<section xmlns="http://docbook.org/ns/docbook"
xmlns:xlink="http://www.w3.org/1999/xlink"
xml:id="sec-conf-file">
<title>Nix configuration file</title>
<para>A number of persistent settings of Nix are stored in the file
<filename><replaceable>sysconfdir</replaceable>/nix/nix.conf</filename>.
This file is a list of <literal><replaceable>name</replaceable> =
<replaceable>value</replaceable></literal> pairs, one per line.
Comments start with a <literal>#</literal> character. An example
configuration file is shown in <xref linkend="ex-nix-conf" />.</para>
<example xml:id='ex-nix-conf'><title>Nix configuration file</title>
<programlisting>
gc-keep-outputs = true # Nice for developers
gc-keep-derivations = true # Idem
env-keep-derivations = false
</programlisting>
</example>
<para>The following variables are currently available:
<variablelist>
<varlistentry xml:id="conf-gc-keep-outputs"><term><literal>gc-keep-outputs</literal></term>
<listitem><para>If <literal>true</literal>, the garbage collector
will keep the outputs of non-garbage derivations. If
<literal>false</literal> (default), outputs will be deleted unless
they are GC roots themselves (or reachable from other roots).</para>
<para>In general, outputs must be registered as roots separately.
However, even if the output of a derivation is registered as a
root, the collector will still delete store paths that are used
only at build time (e.g., the C compiler, or source tarballs
downloaded from the network). To prevent it from doing so, set
this option to <literal>true</literal>.</para></listitem>
</varlistentry>
<varlistentry xml:id="conf-gc-keep-derivations"><term><literal>gc-keep-derivations</literal></term>
<listitem><para>If <literal>true</literal> (default), the garbage
collector will keep the derivations from which non-garbage store
paths were built. If <literal>false</literal>, they will be
deleted unless explicitly registered as a root (or reachable from
other roots).</para>
<para>Keeping derivation around is useful for querying and
traceability (e.g., it allows you to ask with what dependencies or
options a store path was built), so by default this option is on.
Turn it off to safe a bit of disk space (or a lot if
<literal>gc-keep-outputs</literal> is also turned on).</para></listitem>
</varlistentry>
<varlistentry><term><literal>env-keep-derivations</literal></term>
<listitem><para>If <literal>false</literal> (default), derivations
are not stored in Nix user environments. That is, the derivation
any build-time-only dependencies may be garbage-collected.</para>
<para>If <literal>true</literal>, when you add a Nix derivation to
a user environment, the path of the derivation is stored in the
user environment. Thus, the derivation will not be
garbage-collected until the user environment generation is deleted
(<command>nix-env --delete-generations</command>). To prevent
build-time-only dependencies from being collected, you should also
turn on <literal>gc-keep-outputs</literal>.</para>
<para>The difference between this option and
<literal>gc-keep-derivations</literal> is that this one is
“sticky”: it applies to any user environment created while this
option was enabled, while <literal>gc-keep-derivations</literal>
only applies at the moment the garbage collector is
run.</para></listitem>
</varlistentry>
<varlistentry xml:id="conf-build-max-jobs"><term><literal>build-max-jobs</literal></term>
<listitem><para>This option defines the maximum number of jobs
that Nix will try to build in parallel. The default is
<literal>1</literal>. You should generally set it to the number
of CPUs in your system (e.g., <literal>2</literal> on a Athlon 64
X2). It can be overriden using the <option
linkend='opt-max-jobs'>--max-jobs</option> (<option>-j</option>)
command line switch.</para></listitem>
</varlistentry>
<varlistentry xml:id="conf-build-cores"><term><literal>build-cores</literal></term>
<listitem><para>Sets the value of the
<envar>NIX_BUILD_CORES</envar> environment variable in the
invocation of builders. Builders can use this variable at their
discretion to control the maximum amount of parallelism. For
instance, in Nixpkgs, if the derivation attribute
<varname>enableParallelBuilding</varname> is set to
<literal>true</literal>, the builder passes the
<option>-j<replaceable>N</replaceable></option> flag to GNU Make.
It can be overriden using the <option
linkend='opt-cores'>--cores</option> command line switch and
defaults to <literal>1</literal>. The value <literal>0</literal>
means that the builder should use all available CPU cores in the
system.</para></listitem>
</varlistentry>
<varlistentry xml:id="conf-build-max-silent-time"><term><literal>build-max-silent-time</literal></term>
<listitem>
<para>This option defines the maximum number of seconds that a
builder can go without producing any data on standard output or
standard error. This is useful (for instance in a automated
build system) to catch builds that are stuck in an infinite
loop, or to catch remote builds that are hanging due to network
problems. It can be overriden using the <option
linkend="opt-max-silent-time">--max-silent-time</option> command
line switch.</para>
<para>The value <literal>0</literal> means that there is no
timeout. This is also the default.</para>
</listitem>
</varlistentry>
<varlistentry xml:id="conf-build-timeout"><term><literal>build-timeout</literal></term>
<listitem>
<para>This option defines the maximum number of seconds that a
builder can run. This is useful (for instance in a automated
build system) to catch builds that are stuck in an infinite loop
but keep writing to their standard output or standard error. It
can be overriden using the <option
linkend="opt-timeout">--timeout</option> command line
switch.</para>
<para>The value <literal>0</literal> means that there is no
timeout. This is also the default.</para>
</listitem>
</varlistentry>
<varlistentry xml:id="conf-build-users-group"><term><literal>build-users-group</literal></term>
<listitem><para>This options specifies the Unix group containing
the Nix build user accounts. In multi-user Nix installations,
builds should not be performed by the Nix account since that would
allow users to arbitrarily modify the Nix store and database by
supplying specially crafted builders; and they cannot be performed
by the calling user since that would allow him/her to influence
the build result.</para>
<para>Therefore, if this option is non-empty and specifies a valid
group, builds will be performed under the user accounts that are a
member of the group specified here (as listed in
<filename>/etc/group</filename>). Those user accounts should not
be used for any other purpose!</para>
<para>Nix will never run two builds under the same user account at
the same time. This is to prevent an obvious security hole: a
malicious user writing a Nix expression that modifies the build
result of a legitimate Nix expression being built by another user.
Therefore it is good to have as many Nix build user accounts as
you can spare. (Remember: uids are cheap.)</para>
<para>The build users should have permission to create files in
the Nix store, but not delete them. Therefore,
<filename>/nix/store</filename> should be owned by the Nix
account, its group should be the group specified here, and its
mode should be <literal>1775</literal>.</para>
<para>If the build users group is empty, builds will be performed
under the uid of the Nix process (that is, the uid of the caller
if <envar>NIX_REMOTE</envar> is empty, the uid under which the Nix
daemon runs if <envar>NIX_REMOTE</envar> is
<literal>daemon</literal>, or the uid that owns the setuid
<command>nix-worker</command> program if <envar>NIX_REMOTE</envar>
is <literal>slave</literal>). Obviously, this should not be used
in multi-user settings with untrusted users.</para>
</listitem>
</varlistentry>
<varlistentry><term><literal>build-use-chroot</literal></term>
<listitem><para>If set to <literal>true</literal>, builds will be
performed in a <emphasis>chroot environment</emphasis>, i.e., the
build will be isolated from the normal file system hierarchy and
will only see the Nix store, the temporary build directory, and
the directories configured with the <link
linkend='conf-build-chroot-dirs'><literal>build-chroot-dirs</literal>
option</link> (such as <filename>/proc</filename> and
<filename>/dev</filename>). This is useful to prevent undeclared
dependencies on files in directories such as
<filename>/usr/bin</filename>.</para>
<para>The use of a chroot requires that Nix is run as root (but
you can still use the <link
linkend='conf-build-users-group'>“build users” feature</link> to
perform builds under different users than root). Currently,
chroot builds only work on Linux because Nix uses “bind mounts” to
make the Nix store and other directories available inside the
chroot.</para>
</listitem>
</varlistentry>
<varlistentry><term><literal>build-use-substitutes</literal></term>
<listitem><para>If set to <literal>true</literal> (default), Nix
will use binary substitutes if available. This option can be
disabled to force building from source.</para></listitem>
</varlistentry>
<varlistentry xml:id="conf-build-chroot-dirs"><term><literal>build-chroot-dirs</literal></term>
<listitem><para>When builds are performed in a chroot environment,
Nix will mount (using <command>mount --bind</command> on Linux)
some directories from the normal file system hierarchy inside the
chroot. These are the Nix store, the temporary build directory
(usually
<filename>/tmp/nix-<replaceable>pid</replaceable>-<replaceable>number</replaceable></filename>)
and the directories listed here. The default is <literal>dev
/proc</literal>. Files in <filename>/dev</filename> (such as
<filename>/dev/null</filename>) are needed by many builds, and
some files in <filename>/proc</filename> may also be needed
occasionally.</para>
<para>The value used on NixOS is
<programlisting>
build-use-chroot = /dev /proc /bin</programlisting>
to make the <filename>/bin/sh</filename> symlink available (which
is still needed by many builders).</para>
</listitem>
</varlistentry>
<varlistentry><term><literal>build-cache-failures</literal></term>
<listitem><para>If set to <literal>true</literal>, Nix will
“cache” build failures, meaning that it will remember (in its
database) that a derivation previously failed. If you then try to
build the derivation again, Nix will immediately fail rather than
perform the build again. Failures in fixed-output derivations
(such as <function>fetchurl</function> calls) are never cached.
The “failed” status of a derivation can be cleared using
<command>nix-store --clear-failed-paths</command>. By default,
failure caching is disabled.</para></listitem>
</varlistentry>
<varlistentry><term><literal>system</literal></term>
<listitem><para>This option specifies the canonical Nix system
name of the current installation, such as
<literal>i686-linux</literal> or
<literal>powerpc-darwin</literal>. Nix can only build derivations
whose <literal>system</literal> attribute equals the value
specified here. In general, it never makes sense to modify this
value from its default, since you can use it to ‘lie’ about the
platform you are building on (e.g., perform a Mac OS build on a
Linux machine; the result would obviously be wrong). It only
makes sense if the Nix binaries can run on multiple platforms,
e.g., ‘universal binaries’ that run on <literal>powerpc-darwin</literal> and
<literal>i686-darwin</literal>.</para>
<para>It defaults to the canonical Nix system name detected by
<filename>configure</filename> at build time.</para></listitem>
</varlistentry>
<varlistentry><term><literal>fsync-metadata</literal></term>
<listitem><para>If set to <literal>true</literal>, changes to the
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>true</literal>.</para></listitem>
</varlistentry>
</variablelist>
</para>
</section>
