.hash) % refs % info.deriver % - (oldInfo.registrationTime ? oldInfo.registrationTime : time(0))).str(); - - /* Atomically rewrite the info file. */ - Path tmpFile = tmpFileForAtomicUpdate(infoFile); - writeFile(tmpFile, s, doFsync); - if (rename(tmpFile.c_str(), infoFile.c_str()) == -1) - throw SysError(format("cannot rename `%1%' to `%2%'") % tmpFile % infoFile); - - pathInfoCache[info.path] = info; + return res; } -void LocalStore::registerFailedPath(const Path & path) +void LocalStore::clearFailedPaths(const PathSet & paths) { - /* Write an empty file in the .../failed directory to denote the - failure of the builder for `path'. */ - writeFile(failedFileFor(path), ""); -} + SQLiteTxn txn(db); + foreach (PathSet::const_iterator, i, paths) { + SQLiteStmtUse use(stmtClearFailedPath); + stmtClearFailedPath.bind(*i); + if (sqlite3_step(stmtClearFailedPath) != SQLITE_DONE) + throwSQLiteError(db, format("clearing failed path `%1%' in database") % *i); + } -bool LocalStore::hasPathFailed(const Path & path) -{ - return pathExists(failedFileFor(path)); + txn.commit(); } @@ -387,91 +587,109 @@ Hash parseHashField(const Path & path, const string & s) } -ValidPathInfo LocalStore::queryPathInfo(const Path & path, bool ignoreErrors) +ValidPathInfo LocalStore::queryPathInfo(const Path & path) { - ValidPathInfo res; - res.path = path; + ValidPathInfo info; + info.path = path; assertStorePath(path); - if (!isValidPath(path)) - throw Error(format("path `%1%' is not valid") % path); + /* Get the path info. */ + SQLiteStmtUse use1(stmtQueryPathInfo); - std::map<Path, ValidPathInfo>::iterator lookup = pathInfoCache.find(path); - if (lookup != pathInfoCache.end()) return lookup->second; + stmtQueryPathInfo.bind(path); - /* Read the info file. */ - Path infoFile = infoFileFor(path); - if (!pathExists(infoFile)) - throw Error(format("path `%1%' is not valid") % path); - string info = readFile(infoFile); + int r = sqlite3_step(stmtQueryPathInfo); + if (r == SQLITE_DONE) throw Error(format("path `%1%' is not valid") % path); + if (r != SQLITE_ROW) throwSQLiteError(db, "querying path in database"); - /* Parse it. */ - Strings lines = tokenizeString(info, "\n"); + info.id = sqlite3_column_int(stmtQueryPathInfo, 0); - foreach (Strings::iterator, i, lines) { - string::size_type p = i->find(':'); - if (p == string::npos) { - if (!ignoreErrors) - throw Error(format("corrupt line in `%1%': %2%") % infoFile % *i); - continue; /* bad line */ - } - string name(*i, 0, p); - string value(*i, p + 2); - if (name == "References") { - Strings refs = tokenizeString(value, " "); - res.references = PathSet(refs.begin(), refs.end()); - } else if (name == "Deriver") { - res.deriver = value; - } else if (name == "Hash") { - try { - res.hash = parseHashField(path, value); - } catch (Error & e) { - if (!ignoreErrors) throw; - printMsg(lvlError, format("cannot parse hash field in `%1%': %2%") % infoFile % e.msg()); - } - } else if (name == "Registered-At") { - int n = 0; - string2Int(value, n); - res.registrationTime = n; - } + const char * s = (const char *) sqlite3_column_text(stmtQueryPathInfo, 1); + assert(s); + info.hash = parseHashField(path, s); + + info.registrationTime = sqlite3_column_int(stmtQueryPathInfo, 2); + + s = (const char *) sqlite3_column_text(stmtQueryPathInfo, 3); + if (s) info.deriver = s; + + /* Note that narSize = NULL yields 0. */ + info.narSize = sqlite3_column_int64(stmtQueryPathInfo, 4); + + /* Get the references. */ + SQLiteStmtUse use2(stmtQueryReferences); + + stmtQueryReferences.bind(info.id); + + while ((r = sqlite3_step(stmtQueryReferences)) == SQLITE_ROW) { + s = (const char *) sqlite3_column_text(stmtQueryReferences, 0); + assert(s); + info.references.insert(s); } - return pathInfoCache[path] = res; + if (r != SQLITE_DONE) + throwSQLiteError(db, format("error getting references of `%1%'") % path); + + return info; } -bool LocalStore::isValidPath(const Path & path) +/* Update path info in the database. Currently only updated the + narSize field. */ +void LocalStore::updatePathInfo(const ValidPathInfo & info) { - /* Files in the info directory starting with a `.' are temporary - files. */ - if (baseNameOf(path).at(0) == '.') return false; + SQLiteStmtUse use(stmtUpdatePathInfo); + if (info.narSize != 0) + stmtUpdatePathInfo.bind64(info.narSize); + else + stmtUpdatePathInfo.bind(); // null + stmtUpdatePathInfo.bind(info.path); + if (sqlite3_step(stmtUpdatePathInfo) != SQLITE_DONE) + throwSQLiteError(db, format("updating info of path `%1%' in database") % info.path); +} - /* A path is valid if its info file exists and has a non-zero - size. (The non-zero size restriction is to be robust to - certain kinds of filesystem corruption, particularly with - ext4.) */ - Path infoFile = infoFileFor(path); - struct stat st; - if (lstat(infoFile.c_str(), &st)) { - if (errno == ENOENT) return false; - throw SysError(format("getting status of `%1%'") % infoFile); - } +unsigned long long LocalStore::queryValidPathId(const Path & path) +{ + SQLiteStmtUse use(stmtQueryPathInfo); + stmtQueryPathInfo.bind(path); + int res = sqlite3_step(stmtQueryPathInfo); + if (res == SQLITE_ROW) return sqlite3_column_int(stmtQueryPathInfo, 0); + if (res == SQLITE_DONE) throw Error(format("path `%1%' is not valid") % path); + throwSQLiteError(db, "querying path in database"); +} - if (st.st_size == 0) return false; - - return true; + +bool LocalStore::isValidPath(const Path & path) +{ + SQLiteStmtUse use(stmtQueryPathInfo); + stmtQueryPathInfo.bind(path); + int res = sqlite3_step(stmtQueryPathInfo); + if (res != SQLITE_DONE && res != SQLITE_ROW) + throwSQLiteError(db, "querying path in database"); + return res == SQLITE_ROW; } PathSet LocalStore::queryValidPaths() { - PathSet paths; - Strings entries = readDirectory(nixDBPath + "/info"); - foreach (Strings::iterator, i, entries) - if (i->at(0) != '.') paths.insert(nixStore + "/" + *i); - return paths; + SQLiteStmt stmt; + stmt.create(db, "select path from ValidPaths"); + + PathSet res; + + int r; + while ((r = sqlite3_step(stmt)) == SQLITE_ROW) { + const char * s = (const char *) sqlite3_column_text(stmt, 0); + assert(s); + res.insert(s); + } + + if (r != SQLITE_DONE) + throwSQLiteError(db, "error getting valid paths"); + + return res; } @@ -483,45 +701,73 @@ void LocalStore::queryReferences(const Path & path, } -bool LocalStore::queryReferrersInternal(const Path & path, PathSet & referrers) +void LocalStore::queryReferrers(const Path & path, PathSet & referrers) { - bool allValid = true; - - if (!isValidPath(path)) - throw Error(format("path `%1%' is not valid") % path); + assertStorePath(path); - /* No locking is necessary here: updates are only done by - appending or by atomically replacing the file. When appending, - there is a possibility that we see a partial entry, but it will - just be filtered out below (the partially written path will not - be valid, so it will be ignored). */ + SQLiteStmtUse use(stmtQueryReferrers); - Path referrersFile = referrersFileFor(path); - if (!pathExists(referrersFile)) return true; - - AutoCloseFD fd = open(referrersFile.c_str(), O_RDONLY); - if (fd == -1) throw SysError(format("opening file `%1%'") % referrersFile); + stmtQueryReferrers.bind(path); - Paths refs = tokenizeString(readFile(fd), " "); + int r; + while ((r = sqlite3_step(stmtQueryReferrers)) == SQLITE_ROW) { + const char * s = (const char *) sqlite3_column_text(stmtQueryReferrers, 0); + assert(s); + referrers.insert(s); + } - foreach (Paths::iterator, i, refs) - /* Referrers can be invalid (see registerValidPath() for the - invariant), so we only return one if it is valid. */ - if (isStorePath(*i) && isValidPath(*i)) referrers.insert(*i); else allValid = false; + if (r != SQLITE_DONE) + throwSQLiteError(db, format("error getting references of `%1%'") % path); +} - return allValid; + +Path LocalStore::queryDeriver(const Path & path) +{ + return queryPathInfo(path).deriver; } -void LocalStore::queryReferrers(const Path & path, PathSet & referrers) +PathSet LocalStore::queryValidDerivers(const Path & path) { - queryReferrersInternal(path, referrers); + assertStorePath(path); + + SQLiteStmtUse use(stmtQueryValidDerivers); + stmtQueryValidDerivers.bind(path); + + PathSet derivers; + int r; + while ((r = sqlite3_step(stmtQueryValidDerivers)) == SQLITE_ROW) { + const char * s = (const char *) sqlite3_column_text(stmtQueryValidDerivers, 1); + assert(s); + derivers.insert(s); + } + + if (r != SQLITE_DONE) + throwSQLiteError(db, format("error getting valid derivers of `%1%'") % path); + + return derivers; } -Path LocalStore::queryDeriver(const Path & path) +PathSet LocalStore::queryDerivationOutputs(const Path & path) { - return queryPathInfo(path).deriver; + SQLiteTxn txn(db); + + SQLiteStmtUse use(stmtQueryDerivationOutputs); + stmtQueryDerivationOutputs.bind(queryValidPathId(path)); + + PathSet outputs; + int r; + while ((r = sqlite3_step(stmtQueryDerivationOutputs)) == SQLITE_ROW) { + const char * s = (const char *) sqlite3_column_text(stmtQueryDerivationOutputs, 1); + assert(s); + outputs.insert(s); + } + + if (r != SQLITE_DONE) + throwSQLiteError(db, format("error getting outputs of `%1%'") % path); + + return outputs; } @@ -615,6 +861,7 @@ bool LocalStore::querySubstitutablePathInfo(const Path & substituter, info.references.insert(p); } info.downloadSize = getIntLine<long long>(run.from); + info.narSize = getIntLine<long long>(run.from); return true; } @@ -635,39 +882,40 @@ Hash LocalStore::queryPathHash(const Path & path) } -static void dfsVisit(std::map<Path, ValidPathInfo> & infos, - const Path & path, PathSet & visited, Paths & sorted) +void LocalStore::registerValidPath(const ValidPathInfo & info) { - if (visited.find(path) != visited.end()) return; - visited.insert(path); - - ValidPathInfo & info(infos[path]); - - foreach (PathSet::iterator, i, info.references) - if (infos.find(*i) != infos.end()) - dfsVisit(infos, *i, visited, sorted); - - sorted.push_back(path); + ValidPathInfos infos; + infos.push_back(info); + registerValidPaths(infos); } void LocalStore::registerValidPaths(const ValidPathInfos & infos) { - std::map<Path, ValidPathInfo> infosMap; + while (1) { + try { + SQLiteTxn txn(db); - /* Sort the paths topologically under the references relation, so - that if path A is referenced by B, then A is registered before - B. */ - foreach (ValidPathInfos::const_iterator, i, infos) - infosMap[i->path] = *i; - - PathSet visited; - Paths sorted; - foreach (ValidPathInfos::const_iterator, i, infos) - dfsVisit(infosMap, i->path, visited, sorted); - - foreach (Paths::iterator, i, sorted) - registerValidPath(infosMap[*i]); + foreach (ValidPathInfos::const_iterator, i, infos) { + assert(i->hash.type == htSHA256); + /* !!! Maybe the registration info should be updated if the + path is already valid. */ + if (!isValidPath(i->path)) addValidPath(*i); + } + + foreach (ValidPathInfos::const_iterator, i, infos) { + unsigned long long referrer = queryValidPathId(i->path); + foreach (PathSet::iterator, j, i->references) + addReference(referrer, queryValidPathId(*j)); + } + + txn.commit(); + break; + } catch (SQLiteBusy & e) { + /* Retry; the `txn' destructor will roll back the current + transaction. */ + } + } } @@ -677,43 +925,15 @@ void LocalStore::invalidatePath(const Path & path) { debug(format("invalidating path `%1%'") % path); - ValidPathInfo info; + SQLiteStmtUse use(stmtInvalidatePath); - if (pathExists(infoFileFor(path))) { - info = queryPathInfo(path); + stmtInvalidatePath.bind(path); - /* Remove the info file. */ - Path p = infoFileFor(path); - if (unlink(p.c_str()) == -1) - throw SysError(format("unlinking `%1%'") % p); - } + if (sqlite3_step(stmtInvalidatePath) != SQLITE_DONE) + throwSQLiteError(db, format("invalidating path `%1%' in database") % path); - /* Remove the referrers file for `path'. */ - Path p = referrersFileFor(path); - if (pathExists(p) && unlink(p.c_str()) == -1) - throw SysError(format("unlinking `%1%'") % p); - - /* Clear `path' from the info cache. */ - pathInfoCache.erase(path); - delayedUpdates.erase(path); - - /* Cause the referrer files for each path referenced by this one - to be updated. This has to happen after removing the info file - to preserve the invariant (see registerValidPath()). - - The referrer files are updated lazily in flushDelayedUpdates() - to prevent quadratic performance in the garbage collector - (i.e., when N referrers to some path X are deleted, we have to - rewrite the referrers file for X N times, causing O(N^2) I/O). - - What happens if we die before the referrer file can be updated? - That's not a problem, because stale (invalid) entries in the - referrer file are ignored by queryReferrers(). Thus a referrer - file is allowed to have stale entries; removing them is just an - optimisation. verifyStore() gets rid of them eventually. - */ - foreach (PathSet::iterator, i, info.references) - if (*i != path) delayedUpdates.insert(*i); + /* Note that the foreign key constraints on the Refs table take + care of deleting the references entries for `path'. */ } @@ -749,10 +969,18 @@ Path LocalStore::addToStoreFromDump(const string & dump, const string & name, the path in the database. We may just have computed it above (if called with recursive == true and hashAlgo == sha256); otherwise, compute it here. */ - registerValidPath(dstPath, - (recursive && hashAlgo == htSHA256) ? h : - (recursive ? hashString(htSHA256, dump) : hashPath(htSHA256, dstPath)), - PathSet(), ""); + HashResult hash; + if (recursive) { + hash.first = hashAlgo == htSHA256 ? h : hashString(htSHA256, dump); + hash.second = dump.size(); + } else + hash = hashPath(htSHA256, dstPath); + + ValidPathInfo info; + info.path = dstPath; + info.hash = hash.first; + info.narSize = hash.second; + registerValidPath(info); } outputLock.setDeletion(true); @@ -799,9 +1027,15 @@ Path LocalStore::addTextToStore(const string & name, const string & s, writeFile(dstPath, s); canonicalisePathMetaData(dstPath); + + HashResult hash = hashPath(htSHA256, dstPath); - registerValidPath(dstPath, - hashPath(htSHA256, dstPath), references, ""); + ValidPathInfo info; + info.path = dstPath; + info.hash = hash.first; + info.narSize = hash.second; + info.references = references; + registerValidPath(info); } outputLock.setDeletion(true); @@ -815,16 +1049,19 @@ struct HashAndWriteSink : Sink { Sink & writeSink; HashSink hashSink; - bool hashing; HashAndWriteSink(Sink & writeSink) : writeSink(writeSink), hashSink(htSHA256) { - hashing = true; } virtual void operator () (const unsigned char * data, unsigned int len) { writeSink(data, len); - if (hashing) hashSink(data, len); + hashSink(data, len); + } + Hash currentHash() + { + HashSink hashSinkClone(hashSink); + return hashSinkClone.finish().first; } }; @@ -855,6 +1092,15 @@ void LocalStore::exportPath(const Path & path, bool sign, dumpPath(path, hashAndWriteSink); + /* Refuse to export paths that have changed. This prevents + filesystem corruption from spreading to other machines. + Don't complain if the stored hash is zero (unknown). */ + Hash hash = hashAndWriteSink.currentHash(); + Hash storedHash = queryPathHash(path); + if (hash != storedHash && storedHash != Hash(storedHash.type)) + throw Error(format("hash of path `%1%' has changed from `%2%' to `%3%'!") % path + % printHash(storedHash) % printHash(hash)); + writeInt(EXPORT_MAGIC, hashAndWriteSink); writeString(path, hashAndWriteSink); @@ -867,14 +1113,11 @@ void LocalStore::exportPath(const Path & path, bool sign, writeString(deriver, hashAndWriteSink); if (sign) { - Hash hash = hashAndWriteSink.hashSink.finish(); - hashAndWriteSink.hashing = false; - + Hash hash = hashAndWriteSink.currentHash(); + writeInt(1, hashAndWriteSink); Path tmpDir = createTempDir(); - PathLocks tmpDirLock(singleton<PathSet, Path>(tmpDir)); - tmpDirLock.setDeletion(true); AutoDelete delTmp(tmpDir); Path hashFile = tmpDir + "/hash"; writeFile(hashFile, printHash(hash)); @@ -916,6 +1159,22 @@ struct HashAndReadSource : Source }; +/* Create a temporary directory in the store that won't be + garbage-collected. */ +Path LocalStore::createTempDirInStore() +{ + Path tmpDir; + do { + /* There is a slight possibility that `tmpDir' gets deleted by + the GC between createTempDir() and addTempRoot(), so repeat + until `tmpDir' exists. */ + tmpDir = createTempDir(nixStore); + addTempRoot(tmpDir); + } while (!pathExists(tmpDir)); + return tmpDir; +} + + Path LocalStore::importPath(bool requireSignature, Source & source) { HashAndReadSource hashAndReadSource(source); @@ -923,10 +1182,8 @@ Path LocalStore::importPath(bool requireSignature, Source & source) /* We don't yet know what store path this archive contains (the store path follows the archive data proper), and besides, we don't know yet whether the signature is valid. */ - Path tmpDir = createTempDir(nixStore); - PathLocks tmpDirLock(singleton<PathSet, Path>(tmpDir)); - tmpDirLock.setDeletion(true); - AutoDelete delTmp(tmpDir); /* !!! could be GC'ed! */ + Path tmpDir = createTempDirInStore(); + AutoDelete delTmp(tmpDir); Path unpacked = tmpDir + "/unpacked"; restorePath(unpacked, hashAndReadSource); @@ -942,7 +1199,7 @@ Path LocalStore::importPath(bool requireSignature, Source & source) Path deriver = readString(hashAndReadSource); if (deriver != "") assertStorePath(deriver); - Hash hash = hashAndReadSource.hashSink.finish(); + Hash hash = hashAndReadSource.hashSink.finish().first; hashAndReadSource.hashing = false; bool haveSignature = readInt(hashAndReadSource) == 1; @@ -1006,9 +1263,15 @@ Path LocalStore::importPath(bool requireSignature, Source & source) /* !!! if we were clever, we could prevent the hashPath() here. */ - if (deriver != "" && !isValidPath(deriver)) deriver = ""; - registerValidPath(dstPath, - hashPath(htSHA256, dstPath), references, deriver); + HashResult hash = hashPath(htSHA256, dstPath); + + ValidPathInfo info; + info.path = dstPath; + info.hash = hash.first; + info.narSize = hash.second; + info.references = references; + info.deriver = deriver != "" && isValidPath(deriver) ? deriver : ""; + registerValidPath(info); } outputLock.setDeletion(true); @@ -1025,163 +1288,205 @@ void LocalStore::deleteFromStore(const Path & path, unsigned long long & bytesFr assertStorePath(path); - if (isValidPath(path)) { - /* Acquire a lock on the referrers file to prevent new - referrers to this path from appearing while we're deleting - it. */ - PathLocks referrersLock(singleton<PathSet, Path>(referrersFileFor(path))); - referrersLock.setDeletion(true); - PathSet referrers; queryReferrers(path, referrers); - referrers.erase(path); /* ignore self-references */ - if (!referrers.empty()) - throw PathInUse(format("cannot delete path `%1%' because it is in use by `%2%'") - % path % showPaths(referrers)); - invalidatePath(path); - } + while (1) { + try { + SQLiteTxn txn(db); + + if (isValidPath(path)) { + PathSet referrers; queryReferrers(path, referrers); + referrers.erase(path); /* ignore self-references */ + if (!referrers.empty()) + throw PathInUse(format("cannot delete path `%1%' because it is in use by `%2%'") + % path % showPaths(referrers)); + invalidatePath(path); + } + txn.commit(); + break; + } catch (SQLiteBusy & e) { }; + } + deletePathWrapped(path, bytesFreed, blocksFreed); } void LocalStore::verifyStore(bool checkContents) { - /* Check whether all valid paths actually exist. */ - printMsg(lvlInfo, "checking path existence"); + printMsg(lvlError, format("reading the Nix store...")); - PathSet validPaths2 = queryValidPaths(), validPaths; + /* Acquire the global GC lock to prevent a garbage collection. */ + AutoCloseFD fdGCLock = openGCLock(ltWrite); - foreach (PathSet::iterator, i, validPaths2) { - checkInterrupt(); - if (!isStorePath(*i)) { - printMsg(lvlError, format("path `%1%' is not in the Nix store") % *i); - invalidatePath(*i); - } else if (!pathExists(*i)) { - printMsg(lvlError, format("path `%1%' disappeared") % *i); - invalidatePath(*i); - } else { - Path infoFile = infoFileFor(*i); - struct stat st; - if (lstat(infoFile.c_str(), &st)) - throw SysError(format("getting status of `%1%'") % infoFile); - if (st.st_size == 0) { - printMsg(lvlError, format("removing corrupt info file `%1%'") % infoFile); - if (unlink(infoFile.c_str()) == -1) - throw SysError(format("unlinking `%1%'") % infoFile); - } - else validPaths.insert(*i); - } - } + Paths entries = readDirectory(nixStore); + PathSet store(entries.begin(), entries.end()); + /* Check whether all valid paths actually exist. */ + printMsg(lvlInfo, "checking path existence..."); - /* Check the store path meta-information. */ - printMsg(lvlInfo, "checking path meta-information"); + PathSet validPaths2 = queryValidPaths(), validPaths, done; - std::map<Path, PathSet> referrersCache; - - foreach (PathSet::iterator, i, validPaths) { - bool update = false; - ValidPathInfo info = queryPathInfo(*i, true); - - /* Check the references: each reference should be valid, and - it should have a matching referrer. */ - foreach (PathSet::iterator, j, info.references) { - if (validPaths.find(*j) == validPaths.end()) { - printMsg(lvlError, format("incomplete closure: `%1%' needs missing `%2%'") - % *i % *j); - /* nothing we can do about it... */ - } else { - if (referrersCache.find(*j) == referrersCache.end()) - queryReferrers(*j, referrersCache[*j]); - if (referrersCache[*j].find(*i) == referrersCache[*j].end()) { - printMsg(lvlError, format("adding missing referrer mapping from `%1%' to `%2%'") - % *j % *i); - appendReferrer(*j, *i, true); + foreach (PathSet::iterator, i, validPaths2) + verifyPath(*i, store, done, validPaths); + + /* Release the GC lock so that checking content hashes (which can + take ages) doesn't block the GC or builds. */ + fdGCLock.close(); + + /* Optionally, check the content hashes (slow). */ + if (checkContents) { + printMsg(lvlInfo, "checking hashes..."); + + foreach (PathSet::iterator, i, validPaths) { + try { + ValidPathInfo info = queryPathInfo(*i); + + /* Check the content hash (optionally - slow). */ + printMsg(lvlTalkative, format("checking contents of `%1%'") % *i); + HashResult current = hashPath(info.hash.type, *i); + + if (current.first != info.hash) { + printMsg(lvlError, format("path `%1%' was modified! " + "expected hash `%2%', got `%3%'") + % *i % printHash(info.hash) % printHash(current.first)); + } else { + /* Fill in missing narSize fields (from old stores). */ + if (info.narSize == 0) { + printMsg(lvlError, format("updating size field on `%1%' to %2%") % *i % current.second); + info.narSize = current.second; + updatePathInfo(info); + } } + + } catch (Error & e) { + /* It's possible that the path got GC'ed, so ignore + errors on invalid paths. */ + if (isValidPath(*i)) throw; + printMsg(lvlError, format("warning: %1%") % e.msg()); } } + } +} - /* Check the deriver. (Note that the deriver doesn't have to - be a valid path.) */ - if (!info.deriver.empty() && !isStorePath(info.deriver)) { - info.deriver = ""; - update = true; - } - /* Check the content hash (optionally - slow). */ - if (info.hash.hashSize == 0) { - printMsg(lvlError, format("re-hashing `%1%'") % *i); - info.hash = hashPath(htSHA256, *i); - update = true; - } else if (checkContents) { - debug(format("checking contents of `%1%'") % *i); - Hash current = hashPath(info.hash.type, *i); - if (current != info.hash) { - printMsg(lvlError, format("path `%1%' was modified! " - "expected hash `%2%', got `%3%'") - % *i % printHash(info.hash) % printHash(current)); - } - } +void LocalStore::verifyPath(const Path & path, const PathSet & store, + PathSet & done, PathSet & validPaths) +{ + checkInterrupt(); + + if (done.find(path) != done.end()) return; + done.insert(path); - if (update) registerValidPath(info); + if (!isStorePath(path)) { + printMsg(lvlError, format("path `%1%' is not in the Nix store") % path); + invalidatePath(path); + return; } - referrersCache.clear(); + if (store.find(baseNameOf(path)) == store.end()) { + /* Check any referrers first. If we can invalidate them + first, then we can invalidate this path as well. */ + bool canInvalidate = true; + PathSet referrers; queryReferrers(path, referrers); + foreach (PathSet::iterator, i, referrers) + if (*i != path) { + verifyPath(*i, store, done, validPaths); + if (validPaths.find(*i) != validPaths.end()) + canInvalidate = false; + } + + if (canInvalidate) { + printMsg(lvlError, format("path `%1%' disappeared, removing from database...") % path); + invalidatePath(path); + } else + printMsg(lvlError, format("path `%1%' disappeared, but it still has valid referrers!") % path); + + return; + } + validPaths.insert(path); +} - /* Check the referrers. */ - printMsg(lvlInfo, "checking referrers"); - std::map<Path, PathSet> referencesCache; - - Strings entries = readDirectory(nixDBPath + "/referrer"); - foreach (Strings::iterator, i, entries) { - Path from = nixStore + "/" + *i; - - if (validPaths.find(from) == validPaths.end()) { - /* !!! This removes lock files as well. Need to check - whether that's okay. */ - printMsg(lvlError, format("removing referrers file for invalid `%1%'") % from); - Path p = referrersFileFor(from); - if (unlink(p.c_str()) == -1) - throw SysError(format("unlinking `%1%'") % p); - continue; - } +/* Functions for upgrading from the pre-SQLite database. */ - PathSet referrers; - bool allValid = queryReferrersInternal(from, referrers); - bool update = false; +PathSet LocalStore::queryValidPathsOld() +{ + PathSet paths; + Strings entries = readDirectory(nixDBPath + "/info"); + foreach (Strings::iterator, i, entries) + if (i->at(0) != '.') paths.insert(nixStore + "/" + *i); + return paths; +} - if (!allValid) { - printMsg(lvlError, format("removing some stale referrers for `%1%'") % from); - update = true; - } - /* Each referrer should have a matching reference. */ - PathSet referrersNew; - foreach (PathSet::iterator, j, referrers) { - if (referencesCache.find(*j) == referencesCache.end()) - queryReferences(*j, referencesCache[*j]); - if (referencesCache[*j].find(from) == referencesCache[*j].end()) { - printMsg(lvlError, format("removing unexpected referrer mapping from `%1%' to `%2%'") - % from % *j); - update = true; - } else referrersNew.insert(*j); - } +ValidPathInfo LocalStore::queryPathInfoOld(const Path & path) +{ + ValidPathInfo res; + res.path = path; - if (update) rewriteReferrers(from, false, referrersNew); + /* Read the info file. */ + string baseName = baseNameOf(path); + Path infoFile = (format("%1%/info/%2%") % nixDBPath % baseName).str(); + if (!pathExists(infoFile)) + throw Error(format("path `%1%' is not valid") % path); + string info = readFile(infoFile); + + /* Parse it. */ + Strings lines = tokenizeString(info, "\n"); + + foreach (Strings::iterator, i, lines) { + string::size_type p = i->find(':'); + if (p == string::npos) + throw Error(format("corrupt line in `%1%': %2%") % infoFile % *i); + string name(*i, 0, p); + string value(*i, p + 2); + if (name == "References") { + Strings refs = tokenizeString(value, " "); + res.references = PathSet(refs.begin(), refs.end()); + } else if (name == "Deriver") { + res.deriver = value; + } else if (name == "Hash") { + res.hash = parseHashField(path, value); + } else if (name == "Registered-At") { + int n = 0; + string2Int(value, n); + res.registrationTime = n; + } } + + return res; } -/* Upgrade from schema 4 (Nix 0.11) to schema 5 (Nix >= 0.12). The - old schema uses Berkeley DB, the new one stores store path - meta-information in files. */ -void LocalStore::upgradeStore12() +/* Upgrade from schema 5 (Nix 0.12) to schema 6 (Nix >= 0.15). */ +void LocalStore::upgradeStore6() { - throw Error( - "Your Nix store has a database in Berkeley DB format,\n" - "which is no longer supported. To convert to the new format,\n" - "please upgrade Nix to version 0.12 first."); + printMsg(lvlError, "upgrading Nix store to new schema (this may take a while)..."); + + openDB(true); + + PathSet validPaths = queryValidPathsOld(); + + SQLiteTxn txn(db); + + foreach (PathSet::iterator, i, validPaths) { + addValidPath(queryPathInfoOld(*i)); + std::cerr << "."; + } + + std::cerr << "|"; + + foreach (PathSet::iterator, i, validPaths) { + ValidPathInfo info = queryPathInfoOld(*i); + unsigned long long referrer = queryValidPathId(*i); + foreach (PathSet::iterator, j, info.references) + addReference(referrer, queryValidPathId(*j)); + std::cerr << "."; + } + + std::cerr << "\n"; + + txn.commit(); } diff --git a/src/libstore/local-store.hh b/src/libstore/local-store.hh index 31f8d91096a3..f270fb723964 100644 --- a/src/libstore/local-store.hh +++ b/src/libstore/local-store.hh @@ -5,6 +5,11 @@ #include "store-api.hh" #include "util.hh" +#include "pathlocks.hh" + + +class sqlite3; +class sqlite3_stmt; namespace nix { @@ -12,8 +17,9 @@ namespace nix { /* Nix store and database schema version. Version 1 (or 0) was Nix <= 0.7. Version 2 was Nix 0.8 and 0.9. Version 3 is Nix 0.10. - Version 4 is Nix 0.11. Version 5 is Nix 0.12*/ -const int nixSchemaVersion = 5; + Version 4 is Nix 0.11. Version 5 is Nix 0.12-0.16. Version 6 is + Nix 1.0. */ +const int nixSchemaVersion = 6; extern string drvsLogDir; @@ -41,6 +47,34 @@ struct RunningSubstituter }; +/* Wrapper object to close the SQLite database automatically. */ +struct SQLite +{ + sqlite3 * db; + SQLite() { db = 0; } + ~SQLite(); + operator sqlite3 * () { return db; } +}; + + +/* Wrapper object to create and destroy SQLite prepared statements. */ +struct SQLiteStmt +{ + sqlite3 * db; + sqlite3_stmt * stmt; + unsigned int curArg; + SQLiteStmt() { stmt = 0; } + void create(sqlite3 * db, const string & s); + void reset(); + ~SQLiteStmt(); + operator sqlite3_stmt * () { return stmt; } + void bind(const string & value); + void bind(int value); + void bind64(long long value); + void bind(); +}; + + class LocalStore : public StoreAPI { private: @@ -64,6 +98,8 @@ public: PathSet queryValidPaths(); + ValidPathInfo queryPathInfo(const Path & path); + Hash queryPathHash(const Path & path); void queryReferences(const Path & path, PathSet & references); @@ -71,6 +107,14 @@ public: void queryReferrers(const Path & path, PathSet & referrers); Path queryDeriver(const Path & path); + + /* Return all currently valid derivations that have `path' as an + output. (Note that the result of `queryDeriver()' is the + derivation that was actually used to produce `path', which may + not exist anymore.) */ + PathSet queryValidDerivers(const Path & path); + + PathSet queryDerivationOutputs(const Path & path); PathSet querySubstitutablePaths(); @@ -132,8 +176,7 @@ public: execution of the derivation (or something equivalent). Also register the hash of the file system contents of the path. The hash must be a SHA-256 hash. */ - void registerValidPath(const Path & path, - const Hash & hash, const PathSet & references, const Path & deriver); + void registerValidPath(const ValidPathInfo & info); void registerValidPaths(const ValidPathInfos & infos); @@ -144,6 +187,10 @@ public: /* Query whether `path' previously failed to build. */ bool hasPathFailed(const Path & path); + PathSet queryFailedPaths(); + + void clearFailedPaths(const PathSet & paths); + private: Path schemaPath; @@ -151,45 +198,63 @@ private: /* Lock file used for upgrading. */ AutoCloseFD globalLock; - /* !!! The cache can grow very big. Maybe it should be pruned - every once in a while. */ - std::map<Path, ValidPathInfo> pathInfoCache; - - /* Store paths for which the referrers file must be purged. */ - PathSet delayedUpdates; - - /* Whether to do an fsync() after writing Nix metadata. */ - bool doFsync; + /* The SQLite database object. */ + SQLite db; + + /* Some precompiled SQLite statements. */ + SQLiteStmt stmtRegisterValidPath; + SQLiteStmt stmtUpdatePathInfo; + SQLiteStmt stmtAddReference; + SQLiteStmt stmtQueryPathInfo; + SQLiteStmt stmtQueryReferences; + SQLiteStmt stmtQueryReferrers; + SQLiteStmt stmtInvalidatePath; + SQLiteStmt stmtRegisterFailedPath; + SQLiteStmt stmtHasPathFailed; + SQLiteStmt stmtQueryFailedPaths; + SQLiteStmt stmtClearFailedPath; + SQLiteStmt stmtAddDerivationOutput; + SQLiteStmt stmtQueryValidDerivers; + SQLiteStmt stmtQueryDerivationOutputs; int getSchema(); - void registerValidPath(const ValidPathInfo & info, bool ignoreValidity = false); + void openDB(bool create); - ValidPathInfo queryPathInfo(const Path & path, bool ignoreErrors = false); + unsigned long long queryValidPathId(const Path & path); + unsigned long long addValidPath(const ValidPathInfo & info); + + void addReference(unsigned long long referrer, unsigned long long reference); + void appendReferrer(const Path & from, const Path & to, bool lock); void rewriteReferrers(const Path & path, bool purge, PathSet referrers); - void flushDelayedUpdates(); - - bool queryReferrersInternal(const Path & path, PathSet & referrers); - void invalidatePath(const Path & path); - - void upgradeStore12(); + + void verifyPath(const Path & path, const PathSet & store, + PathSet & done, PathSet & validPaths); + + void updatePathInfo(const ValidPathInfo & info); + + void upgradeStore6(); + PathSet queryValidPathsOld(); + ValidPathInfo queryPathInfoOld(const Path & path); struct GCState; bool tryToDelete(GCState & state, const Path & path); - PathSet findDerivers(GCState & state, const Path & path); - bool isActiveTempFile(const GCState & state, const Path & path, const string & suffix); + int openGCLock(LockType lockType); + void startSubstituter(const Path & substituter, RunningSubstituter & runningSubstituter); + + Path createTempDirInStore(); }; diff --git a/src/libstore/misc.cc b/src/libstore/misc.cc index f2cc20626915..01d6a97ae6fb 100644 --- a/src/libstore/misc.cc +++ b/src/libstore/misc.cc @@ -27,10 +27,10 @@ void computeFSClosure(const Path & storePath, store->queryReferences(storePath, references); if (includeOutputs && isDerivation(storePath)) { - Derivation drv = derivationFromPath(storePath); - foreach (DerivationOutputs::iterator, i, drv.outputs) - if (store->isValidPath(i->second.path)) - computeFSClosure(i->second.path, paths, flipDirection, true); + PathSet outputs = store->queryDerivationOutputs(storePath); + foreach (PathSet::iterator, i, outputs) + if (store->isValidPath(*i)) + computeFSClosure(*i, paths, flipDirection, true); } foreach (PathSet::iterator, i, references) @@ -48,9 +48,9 @@ Path findOutput(const Derivation & drv, string id) void queryMissing(const PathSet & targets, PathSet & willBuild, PathSet & willSubstitute, PathSet & unknown, - unsigned long long & downloadSize) + unsigned long long & downloadSize, unsigned long long & narSize) { - downloadSize = 0; + downloadSize = narSize = 0; PathSet todo(targets.begin(), targets.end()), done; @@ -88,6 +88,7 @@ void queryMissing(const PathSet & targets, if (store->querySubstitutablePathInfo(p, info)) { willSubstitute.insert(p); downloadSize += info.downloadSize; + narSize += info.narSize; todo.insert(info.references.begin(), info.references.end()); } else unknown.insert(p); diff --git a/src/libstore/misc.hh b/src/libstore/misc.hh index 0bc9a2ee0d9c..abef6237e7f6 100644 --- a/src/libstore/misc.hh +++ b/src/libstore/misc.hh @@ -31,7 +31,7 @@ Path findOutput(const Derivation & drv, string id); will be substituted. */ void queryMissing(const PathSet & targets, PathSet & willBuild, PathSet & willSubstitute, PathSet & unknown, - unsigned long long & downloadSize); + unsigned long long & downloadSize, unsigned long long & narSize); } diff --git a/src/libstore/optimise-store.cc b/src/libstore/optimise-store.cc index 3ed54e24d773..89be6ac6529a 100644 --- a/src/libstore/optimise-store.cc +++ b/src/libstore/optimise-store.cc @@ -68,7 +68,7 @@ static void hashAndLink(bool dryRun, HashToPath & hashToPath, the contents of the symlink (i.e. the result of readlink()), not the contents of the target (which may not even exist). */ - Hash hash = hashPath(htSHA256, path); + Hash hash = hashPath(htSHA256, path).first; stats.totalFiles++; printMsg(lvlDebug, format("`%1%' has hash `%2%'") % path % printHash(hash)); diff --git a/src/libstore/references.cc b/src/libstore/references.cc index a6f6e85fc84b..ade9c9aa20e3 100644 --- a/src/libstore/references.cc +++ b/src/libstore/references.cc @@ -81,7 +81,7 @@ void RefScanSink::operator () (const unsigned char * data, unsigned int len) PathSet scanForReferences(const string & path, - const PathSet & refs, Hash & hash) + const PathSet & refs, HashResult & hash) { RefScanSink sink; std::map<string, Path> backMap; diff --git a/src/libstore/references.hh b/src/libstore/references.hh index 7d068eb51700..158c08a77646 100644 --- a/src/libstore/references.hh +++ b/src/libstore/references.hh @@ -7,7 +7,7 @@ namespace nix { PathSet scanForReferences(const Path & path, const PathSet & refs, - Hash & hash); + HashResult & hash); } diff --git a/src/libstore/remote-store.cc b/src/libstore/remote-store.cc index 3d8b2a0c4263..ae99846a3dfb 100644 --- a/src/libstore/remote-store.cc +++ b/src/libstore/remote-store.cc @@ -97,10 +97,6 @@ void RemoteStore::forkSlave() if (worker == "") worker = nixBinDir + "/nix-worker"; - string verbosityArg = "-"; - for (int i = 1; i < verbosity; ++i) - verbosityArg += "v"; - child = fork(); switch (child) { @@ -120,10 +116,7 @@ void RemoteStore::forkSlave() close(fdSocket); close(fdChild); - execlp(worker.c_str(), worker.c_str(), "--slave", - /* hacky - must be at the end */ - verbosityArg == "-" ? NULL : verbosityArg.c_str(), - NULL); + execlp(worker.c_str(), worker.c_str(), "--slave", NULL); throw SysError(format("executing `%1%'") % worker); @@ -198,9 +191,8 @@ void RemoteStore::setOptions() writeInt(logType, to); writeInt(printBuildTrace, to); } - if (GET_PROTOCOL_MINOR(daemonVersion) >= 6) { + if (GET_PROTOCOL_MINOR(daemonVersion) >= 6) writeInt(buildCores, to); - } processStderr(); } @@ -219,7 +211,9 @@ bool RemoteStore::isValidPath(const Path & path) PathSet RemoteStore::queryValidPaths() { openConnection(); - throw Error("not implemented"); + writeInt(wopQueryValidPaths, to); + processStderr(); + return readStorePaths(from); } @@ -248,10 +242,29 @@ bool RemoteStore::querySubstitutablePathInfo(const Path & path, if (info.deriver != "") assertStorePath(info.deriver); info.references = readStorePaths(from); info.downloadSize = readLongLong(from); + info.narSize = GET_PROTOCOL_MINOR(daemonVersion) >= 7 ? readLongLong(from) : 0; return true; } +ValidPathInfo RemoteStore::queryPathInfo(const Path & path) +{ + openConnection(); + writeInt(wopQueryPathInfo, to); + writeString(path, to); + processStderr(); + ValidPathInfo info; + info.path = path; + info.deriver = readString(from); + if (info.deriver != "") assertStorePath(info.deriver); + info.hash = parseHash(htSHA256, readString(from)); + info.references = readStorePaths(from); + info.registrationTime = readInt(from); + info.narSize = readLongLong(from); + return info; +} + + Hash RemoteStore::queryPathHash(const Path & path) { openConnection(); @@ -299,6 +312,16 @@ Path RemoteStore::queryDeriver(const Path & path) } +PathSet RemoteStore::queryDerivationOutputs(const Path & path) +{ + openConnection(); + writeInt(wopQueryDerivationOutputs, to); + writeString(path, to); + processStderr(); + return readStorePaths(from); +} + + Path RemoteStore::addToStore(const Path & _srcPath, bool recursive, HashType hashAlgo, PathFilter & filter) { @@ -444,6 +467,25 @@ void RemoteStore::collectGarbage(const GCOptions & options, GCResults & results) } +PathSet RemoteStore::queryFailedPaths() +{ + openConnection(); + writeInt(wopQueryFailedPaths, to); + processStderr(); + return readStorePaths(from); +} + + +void RemoteStore::clearFailedPaths(const PathSet & paths) +{ + openConnection(); + writeInt(wopClearFailedPaths, to); + writeStringSet(paths, to); + processStderr(); + readInt(from); +} + + void RemoteStore::processStderr(Sink * sink, Source * source) { unsigned int msg; @@ -467,8 +509,11 @@ void RemoteStore::processStderr(Sink * sink, Source * source) writeToStderr((const unsigned char *) s.c_str(), s.size()); } } - if (msg == STDERR_ERROR) - throw Error(readString(from)); + if (msg == STDERR_ERROR) { + string error = readString(from); + unsigned int status = GET_PROTOCOL_MINOR(daemonVersion) >= 8 ? readInt(from) : 1; + throw Error(error, status); + } else if (msg != STDERR_LAST) throw Error("protocol error processing standard error"); } diff --git a/src/libstore/remote-store.hh b/src/libstore/remote-store.hh index 3d55d23d958e..519f46fd1a6d 100644 --- a/src/libstore/remote-store.hh +++ b/src/libstore/remote-store.hh @@ -29,6 +29,8 @@ public: PathSet queryValidPaths(); + ValidPathInfo queryPathInfo(const Path & path); + Hash queryPathHash(const Path & path); void queryReferences(const Path & path, PathSet & references); @@ -37,6 +39,8 @@ public: Path queryDeriver(const Path & path); + PathSet queryDerivationOutputs(const Path & path); + bool hasSubstitutes(const Path & path); bool querySubstitutablePathInfo(const Path & path, @@ -68,6 +72,10 @@ public: void collectGarbage(const GCOptions & options, GCResults & results); + PathSet queryFailedPaths(); + + void clearFailedPaths(const PathSet & paths); + private: AutoCloseFD fdSocket; FdSink to; diff --git a/src/libstore/schema.sql b/src/libstore/schema.sql new file mode 100644 index 000000000000..c1b4a689afcb --- /dev/null +++ b/src/libstore/schema.sql @@ -0,0 +1,44 @@ +create table if not exists ValidPaths ( + id integer primary key autoincrement not null, + path text unique not null, + hash text not null, + registrationTime integer not null, + deriver text, + narSize integer +); + +create table if not exists Refs ( + referrer integer not null, + reference integer not null, + primary key (referrer, reference), + foreign key (referrer) references ValidPaths(id) on delete cascade, + foreign key (reference) references ValidPaths(id) on delete restrict +); + +create index if not exists IndexReferrer on Refs(referrer); +create index if not exists IndexReference on Refs(reference); + +-- Paths can refer to themselves, causing a tuple (N, N) in the Refs +-- table. This causes a deletion of the corresponding row in +-- ValidPaths to cause a foreign key constraint violation (due to `on +-- delete restrict' on the `reference' column). Therefore, explicitly +-- get rid of self-references. +create trigger if not exists DeleteSelfRefs before delete on ValidPaths + begin + delete from Refs where referrer = old.id and reference = old.id; + end; + +create table if not exists DerivationOutputs ( + drv integer not null, + id text not null, -- symbolic output id, usually "out" + path text not null, + primary key (drv, id), + foreign key (drv) references ValidPaths(id) on delete cascade +); + +create index if not exists IndexDerivationOutputs on DerivationOutputs(path); + +create table if not exists FailedPaths ( + path text primary key not null, + time integer not null +); diff --git a/src/libstore/store-api.cc b/src/libstore/store-api.cc index f0abe61ad1e1..4b04f5751ce8 100644 --- a/src/libstore/store-api.cc +++ b/src/libstore/store-api.cc @@ -1,7 +1,6 @@ #include "store-api.hh" #include "globals.hh" #include "util.hh" -#include "derivations.hh" #include <limits.h> @@ -53,18 +52,6 @@ Path toStorePath(const Path & path) } -string getNameOfStorePath(const Path & path) -{ - Path::size_type slash = path.rfind('/'); - string p = slash == Path::npos ? path : string(path, slash + 1); - Path::size_type dash = p.find('-'); - assert(dash != Path::npos); - string p2 = string(p, dash + 1); - if (isDerivation(p2)) p2 = string(p2, 0, p2.size() - 4); - return p2; -} - - Path followLinksToStore(const Path & _path) { Path path = absPath(_path); @@ -203,7 +190,7 @@ std::pair<Path, Hash> computeStorePathForPath(const Path & srcPath, bool recursive, HashType hashAlgo, PathFilter & filter) { HashType ht(hashAlgo); - Hash h = recursive ? hashPath(ht, srcPath, filter) : hashFile(ht, srcPath); + Hash h = recursive ? hashPath(ht, srcPath, filter).first : hashFile(ht, srcPath); string name = baseNameOf(srcPath); Path dstPath = makeFixedOutputPath(recursive, hashAlgo, h, name); return std::pair<Path, Hash>(dstPath, h); @@ -229,7 +216,7 @@ Path computeStorePathForText(const string & name, const string & s, /* Return a string accepted by decodeValidPathInfo() that registers the specified paths as valid. Note: it's the responsibility of the caller to provide a closure. */ -string makeValidityRegistration(const PathSet & paths, +string StoreAPI::makeValidityRegistration(const PathSet & paths, bool showDerivers, bool showHash) { string s = ""; @@ -237,18 +224,19 @@ string makeValidityRegistration(const PathSet & paths, foreach (PathSet::iterator, i, paths) { s += *i + "\n"; - if (showHash) - s += printHash(store->queryPathHash(*i)) + "\n"; + ValidPathInfo info = queryPathInfo(*i); - Path deriver = showDerivers ? store->queryDeriver(*i) : ""; + if (showHash) { + s += printHash(info.hash) + "\n"; + s += (format("%1%\n") % info.narSize).str(); + } + + Path deriver = showDerivers ? info.deriver : ""; s += deriver + "\n"; - PathSet references; - store->queryReferences(*i, references); + s += (format("%1%\n") % info.references.size()).str(); - s += (format("%1%\n") % references.size()).str(); - - foreach (PathSet::iterator, j, references) + foreach (PathSet::iterator, j, info.references) s += *j + "\n"; } @@ -265,6 +253,8 @@ ValidPathInfo decodeValidPathInfo(std::istream & str, bool hashGiven) string s; getline(str, s); info.hash = parseHash(htSHA256, s); + getline(str, s); + if (!string2Int(s, info.narSize)) throw Error("number expected"); } getline(str, info.deriver); string s; int n; diff --git a/src/libstore/store-api.hh b/src/libstore/store-api.hh index 6fc0689ba1af..40ac887140d7 100644 --- a/src/libstore/store-api.hh +++ b/src/libstore/store-api.hh @@ -87,9 +87,25 @@ struct SubstitutablePathInfo Path deriver; PathSet references; unsigned long long downloadSize; /* 0 = unknown or inapplicable */ + unsigned long long narSize; /* 0 = unknown */ }; +struct ValidPathInfo +{ + Path path; + Path deriver; + Hash hash; + PathSet references; + time_t registrationTime; + unsigned long long narSize; // 0 = unknown + unsigned long long id; // internal use only + ValidPathInfo() : registrationTime(0), narSize(0) { } +}; + +typedef list<ValidPathInfo> ValidPathInfos; + + class StoreAPI { public: @@ -102,6 +118,9 @@ public: /* Query the set of valid paths. */ virtual PathSet queryValidPaths() = 0; + /* Query information about a valid path. */ + virtual ValidPathInfo queryPathInfo(const Path & path) = 0; + /* Queries the hash of a valid path. */ virtual Hash queryPathHash(const Path & path) = 0; @@ -110,33 +129,18 @@ public: virtual void queryReferences(const Path & path, PathSet & references) = 0; - /* Like queryReferences, but with self-references filtered out. */ - PathSet queryReferencesNoSelf(const Path & path) - { - PathSet res; - queryReferences(path, res); - res.erase(path); - return res; - } - /* Queries the set of incoming FS references for a store path. The result is not cleared. */ virtual void queryReferrers(const Path & path, PathSet & referrers) = 0; - /* Like queryReferrers, but with self-references filtered out. */ - PathSet queryReferrersNoSelf(const Path & path) - { - PathSet res; - queryReferrers(path, res); - res.erase(path); - return res; - } - /* Query the deriver of a store path. Return the empty string if no deriver has been set. */ virtual Path queryDeriver(const Path & path) = 0; + /* Query the outputs of the derivation denoted by `path'. */ + virtual PathSet queryDerivationOutputs(const Path & path) = 0; + /* Query whether a path has substitutes. */ virtual bool hasSubstitutes(const Path & path) = 0; @@ -222,6 +226,19 @@ public: /* Perform a garbage collection. */ virtual void collectGarbage(const GCOptions & options, GCResults & results) = 0; + + /* Return the set of paths that have failed to build.*/ + virtual PathSet queryFailedPaths() = 0; + + /* Clear the "failed" status of the given paths. The special + value `*' causes all failed paths to be cleared. */ + virtual void clearFailedPaths(const PathSet & paths) = 0; + + /* Return a string representing information about the path that + can be loaded into the database using `nix-store --load-db' or + `nix-store --register-validity'. */ + string makeValidityRegistration(const PathSet & paths, + bool showDerivers, bool showHash); }; @@ -241,12 +258,6 @@ void checkStoreName(const string & name); Path toStorePath(const Path & path); -/* Get the "name" part of a store path, that is, the part after the - hash and the dash, and with any ".drv" suffix removed - (e.g. /nix/store/<hash>-foo-1.2.3.drv => foo-1.2.3). */ -string getNameOfStorePath(const Path & path); - - /* Follow symlinks until we end up with a path in the Nix store. */ Path followLinksToStore(const Path & path); @@ -321,21 +332,6 @@ boost::shared_ptr<StoreAPI> openStore(); string showPaths(const PathSet & paths); -string makeValidityRegistration(const PathSet & paths, - bool showDerivers, bool showHash); - -struct ValidPathInfo -{ - Path path; - Path deriver; - Hash hash; - PathSet references; - time_t registrationTime; - ValidPathInfo() : registrationTime(0) { } -}; - -typedef list<ValidPathInfo> ValidPathInfos; - ValidPathInfo decodeValidPathInfo(std::istream & str, bool hashGiven = false); diff --git a/src/libstore/worker-protocol.hh b/src/libstore/worker-protocol.hh index 5504773b8f91..acb8bc8b2948 100644 --- a/src/libstore/worker-protocol.hh +++ b/src/libstore/worker-protocol.hh @@ -8,7 +8,7 @@ namespace nix { #define WORKER_MAGIC_1 0x6e697863 #define WORKER_MAGIC_2 0x6478696f -#define PROTOCOL_VERSION 0x106 +#define PROTOCOL_VERSION 0x108 #define GET_PROTOCOL_MAJOR(x) ((x) & 0xff00) #define GET_PROTOCOL_MINOR(x) ((x) & 0x00ff) @@ -34,6 +34,11 @@ typedef enum { wopSetOptions = 19, wopCollectGarbage = 20, wopQuerySubstitutablePathInfo = 21, + wopQueryDerivationOutputs = 22, + wopQueryValidPaths = 23, + wopQueryFailedPaths = 24, + wopClearFailedPaths = 25, + wopQueryPathInfo = 26, } WorkerOp; diff --git a/src/libutil/Makefile.am b/src/libutil/Makefile.am index aa862208c6f9..98f32633b894 100644 --- a/src/libutil/Makefile.am +++ b/src/libutil/Makefile.am @@ -3,7 +3,7 @@ pkglib_LTLIBRARIES = libutil.la libutil_la_SOURCES = util.cc hash.cc serialise.cc \ archive.cc xml-writer.cc -libutil_la_LIBADD = ../boost/format/libformat.la +libutil_la_LIBADD = ../boost/format/libformat.la ${aterm_lib} ${sqlite_lib} pkginclude_HEADERS = util.hh hash.hh serialise.hh \ archive.hh xml-writer.hh types.hh diff --git a/src/libutil/archive.cc b/src/libutil/archive.cc index 8fde4328c47e..999b17cd2f19 100644 --- a/src/libutil/archive.cc +++ b/src/libutil/archive.cc @@ -181,8 +181,6 @@ static void parseContents(ParseSink & sink, Source & source, const Path & path) left -= n; } - sink.finalizeContents(size); - readPadding(size, source); } @@ -317,12 +315,6 @@ struct RestoreSink : ParseSink writeFull(fd, data, len); } - void finalizeContents(unsigned long long size) - { - errno = ftruncate(fd, size); - if (errno) throw SysError(format("truncating file to its allocated length of %1% bytes") % size); - } - void createSymlink(const Path & path, const string & target) { Path p = dstPath + path; diff --git a/src/libutil/archive.hh b/src/libutil/archive.hh index f358a2a6be17..fff62031397c 100644 --- a/src/libutil/archive.hh +++ b/src/libutil/archive.hh @@ -64,7 +64,6 @@ struct ParseSink virtual void isExecutable() { }; virtual void preallocateContents(unsigned long long size) { }; virtual void receiveContents(unsigned char * data, unsigned int len) { }; - virtual void finalizeContents(unsigned long long size) { }; virtual void createSymlink(const Path & path, const string & target) { }; }; diff --git a/src/libutil/hash.cc b/src/libutil/hash.cc index eef01fe4d609..b9e7846992f6 100644 --- a/src/libutil/hash.cc +++ b/src/libutil/hash.cc @@ -286,9 +286,18 @@ Hash hashFile(HashType ht, const Path & path) HashSink::HashSink(HashType ht) : ht(ht) { ctx = new Ctx; + bytes = 0; start(ht, *ctx); } +HashSink::HashSink(const HashSink & h) +{ + ht = h.ht; + bytes = h.bytes; + ctx = new Ctx; + *ctx = *h.ctx; +} + HashSink::~HashSink() { delete ctx; @@ -297,18 +306,20 @@ HashSink::~HashSink() void HashSink::operator () (const unsigned char * data, unsigned int len) { + bytes += len; update(ht, *ctx, data, len); } -Hash HashSink::finish() +HashResult HashSink::finish() { Hash hash(ht); nix::finish(ht, *ctx, hash.hash); - return hash; + return HashResult(hash, bytes); } -Hash hashPath(HashType ht, const Path & path, PathFilter & filter) +HashResult hashPath( + HashType ht, const Path & path, PathFilter & filter) { HashSink sink(ht); dumpPath(path, sink, filter); diff --git a/src/libutil/hash.hh b/src/libutil/hash.hh index 062d97254bfc..13740954d625 100644 --- a/src/libutil/hash.hh +++ b/src/libutil/hash.hh @@ -40,7 +40,6 @@ struct Hash /* For sorting. */ bool operator < (const Hash & h) const; - }; @@ -72,7 +71,8 @@ Hash hashFile(HashType ht, const Path & path); (essentially) hashString(ht, dumpPath(path)). */ struct PathFilter; extern PathFilter defaultPathFilter; -Hash hashPath(HashType ht, const Path & path, +typedef std::pair<Hash, unsigned long long> HashResult; +HashResult hashPath(HashType ht, const Path & path, PathFilter & filter = defaultPathFilter); /* Compress a hash to the specified number of bytes by cyclically @@ -93,16 +93,18 @@ class HashSink : public Sink private: HashType ht; Ctx * ctx; + unsigned long long bytes; public: HashSink(HashType ht); + HashSink(const HashSink & h); ~HashSink(); virtual void operator () (const unsigned char * data, unsigned int len); - Hash finish(); + HashResult finish(); }; } - + #endif /* !__HASH_H */ diff --git a/src/libutil/types.hh b/src/libutil/types.hh index f110188da151..844ad6f76a13 100644 --- a/src/libutil/types.hh +++ b/src/libutil/types.hh @@ -27,7 +27,8 @@ protected: string prefix_; // used for location traces etc. string err; public: - BaseError(const format & f); + unsigned int status; // exit status + BaseError(const format & f, unsigned int status = 1); ~BaseError() throw () { }; const char * what() const throw () { return err.c_str(); } const string & msg() const throw () { return err; } @@ -39,7 +40,7 @@ public: class newClass : public superClass \ { \ public: \ - newClass(const format & f) : superClass(f) { }; \ + newClass(const format & f, unsigned int status = 1) : superClass(f, status) { }; \ }; MakeError(Error, BaseError) @@ -63,7 +64,7 @@ typedef set<Path> PathSet; typedef enum { - lvlError, + lvlError = 0, lvlInfo, lvlTalkative, lvlChatty, diff --git a/src/libutil/util.cc b/src/libutil/util.cc index 32c8fce9a1b1..9adaac40d56e 100644 --- a/src/libutil/util.cc +++ b/src/libutil/util.cc @@ -7,11 +7,8 @@ #include <sstream> #include <cstring> -#include <sys/stat.h> #include <sys/wait.h> -#include <sys/types.h> #include <fcntl.h> -#include <unistd.h> #include <limits.h> #include "util.hh" @@ -23,7 +20,8 @@ extern char * * environ; namespace nix { -BaseError::BaseError(const format & f) +BaseError::BaseError(const format & f, unsigned int status) + : status(status) { err = f.str(); } @@ -149,6 +147,15 @@ string baseNameOf(const Path & path) } +struct stat lstat(const Path & path) +{ + struct stat st; + if (lstat(path.c_str(), &st)) + throw SysError(format("getting status of `%1%'") % path); + return st; +} + + bool pathExists(const Path & path) { int res; @@ -164,9 +171,7 @@ bool pathExists(const Path & path) Path readLink(const Path & path) { checkInterrupt(); - struct stat st; - if (lstat(path.c_str(), &st)) - throw SysError(format("getting status of `%1%'") % path); + struct stat st = lstat(path); if (!S_ISLNK(st.st_mode)) throw Error(format("`%1%' is not a symlink") % path); char buf[st.st_size]; @@ -178,9 +183,7 @@ Path readLink(const Path & path) bool isLink(const Path & path) { - struct stat st; - if (lstat(path.c_str(), &st)) - throw SysError(format("getting status of `%1%'") % path); + struct stat st = lstat(path); return S_ISLNK(st.st_mode); } @@ -228,13 +231,12 @@ string readFile(const Path & path) } -void writeFile(const Path & path, const string & s, bool doFsync) +void writeFile(const Path & path, const string & s) { AutoCloseFD fd = open(path.c_str(), O_WRONLY | O_TRUNC | O_CREAT, 0666); if (fd == -1) throw SysError(format("opening file `%1%'") % path); writeFull(fd, (unsigned char *) s.c_str(), s.size()); - if (doFsync) fsync(fd); } @@ -270,9 +272,7 @@ static void _computePathSize(const Path & path, { checkInterrupt(); - struct stat st; - if (lstat(path.c_str(), &st)) - throw SysError(format("getting attributes of path `%1%'") % path); + struct stat st = lstat(path); bytes += st.st_size; blocks += st.st_blocks; @@ -302,9 +302,7 @@ static void _deletePath(const Path & path, unsigned long long & bytesFreed, printMsg(lvlVomit, format("%1%") % path); - struct stat st; - if (lstat(path.c_str(), &st)) - throw SysError(format("getting attributes of path `%1%'") % path); + struct stat st = lstat(path); if (!S_ISDIR(st.st_mode) && st.st_nlink == 1) { bytesFreed += st.st_size; @@ -351,9 +349,7 @@ void makePathReadOnly(const Path & path) { checkInterrupt(); - struct stat st; - if (lstat(path.c_str(), &st)) - throw SysError(format("getting attributes of path `%1%'") % path); + struct stat st = lstat(path); if (!S_ISLNK(st.st_mode) && (st.st_mode & S_IWUSR)) { if (chmod(path.c_str(), st.st_mode & ~S_IWUSR) == -1) @@ -412,12 +408,18 @@ Paths createDirs(const Path & path) { Paths created; if (path == "/") return created; - if (!pathExists(path)) { + + struct stat st; + if (lstat(path.c_str(), &st) == -1) { created = createDirs(dirOf(path)); - if (mkdir(path.c_str(), 0777) == -1) + if (mkdir(path.c_str(), 0777) == -1 && errno != EEXIST) throw SysError(format("creating directory `%1%'") % path); + st = lstat(path); created.push_back(path); } + + if (!S_ISDIR(st.st_mode)) throw Error(format("`%1%' is not a directory") % path); + return created; } @@ -976,6 +978,17 @@ Strings tokenizeString(const string & s, const string & separators) } +string concatStringsSep(const string & sep, const Strings & ss) +{ + string s; + foreach (Strings::const_iterator, i, ss) { + if (s.size() != 0) s += sep; + s += *i; + } + return s; +} + + string statusToString(int status) { if (!WIFEXITED(status) || WEXITSTATUS(status) != 0) { diff --git a/src/libutil/util.hh b/src/libutil/util.hh index ff710077ceaa..f86290f31694 100644 --- a/src/libutil/util.hh +++ b/src/libutil/util.hh @@ -4,6 +4,7 @@ #include "types.hh" #include <sys/types.h> +#include <sys/stat.h> #include <dirent.h> #include <unistd.h> #include <signal.h> @@ -42,6 +43,9 @@ Path dirOf(const Path & path); following the final `/'. */ string baseNameOf(const Path & path); +/* Get status of `path'. */ +struct stat lstat(const Path & path); + /* Return true iff the given path exists. */ bool pathExists(const Path & path); @@ -60,7 +64,7 @@ string readFile(int fd); string readFile(const Path & path); /* Write a string to a file. */ -void writeFile(const Path & path, const string & s, bool doFsync = false); +void writeFile(const Path & path, const string & s); /* Read a line from a file descriptor. */ string readLine(int fd); @@ -280,6 +284,11 @@ MakeError(Interrupted, BaseError) Strings tokenizeString(const string & s, const string & separators = " \t\n\r"); +/* Concatenate the given strings with a separator between the + elements. */ +string concatStringsSep(const string & sep, const Strings & ss); + + /* Convert the exit status of a child as returned by wait() into an error string. */ string statusToString(int status); diff --git a/src/nix-env/Makefile.am b/src/nix-env/Makefile.am index 7dfa7425a062..113baabc41cd 100644 --- a/src/nix-env/Makefile.am +++ b/src/nix-env/Makefile.am @@ -4,7 +4,7 @@ nix_env_SOURCES = nix-env.cc profiles.cc profiles.hh user-env.cc user-env.hh hel nix_env_LDADD = ../libmain/libmain.la ../libexpr/libexpr.la \ ../libstore/libstore.la ../libutil/libutil.la \ - ../boost/format/libformat.la @ADDITIONAL_NETWORK_LIBS@ + ../boost/format/libformat.la nix-env.o: help.txt.hh diff --git a/src/nix-env/nix-env.cc b/src/nix-env/nix-env.cc index 58bc9af12598..bfcc25872715 100644 --- a/src/nix-env/nix-env.cc +++ b/src/nix-env/nix-env.cc @@ -69,7 +69,7 @@ typedef void (* Operation) (Globals & globals, void printHelp() { - cout << string((char *) helpText, sizeof helpText); + cout << string((char *) helpText); } diff --git a/src/nix-hash/Makefile.am b/src/nix-hash/Makefile.am index 5f84eb34d6da..a4fdb324629d 100644 --- a/src/nix-hash/Makefile.am +++ b/src/nix-hash/Makefile.am @@ -2,7 +2,7 @@ bin_PROGRAMS = nix-hash nix_hash_SOURCES = nix-hash.cc help.txt nix_hash_LDADD = ../libmain/libmain.la ../libstore/libstore.la ../libutil/libutil.la \ - ../boost/format/libformat.la @ADDITIONAL_NETWORK_LIBS@ + ../boost/format/libformat.la nix-hash.o: help.txt.hh diff --git a/src/nix-hash/nix-hash.cc b/src/nix-hash/nix-hash.cc index f268e4986701..4867234bffc1 100644 --- a/src/nix-hash/nix-hash.cc +++ b/src/nix-hash/nix-hash.cc @@ -10,7 +10,7 @@ using namespace nix; void printHelp() { - std::cout << string((char *) helpText, sizeof helpText); + std::cout << string((char *) helpText); } @@ -44,7 +44,7 @@ void run(Strings args) if (op == opHash) { for (Strings::iterator i = ss.begin(); i != ss.end(); ++i) { - Hash h = flat ? hashFile(ht, *i) : hashPath(ht, *i); + Hash h = flat ? hashFile(ht, *i) : hashPath(ht, *i).first; if (truncate && h.hashSize > 20) h = compressHash(h, 20); std::cout << format("%1%\n") % (base32 ? printHash32(h) : printHash(h)); diff --git a/src/nix-instantiate/Makefile.am b/src/nix-instantiate/Makefile.am index a65907a8d40a..b48dbd9d410f 100644 --- a/src/nix-instantiate/Makefile.am +++ b/src/nix-instantiate/Makefile.am @@ -3,7 +3,7 @@ bin_PROGRAMS = nix-instantiate nix_instantiate_SOURCES = nix-instantiate.cc help.txt nix_instantiate_LDADD = ../libmain/libmain.la ../libexpr/libexpr.la \ ../libstore/libstore.la ../libutil/libutil.la \ - ../boost/format/libformat.la @ADDITIONAL_NETWORK_LIBS@ + ../boost/format/libformat.la nix-instantiate.o: help.txt.hh diff --git a/src/nix-instantiate/nix-instantiate.cc b/src/nix-instantiate/nix-instantiate.cc index b330f0cf11f8..3d3b643cfd6e 100644 --- a/src/nix-instantiate/nix-instantiate.cc +++ b/src/nix-instantiate/nix-instantiate.cc @@ -19,7 +19,7 @@ using namespace nix; void printHelp() { - std::cout << string((char *) helpText, sizeof helpText); + std::cout << string((char *) helpText); } diff --git a/src/nix-log2xml/log2xml.cc b/src/nix-log2xml/log2xml.cc index b2a25eefa269..6645dc500fed 100644 --- a/src/nix-log2xml/log2xml.cc +++ b/src/nix-log2xml/log2xml.cc @@ -18,6 +18,8 @@ struct Decoder int priority; bool ignoreLF; int lineNo, charNo; + bool warning; + bool error; Decoder() { @@ -29,6 +31,8 @@ struct Decoder ignoreLF = false; lineNo = 1; charNo = 0; + warning = false; + error = false; } void pushChar(char c); @@ -95,6 +99,12 @@ void Decoder::pushChar(char c) case 'b': ignoreLF = false; break; + case 'e': + error = true; + break; + case 'w': + warning = true; + break; } } else if (c >= '0' && c <= '9') { int n = 0; @@ -118,6 +128,8 @@ void Decoder::finishLine() string tag = inHeader ? "head" : "line"; cout << "<" << tag; if (priority != 1) cout << " priority='" << priority << "'"; + if (warning) cout << " warning='1'"; + if (error) cout << " error='1'"; cout << ">"; for (unsigned int i = 0; i < line.size(); i++) { @@ -158,6 +170,8 @@ void Decoder::finishLine() line = ""; inHeader = false; priority = 1; + warning = false; + error = false; } diff --git a/src/nix-setuid-helper/Makefile.am b/src/nix-setuid-helper/Makefile.am index 35528458c5f6..a04701636d84 100644 --- a/src/nix-setuid-helper/Makefile.am +++ b/src/nix-setuid-helper/Makefile.am @@ -4,5 +4,4 @@ nix_setuid_helper_SOURCES = nix-setuid-helper.cc nix_setuid_helper_LDADD = ../libutil/libutil.la \ ../boost/format/libformat.la -AM_CXXFLAGS = \ - -I$(srcdir)/.. -I$(srcdir)/../libutil +AM_CXXFLAGS = -I$(srcdir)/.. -I$(srcdir)/../libutil diff --git a/src/nix-store/Makefile.am b/src/nix-store/Makefile.am index b6dd37a6114a..44ff54674ade 100644 --- a/src/nix-store/Makefile.am +++ b/src/nix-store/Makefile.am @@ -5,7 +5,7 @@ nix_store_SOURCES = \ xmlgraph.cc xmlgraph.hh nix_store_LDADD = ../libmain/libmain.la ../libstore/libstore.la ../libutil/libutil.la \ - ../boost/format/libformat.la @ADDITIONAL_NETWORK_LIBS@ + ../boost/format/libformat.la nix-store.o: help.txt.hh diff --git a/src/nix-store/help.txt b/src/nix-store/help.txt index 071934b2000f..4782518517dd 100644 --- a/src/nix-store/help.txt +++ b/src/nix-store/help.txt @@ -16,7 +16,7 @@ Operations: --gc: run the garbage collector - --dump: dump a path as a Nix archive, forgetting dependencies + --dump: dump a path as a Nix archive (NAR), forgetting dependencies --restore: restore a path from a Nix archive, without registering validity @@ -27,6 +27,9 @@ Operations: --verify: verify Nix structures --optimise: optimise the Nix store by hard-linking identical files + --query-failed-paths: list paths that failed to build (if enabled) + --clear-failed-paths: clear the failed status of the given paths + --version: output version information --help: display help @@ -41,6 +44,7 @@ Query flags: --graph: print a dot graph rooted at given path --xml: emit an XML representation of the graph rooted at the given path --hash: print the SHA-256 hash of the contents of the path + --size: print the size of the NAR dump of the path --roots: print the garbage collector roots that point to the path Query switches (not applicable to all queries): diff --git a/src/nix-store/nix-store.cc b/src/nix-store/nix-store.cc index ff6538137a8a..49a705585ebf 100644 --- a/src/nix-store/nix-store.cc +++ b/src/nix-store/nix-store.cc @@ -22,7 +22,7 @@ typedef void (* Operation) (Strings opFlags, Strings opArgs); void printHelp() { - cout << string((char *) helpText, sizeof helpText); + cout << string((char *) helpText); } @@ -34,7 +34,7 @@ static bool indirectRoot = false; LocalStore & ensureLocalStore() { LocalStore * store2(dynamic_cast<LocalStore *>(store.get())); - if (!store2) throw Error("you don't have sufficient rights to use --verify"); + if (!store2) throw Error("you don't have sufficient rights to use this command"); return *store2; } @@ -226,7 +226,7 @@ static void printTree(const Path & path, static void opQuery(Strings opFlags, Strings opArgs) { enum { qOutputs, qRequisites, qReferences, qReferrers - , qReferrersClosure, qDeriver, qBinding, qHash + , qReferrersClosure, qDeriver, qBinding, qHash, qSize , qTree, qGraph, qXml, qResolve, qRoots } query = qOutputs; bool useOutput = false; bool includeOutputs = false; @@ -248,6 +248,7 @@ static void opQuery(Strings opFlags, Strings opArgs) query = qBinding; } else if (*i == "--hash") query = qHash; + else if (*i == "--size") query = qSize; else if (*i == "--tree") query = qTree; else if (*i == "--graph") query = qGraph; else if (*i == "--xml") query = qXml; @@ -310,11 +311,15 @@ static void opQuery(Strings opFlags, Strings opArgs) break; case qHash: + case qSize: foreach (Strings::iterator, i, opArgs) { Path path = maybeUseOutput(followLinksToStorePath(*i), useOutput, forceRealise); - Hash hash = store->queryPathHash(path); - assert(hash.type == htSHA256); - cout << format("sha256:%1%\n") % printHash32(hash); + ValidPathInfo info = store->queryPathInfo(path); + if (query == qHash) { + assert(info.hash.type == htSHA256); + cout << format("sha256:%1%\n") % printHash32(info.hash); + } else if (query == qSize) + cout << format("%1%\n") % info.narSize; } break; @@ -393,9 +398,8 @@ static void opDumpDB(Strings opFlags, Strings opArgs) if (!opArgs.empty()) throw UsageError("no arguments expected"); PathSet validPaths = store->queryValidPaths(); - foreach (PathSet::iterator, i, validPaths) { - cout << makeValidityRegistration(singleton<PathSet>(*i), true, true); - } + foreach (PathSet::iterator, i, validPaths) + cout << store->makeValidityRegistration(singleton<PathSet>(*i), true, true); } @@ -410,8 +414,11 @@ static void registerValidity(bool reregister, bool hashGiven, bool canonicalise) /* !!! races */ if (canonicalise) canonicalisePathMetaData(info.path); - if (!hashGiven) - info.hash = hashPath(htSHA256, info.path); + if (!hashGiven) { + HashResult hash = hashPath(htSHA256, info.path); + info.hash = hash.first; + info.narSize = hash.second; + } infos.push_back(info); } } @@ -661,8 +668,7 @@ static void opOptimise(Strings opFlags, Strings opArgs) bool dryRun = false; - for (Strings::iterator i = opFlags.begin(); - i != opFlags.end(); ++i) + foreach (Strings::iterator, i, opFlags) if (*i == "--dry-run") dryRun = true; else throw UsageError(format("unknown flag `%1%'") % *i); @@ -677,6 +683,24 @@ static void opOptimise(Strings opFlags, Strings opArgs) } +static void opQueryFailedPaths(Strings opFlags, Strings opArgs) +{ + if (!opArgs.empty() || !opFlags.empty()) + throw UsageError("no arguments expected"); + PathSet failed = store->queryFailedPaths(); + foreach (PathSet::iterator, i, failed) + cout << format("%1%\n") % *i; +} + + +static void opClearFailedPaths(Strings opFlags, Strings opArgs) +{ + if (!opFlags.empty()) + throw UsageError("no flags expected"); + store->clearFailedPaths(PathSet(opArgs.begin(), opArgs.end())); +} + + /* Scan the arguments; find the operation, set global flags, put all other flags in a list, and put all other arguments in another list. */ @@ -728,6 +752,10 @@ void run(Strings args) op = opVerify; else if (arg == "--optimise") op = opOptimise; + else if (arg == "--query-failed-paths") + op = opQueryFailedPaths; + else if (arg == "--clear-failed-paths") + op = opClearFailedPaths; else if (arg == "--add-root") { if (i == args.end()) throw UsageError("`--add-root requires an argument"); diff --git a/src/nix-worker/Makefile.am b/src/nix-worker/Makefile.am index 50c8ae36d2e8..b6094a2a038c 100644 --- a/src/nix-worker/Makefile.am +++ b/src/nix-worker/Makefile.am @@ -2,7 +2,7 @@ bin_PROGRAMS = nix-worker nix_worker_SOURCES = nix-worker.cc help.txt nix_worker_LDADD = ../libmain/libmain.la ../libstore/libstore.la ../libutil/libutil.la \ - ../boost/format/libformat.la @ADDITIONAL_NETWORK_LIBS@ + ../boost/format/libformat.la nix-worker.o: help.txt.hh diff --git a/src/nix-worker/nix-worker.cc b/src/nix-worker/nix-worker.cc index 13cecf654e7a..0fa1b40aede9 100644 --- a/src/nix-worker/nix-worker.cc +++ b/src/nix-worker/nix-worker.cc @@ -178,7 +178,7 @@ static void startWork() /* stopWork() means that we're done; stop sending stderr to the client. */ -static void stopWork(bool success = true, const string & msg = "") +static void stopWork(bool success = true, const string & msg = "", unsigned int status = 0) { /* Stop handling async client death; we're going to a state where we're either sending or receiving from the client, so we'll be @@ -192,6 +192,7 @@ static void stopWork(bool success = true, const string & msg = "") else { writeInt(STDERR_ERROR, to); writeString(msg, to); + if (status != 0) writeInt(status, to); } } @@ -315,14 +316,16 @@ static void performOp(unsigned int clientVersion, } case wopQueryReferences: - case wopQueryReferrers: { + case wopQueryReferrers: + case wopQueryDerivationOutputs: { Path path = readStorePath(from); startWork(); PathSet paths; if (op == wopQueryReferences) store->queryReferences(path, paths); - else + else if (op == wopQueryReferrers) store->queryReferrers(path, paths); + else paths = store->queryDerivationOutputs(path); stopWork(); writeStringSet(paths, to); break; @@ -519,10 +522,50 @@ static void performOp(unsigned int clientVersion, writeString(info.deriver, to); writeStringSet(info.references, to); writeLongLong(info.downloadSize, to); + if (GET_PROTOCOL_MINOR(clientVersion) >= 7) + writeLongLong(info.narSize, to); } break; } + case wopQueryValidPaths: { + startWork(); + PathSet paths = store->queryValidPaths(); + stopWork(); + writeStringSet(paths, to); + break; + } + + case wopQueryFailedPaths: { + startWork(); + PathSet paths = store->queryFailedPaths(); + stopWork(); + writeStringSet(paths, to); + break; + } + + case wopClearFailedPaths: { + PathSet paths = readStringSet(from); + startWork(); + store->clearFailedPaths(paths); + stopWork(); + writeInt(1, to); + break; + } + + case wopQueryPathInfo: { + Path path = readStorePath(from); + startWork(); + ValidPathInfo info = store->queryPathInfo(path); + stopWork(); + writeString(info.deriver, to); + writeString(printHash(info.hash), to); + writeStringSet(info.references, to); + writeInt(info.registrationTime, to); + writeLongLong(info.narSize, to); + break; + } + default: throw Error(format("invalid operation %1%") % op); } @@ -595,7 +638,7 @@ static void processConnection() try { performOp(clientVersion, from, to, op); } catch (Error & e) { - stopWork(false, e.msg()); + stopWork(false, e.msg(), GET_PROTOCOL_MINOR(clientVersion) >= 8 ? e.status : 0); } assert(!canSendStderr); @@ -670,9 +713,8 @@ static void daemonLoop() while (1) { try { - /* Important: the server process *cannot* open the - Berkeley DB environment, because it doesn't like forks - very much. */ + /* Important: the server process *cannot* open the SQLite + database, because it doesn't like forks very much. */ assert(!store); /* Accept a connection. */ @@ -770,7 +812,7 @@ void run(Strings args) void printHelp() { - std::cout << string((char *) helpText, sizeof helpText); + std::cout << string((char *) helpText); } diff --git a/substitute.mk b/substitute.mk index f502b942d66e..d83399cfde78 100644 --- a/substitute.mk +++ b/substitute.mk @@ -24,6 +24,7 @@ -e "s^@xmllint\@^$(xmllint)^g" \ -e "s^@xmlflags\@^$(xmlflags)^g" \ -e "s^@xsltproc\@^$(xsltproc)^g" \ + -e "s^@sqlite_bin\@^$(sqlite_bin)^g" \ -e "s^@version\@^$(VERSION)^g" \ -e "s^@testPath\@^$(coreutils):$$(dirname $$(type -p expr))^g" \ < $< > $@ || rm $@ diff --git a/tests/Makefile.am b/tests/Makefile.am index beb7852db31c..fb4a2285c1d5 100644 --- a/tests/Makefile.am +++ b/tests/Makefile.am @@ -7,7 +7,8 @@ TESTS = init.sh hash.sh lang.sh add.sh simple.sh dependencies.sh \ fallback.sh nix-push.sh gc.sh gc-concurrent.sh verify.sh nix-pull.sh \ referrers.sh user-envs.sh logging.sh nix-build.sh misc.sh fixed.sh \ gc-runtime.sh install-package.sh check-refs.sh filter-source.sh \ - remote-store.sh export.sh export-graph.sh negative-caching.sh + remote-store.sh export.sh export-graph.sh negative-caching.sh \ + binary-patching.sh XFAIL_TESTS = @@ -31,5 +32,6 @@ EXTRA_DIST = $(TESTS) \ filter-source.nix \ export-graph.nix \ negative-caching.nix \ + binary-patching.nix \ $(wildcard lang/*.nix) $(wildcard lang/*.exp) $(wildcard lang/*.exp.xml) $(wildcard lang/*.flags) \ common.sh.in diff --git a/tests/binary-patching.nix b/tests/binary-patching.nix new file mode 100644 index 000000000000..0de3028bd64d --- /dev/null +++ b/tests/binary-patching.nix @@ -0,0 +1,18 @@ +{ version }: + +with import ./config.nix; + +mkDerivation { + name = "foo-${toString version}"; + builder = builtins.toFile "builder.sh" + '' + mkdir $out + seq 1 1000000 > $out/foo + ${if version != 1 then '' + seq 1000000 1010000 >> $out/foo + '' else ""} + ${if version == 3 then '' + echo foobar >> $out/foo + '' else ""} + ''; +} diff --git a/tests/binary-patching.sh b/tests/binary-patching.sh new file mode 100644 index 000000000000..60e57b4b00dc --- /dev/null +++ b/tests/binary-patching.sh @@ -0,0 +1,58 @@ +source common.sh + +clearManifests + +mkdir -p $TEST_ROOT/cache2 $TEST_ROOT/patches + +RESULT=$TEST_ROOT/result + +# Build version 1 and 2 of the "foo" package. +$NIX_BIN_DIR/nix-push --copy $TEST_ROOT/cache2 $TEST_ROOT/manifest1 \ + $($nixbuild -o $RESULT binary-patching.nix --arg version 1) + +out2=$($nixbuild -o $RESULT binary-patching.nix --arg version 2) +$NIX_BIN_DIR/nix-push --copy $TEST_ROOT/cache2 $TEST_ROOT/manifest2 $out2 + +out3=$($nixbuild -o $RESULT binary-patching.nix --arg version 3) +$NIX_BIN_DIR/nix-push --copy $TEST_ROOT/cache2 $TEST_ROOT/manifest3 $out3 + +rm $RESULT + +# Generate binary patches. +$NIX_BIN_DIR/nix-generate-patches $TEST_ROOT/cache2 $TEST_ROOT/patches \ + file://$TEST_ROOT/patches $TEST_ROOT/manifest1 $TEST_ROOT/manifest2 + +$NIX_BIN_DIR/nix-generate-patches $TEST_ROOT/cache2 $TEST_ROOT/patches \ + file://$TEST_ROOT/patches $TEST_ROOT/manifest2 $TEST_ROOT/manifest3 + +grep -q "patch {" $TEST_ROOT/manifest3 + +# Get rid of versions 2 and 3. +$nixstore --delete $out2 $out3 + +# Pull the manifest containing the patches. +clearManifests +$NIX_BIN_DIR/nix-pull file://$TEST_ROOT/manifest3 + +# Make sure that the download size prediction uses the patches rather +# than the full download. +$nixbuild -o $RESULT binary-patching.nix --arg version 3 --dry-run 2>&1 | grep -q "0.01 MiB" + +# Now rebuild it. This should use the two patches generated above. +rm -f $TEST_ROOT/var/log/nix/downloads +$nixbuild -o $RESULT binary-patching.nix --arg version 3 +rm $RESULT +[ "$(grep ' patch ' $TEST_ROOT/var/log/nix/downloads | wc -l)" -eq 2 ] + +# Add a patch from version 1 directly to version 3. +$NIX_BIN_DIR/nix-generate-patches $TEST_ROOT/cache2 $TEST_ROOT/patches \ + file://$TEST_ROOT/patches $TEST_ROOT/manifest1 $TEST_ROOT/manifest3 + +# Rebuild version 3. This should use the direct patch rather than the +# sequence of two patches. +$nixstore --delete $out2 $out3 +clearManifests +rm $TEST_ROOT/var/log/nix/downloads +$NIX_BIN_DIR/nix-pull file://$TEST_ROOT/manifest3 +$nixbuild -o $RESULT binary-patching.nix --arg version 3 +[ "$(grep ' patch ' $TEST_ROOT/var/log/nix/downloads | wc -l)" -eq 1 ] diff --git a/tests/build-hook.hook.sh b/tests/build-hook.hook.sh index 83fa3bf78757..ff709985d3af 100755 --- a/tests/build-hook.hook.sh +++ b/tests/build-hook.hook.sh @@ -2,20 +2,22 @@ #set -x -drv=$4 +while read x y drv rest; do -echo "HOOK for $drv" >&2 + echo "HOOK for $drv" >&2 -outPath=`sed 's/Derive(\[("out",\"\([^\"]*\)\".*/\1/' $drv` + outPath=`sed 's/Derive(\[("out",\"\([^\"]*\)\".*/\1/' $drv` -echo "output path is $outPath" >&2 + echo "output path is $outPath" >&2 -if `echo $outPath | grep -q input-1`; then - echo "# accept" >&2 - read x - echo "got $x" - mkdir $outPath - echo "BAR" > $outPath/foo -else - echo "# decline" >&2 -fi + if `echo $outPath | grep -q input-1`; then + echo "# accept" >&2 + read inputs + read outputs + mkdir $outPath + echo "BAR" > $outPath/foo + else + echo "# decline" >&2 + fi + +done \ No newline at end of file diff --git a/tests/common.sh.in b/tests/common.sh.in index b30bc44b5cc0..2c47d75eea23 100644 --- a/tests/common.sh.in +++ b/tests/common.sh.in @@ -38,6 +38,7 @@ export dot=@dot@ export xmllint="@xmllint@" export xmlflags="@xmlflags@" export xsltproc="@xsltproc@" +export sqlite3="@sqlite_bin@/bin/sqlite3" export SHELL="@shell@" export version=@version@ diff --git a/tests/init.sh b/tests/init.sh index 0639a70662cd..104da432eaef 100644 --- a/tests/init.sh +++ b/tests/init.sh @@ -23,28 +23,33 @@ ln -s $nixinstantiate $NIX_BIN_DIR/ ln -s $nixhash $NIX_BIN_DIR/ ln -s $nixenv $NIX_BIN_DIR/ ln -s $nixworker $NIX_BIN_DIR/ +ln -s $TOP/src/bsdiff-*/bsdiff $NIX_BIN_DIR/ +ln -s $TOP/src/bsdiff-*/bspatch $NIX_BIN_DIR/ ln -s $TOP/scripts/nix-prefetch-url $NIX_BIN_DIR/ ln -s $TOP/scripts/nix-collect-garbage $NIX_BIN_DIR/ ln -s $TOP/scripts/nix-build $NIX_BIN_DIR/ ln -s $TOP/scripts/nix-install-package $NIX_BIN_DIR/ ln -s $TOP/scripts/nix-push $NIX_BIN_DIR/ ln -s $TOP/scripts/nix-pull $NIX_BIN_DIR/ +ln -s $TOP/scripts/nix-generate-patches $NIX_BIN_DIR/ mkdir $NIX_BIN_DIR/nix ln -s $bzip2_bin_test/bzip2 $NIX_BIN_DIR/nix/ ln -s $bzip2_bin_test/bunzip2 $NIX_BIN_DIR/nix/ ln -s $TOP/scripts/copy-from-other-stores.pl $NIX_BIN_DIR/nix/ ln -s $TOP/scripts/download-using-manifests.pl $NIX_BIN_DIR/nix/ -ln -s $TOP/scripts/readmanifest.pm $NIX_BIN_DIR/nix/ +ln -s $TOP/scripts/GeneratePatches.pm $NIX_BIN_DIR/nix/ +ln -s $TOP/scripts/NixManifest.pm $NIX_BIN_DIR/nix/ cat > "$NIX_CONF_DIR"/nix.conf <<EOF gc-keep-outputs = false gc-keep-derivations = false env-keep-derivations = false +fsync-metadata = false EOF mkdir $NIX_DATA_DIR/nix cp -pr $TOP/corepkgs $NIX_DATA_DIR/nix/ -# Bah, script has the prefix hard-coded. This is really messy stuff +# Bah, scripts have the prefix hard-coded. This is really messy stuff # (and likely to fail). for i in \ $NIX_DATA_DIR/nix/corepkgs/nar/nar.sh \ @@ -56,7 +61,9 @@ for i in \ $NIX_BIN_DIR/nix-install-package \ $NIX_BIN_DIR/nix-push \ $NIX_BIN_DIR/nix-pull \ - $NIX_BIN_DIR/nix/readmanifest.pm \ + $NIX_BIN_DIR/nix-generate-patches \ + $NIX_BIN_DIR/nix/NixManifest.pm \ + $NIX_BIN_DIR/nix/GeneratePatches.pm \ ; do sed < $i > $i.tmp \ -e "s^$REAL_BIN_DIR/nix-store^$NIX_BIN_DIR/nix-store^" \ @@ -96,7 +103,6 @@ mv $NIX_BIN_DIR/nix/download-using-manifests.pl $NIX_BIN_DIR/nix/substituters/do $nixstore --init # Did anything happen? -test -e "$NIX_DB_DIR"/info -test -e "$NIX_DB_DIR"/referrer +test -e "$NIX_DB_DIR"/db.sqlite echo 'Hello World' > ./dummy diff --git a/tests/lang/eval-okay-substring.exp b/tests/lang/eval-okay-substring.exp index d936b7e96f94..6aace04b0f57 100644 --- a/tests/lang/eval-okay-substring.exp +++ b/tests/lang/eval-okay-substring.exp @@ -1 +1 @@ -"ooxfoobarybarzobaabb" +"ooxfoobarybarzobaabbc" diff --git a/tests/lang/eval-okay-substring.nix b/tests/lang/eval-okay-substring.nix index 184d72580c51..424af00d9b3b 100644 --- a/tests/lang/eval-okay-substring.nix +++ b/tests/lang/eval-okay-substring.nix @@ -17,3 +17,5 @@ substring 1 2 s + substring 3 0 s + "b" + substring 3 1 s ++ "c" ++ substring 5 10 "perl" diff --git a/tests/nix-push.sh b/tests/nix-push.sh index a9a2f5f9c06a..0a35e3b97ab7 100644 --- a/tests/nix-push.sh +++ b/tests/nix-push.sh @@ -5,7 +5,7 @@ outPath=$($nixstore -r $drvPath) echo "pushing $drvPath" -mkdir $TEST_ROOT/cache +mkdir -p $TEST_ROOT/cache $NIX_BIN_DIR/nix-push \ --copy $TEST_ROOT/cache $TEST_ROOT/manifest $drvPath diff --git a/tests/referrers.sh b/tests/referrers.sh index aa5edb555139..a0c195d5ab6c 100644 --- a/tests/referrers.sh +++ b/tests/referrers.sh @@ -1,9 +1,8 @@ source common.sh -# This takes way to long on Cygwin (because process creation is so slow...). -if test "$system" = i686-cygwin; then exit 0; fi +clearStore -max=1000 +max=500 reference=$NIX_STORE_DIR/abcdef touch $reference @@ -13,46 +12,23 @@ echo "making registration..." for ((n = 0; n < $max; n++)); do storePath=$NIX_STORE_DIR/$n - touch $storePath + echo -n > $storePath ref2=$NIX_STORE_DIR/$((n+1)) if test $((n+1)) = $max; then ref2=$reference fi - (echo $storePath && echo && echo 2 && echo $reference && echo $ref2) + echo $storePath; echo; echo 2; echo $reference; echo $ref2 done > $TEST_ROOT/reg_info echo "registering..." -time $nixstore --register-validity < $TEST_ROOT/reg_info - -oldTime=$(cat test-tmp/db/info/1 | grep Registered-At) - -echo "sleeping..." - -sleep 2 - -echo "reregistering..." - -time $nixstore --register-validity --reregister < $TEST_ROOT/reg_info - -newTime=$(cat test-tmp/db/info/1 | grep Registered-At) - -if test "$newTime" != "$oldTime"; then - echo "reregistration changed original registration time" - exit 1 -fi - -if test "$(cat test-tmp/db/referrer/1 | wc -w)" -ne 1; then - echo "reregistration duplicated referrers" - exit 1 -fi +$nixstore --register-validity < $TEST_ROOT/reg_info echo "collecting garbage..." ln -sfn $reference "$NIX_STATE_DIR"/gcroots/ref -time $nixstore --gc +$nixstore --gc -if test "$(cat test-tmp/db/referrer/abcdef | wc -w)" -ne 0; then +if test "$(sqlite3 ./test-tmp/db/db.sqlite 'select count(*) from Refs')" -ne 0; then echo "referrers not cleaned up" exit 1 fi - diff --git a/tests/substituter.sh b/tests/substituter.sh index 94ae26472501..65c4fa1856d0 100755 --- a/tests/substituter.sh +++ b/tests/substituter.sh @@ -16,7 +16,8 @@ if test $1 = "--query"; then echo 1 echo "" # deriver echo 0 # nr of refs - echo 0 # download size + echo $((1 * 1024 * 1024)) # download size + echo $((2 * 1024 * 1024)) # nar size else echo "bad command $cmd" exit 1 diff --git a/tests/substituter2.sh b/tests/substituter2.sh index 70f7d0943b98..c56a1bc47b5c 100755 --- a/tests/substituter2.sh +++ b/tests/substituter2.sh @@ -16,6 +16,7 @@ if test $1 = "--query"; then echo "" # deriver echo 0 # nr of refs echo 0 # download size + echo 0 # nar size else echo "bad command $cmd" exit 1 diff --git a/tests/substitutes.sh b/tests/substitutes.sh index 042827314bd9..b48576c8c302 100644 --- a/tests/substitutes.sh +++ b/tests/substitutes.sh @@ -14,6 +14,8 @@ echo $outPath > $TEST_ROOT/sub-paths export NIX_SUBSTITUTERS=$(pwd)/substituter.sh +$nixstore -r "$drvPath" --dry-run 2>&1 | grep -q "1.00 MiB.*2.00 MiB" + $nixstore -rvv "$drvPath" text=$(cat "$outPath"/hello) |