about summary refs log tree commit diff
path: root/tvix/eval/src/compiler
AgeCommit message (Collapse)AuthorFilesLines
2022-09-16 r/4873 refactor(tvix/eval): move compile_inherit_attrs to compiler::attrsVincent Ambo2-62/+62
Plain move, no other changes. Change-Id: Ic4f89709f5c2cbc03182a848af080c820e39a0fd Reviewed-on: https://cl.tvl.fyi/c/depot/+/6607 Reviewed-by: sterni <sternenseemann@systemli.org> Tested-by: BuildkiteCI
2022-09-16 r/4872 refactor(tvix/eval): explicitly construct attrs in phasesVincent Ambo2-32/+101
This makes the phases of attribute set construction that Nix has very explicit (inherits, static keys, dynamic keys). This change focuses on the split between dynamic/static keys by collecting all dynamic ones while compiling the static ones, and then phasing them in afterwards. It's possible we also need to do some additional splitting inside of the inherits. Change-Id: Icae782e2a5c106e3ce0831dda47ed81c923c0a42 Reviewed-on: https://cl.tvl.fyi/c/depot/+/6530 Reviewed-by: sterni <sternenseemann@systemli.org> Tested-by: BuildkiteCI
2022-09-15 r/4861 refactor(tvix/eval): don't move parts Vec in compile_str_partssterni1-7/+5
This allows us to get rid of the count local variable which was a bit confusing. Calling parts.len() multiple times is fine, since the length doesn't need to be computed. Change-Id: I4f626729ad1bf23a93cb701385c3f4b50c57456d Reviewed-on: https://cl.tvl.fyi/c/depot/+/6584 Autosubmit: sterni <sternenseemann@systemli.org> Reviewed-by: tazjin <tazjin@tvl.su> Tested-by: BuildkiteCI
2022-09-15 r/4860 fix(tvix/eval): coerce string interpolation parts to stringsterni1-35/+45
With this puzzle piece of string compilation in place, `compile_str` becomes less redundant, as every part now needs to be compiled the same. The thunking logic becomes a bit trickier, since we need to thunk even in the case of `count == 1` if the single part is interpolating. Splitting the inner (shared) code in a separate function turned out to be easier for making rustc content. Change-Id: I6a554ca599926ae5907d7acffce349c9616f568f Reviewed-on: https://cl.tvl.fyi/c/depot/+/6582 Reviewed-by: tazjin <tazjin@tvl.su> Tested-by: BuildkiteCI
2022-09-15 r/4859 fix(tvix/eval): thunk string interpolationsterni1-18/+32
If we have multiple string parts, we need to thunk assembling the string. If we have a single literal, it is strict (like all literals), but a single interpolation part may compile to a thunk, depending on how the expression inside is compiled – we can avoid forcing to early here compared to the previous behavior. Note that this CL retains the bug that `"${x}"` is erroneously translated to `x`, implying e.g. `"${12}" == 12`. The use of `parts.len()` is unproblematic, since normalized_parts() builds a `Vec` instead of returning an iterator. Change-Id: I3aecbfefef65cc627b1b8a65be27cbaeada3582b Reviewed-on: https://cl.tvl.fyi/c/depot/+/6580 Autosubmit: sterni <sternenseemann@systemli.org> Reviewed-by: tazjin <tazjin@tvl.su> Tested-by: BuildkiteCI
2022-09-13 r/4849 fix(tvix/eval): force left argument of `?` before checking for attrssterni1-0/+2
OpAttrsIsSet and OpAttrsTrySelect will fail silently if the attribute set value on the stack is actually a thunk, so we need to make sure to force at every step of the way. Emitting the force instructions in the compiler because it is easier to add, but maybe the VM should do this when handling the relevant opcodes? Comments welcome. Change-Id: I65c5ef348d59b2d07c9bb06abb24f9f3e6a0fdb2 Reviewed-on: https://cl.tvl.fyi/c/depot/+/6540 Reviewed-by: grfn <grfn@gws.fyi> Autosubmit: sterni <sternenseemann@systemli.org> Tested-by: BuildkiteCI Reviewed-by: tazjin <tazjin@tvl.su>
2022-09-13 r/4847 fix(tvix/eval): force exprs inside string interpolationsterni1-1/+5
The expression inside ${…} may return arbitrary values, including thunks, so we need to make sure to force them just in case. Change-Id: Ic11ba00c4c92a10a83becd91233db5f57f6e59c8 Reviewed-on: https://cl.tvl.fyi/c/depot/+/6541 Autosubmit: sterni <sternenseemann@systemli.org> Reviewed-by: grfn <grfn@gws.fyi> Tested-by: BuildkiteCI Reviewed-by: tazjin <tazjin@tvl.su>
2022-09-13 r/4844 fix(tvix/eval): use fragment span for OpAttrsSelectVincent Ambo1-2/+2
Pointed out by sterni in cl/6389 Change-Id: I648056a760266a8cfd7adcdc478c7ff2132991f7 Reviewed-on: https://cl.tvl.fyi/c/depot/+/6568 Reviewed-by: sterni <sternenseemann@systemli.org> Tested-by: BuildkiteCI
2022-09-13 r/4843 refactor(tvix/eval): point `OpPushWith` span at namespaceVincent Ambo1-1/+1
Pointed out by sterni in cl/6395 Change-Id: I2dda2bb11fef702df05fd7a4fd93b9e717a85dad Reviewed-on: https://cl.tvl.fyi/c/depot/+/6567 Reviewed-by: sterni <sternenseemann@systemli.org> Tested-by: BuildkiteCI
2022-09-13 r/4842 refactor(tvix/eval): point `OpAssert` span at conditionVincent Ambo1-1/+1
This is more useful than pointing it at the entire assert expression, as that includes the body as well which is not going to be relevant in the error. Pointed out by sterni in cl/6391 Change-Id: I95a5d1edf90df65e7fa53d4d04502afd6e99e89a Reviewed-on: https://cl.tvl.fyi/c/depot/+/6566 Reviewed-by: sterni <sternenseemann@systemli.org> Tested-by: BuildkiteCI
2022-09-13 r/4841 chore(tvix/eval): do not inherit scope depth in new scopesVincent Ambo1-1/+0
This is no longer required, resolution is now more sane. Pointed out by sterni in cl/6422. Change-Id: Icc8983c648f864e66813948df6e2d4bad6a7f312 Reviewed-on: https://cl.tvl.fyi/c/depot/+/6565 Reviewed-by: sterni <sternenseemann@systemli.org> Tested-by: BuildkiteCI
2022-09-13 r/4840 refactor(tvix/eval): encapsulate scope_depth in compiler::scopeVincent Ambo3-15/+20
This field no longer needs to be directly accessible by the compiler. Addresses a sterni lint from cl/6466 Change-Id: I5e6791943d7f0ab3d9b7a30bb1654c4a6a435b1f Reviewed-on: https://cl.tvl.fyi/c/depot/+/6564 Reviewed-by: sterni <sternenseemann@systemli.org> Tested-by: BuildkiteCI
2022-09-13 r/4836 feat(tvix/eval): implement initial fancy formatting for errorsVincent Ambo1-10/+6
This very closely follows the way it's done for warnings, but errors have a lot more information available in some cases which we do not surface yet. Note also that due to requiring the `CodeMap`, this is not yet called from eval.rs as the way that is threaded through needs to be refactored, so only the method for reporting these errors as strings is implemented so far. Next steps for this will be to add a generic diagnostics module that reduces some of the boilerplate for this between warnings & errors, and which will also give us a good point in the future to switch to a fancier diagnostics crate. Change-Id: If6bb209f8e7a568d866e516a90335b9b2afbf66d Reviewed-on: https://cl.tvl.fyi/c/depot/+/6534 Reviewed-by: grfn <grfn@gws.fyi> Tested-by: BuildkiteCI
2022-09-13 r/4830 feat(tvix/eval): implement initial fancy display for warningsVincent Ambo1-1/+1
This implements an initial fancy display for warnings emitted by the tvix compiler, using the codemap_diagnostic crate. Each warning variant has an associated message, and optionally an associated annotation for the span displayed to the user. In theory we could get a lot more fancy with the display for specific variants if needed (e.g. re-parse the AST and actually add multiple semantic spans based on context), but this is already a good start. Example: tvix-repl> let toString = https://tvl.fyi; in let inherit toString; in ({}: 42) rec {} warning[W004]: declared variable 'toString' shadows a built-in global! --> [tvix-repl]:1:5 | 1 | let toString = https://tvl.fyi; in let inherit toString; in ({}: 42) rec {} | ^^^^^^^^ variable declared here warning[W001]: URL literal syntax is deprecated, use a quoted string instead --> [tvix-repl]:1:16 | 1 | let toString = https://tvl.fyi; in let inherit toString; in ({}: 42) rec {} | ^^^^^^^^^^^^^^^ warning[W002]: inherited variable already exists with the same value --> [tvix-repl]:1:40 | 1 | let toString = https://tvl.fyi; in let inherit toString; in ({}: 42) rec {} | ^^^^^^^^^^^^^^^^^ warning[W999]: feature not yet implemented in tvix: recursive attribute sets --> [tvix-repl]:1:70 | 1 | let toString = https://tvl.fyi; in let inherit toString; in ({}: 42) rec {} | ^^^^^^ warning[W999]: feature not yet implemented in tvix: closed formals --> [tvix-repl]:1:62 | 1 | let toString = https://tvl.fyi; in let inherit toString; in ({}: 42) rec {} | ^^ warning[W003]: variable 'toString' is declared, but never used: --> [tvix-repl]:1:5 | 1 | let toString = https://tvl.fyi; in let inherit toString; in ({}: 42) rec {} | ^^^^^^^^ variable declared here => 42 :: int These are coloured when output to a terminal. Change-Id: If315648a07e333895db4ae1d0915ee2013806585 Reviewed-on: https://cl.tvl.fyi/c/depot/+/6532 Autosubmit: tazjin <tazjin@tvl.su> Reviewed-by: grfn <grfn@gws.fyi> Tested-by: BuildkiteCI
2022-09-11 r/4824 chore(tvix/eval): address current clippy lintsVincent Ambo1-2/+1
Change-Id: I76326c20a525044e89d3cd1392a29faa3414ca04 Reviewed-on: https://cl.tvl.fyi/c/depot/+/6529 Reviewed-by: sterni <sternenseemann@systemli.org> Tested-by: BuildkiteCI
2022-09-11 r/4823 refactor(tvix/eval): remove `todo!()` calls in compilerVincent Ambo2-4/+23
It is impossible for tvixbolt to recover from panics, so the user experience of typing an expression using an unsupported feature was that it would get sad and stop responding to input. Instead, raise a normal value-level error of a new variant and continue where possible. Change-Id: Ibe016c92cacb87b85095c0f83758eddc6468053e Reviewed-on: https://cl.tvl.fyi/c/depot/+/6528 Autosubmit: tazjin <tazjin@tvl.su> Tested-by: BuildkiteCI Reviewed-by: sterni <sternenseemann@systemli.org>
2022-09-11 r/4814 fix(tvix/eval): pass correct slot when compiling attr valuesVincent Ambo1-1/+1
Change-Id: I90722d59dea4c7694eb5a7cf505db31196ba6c6c Reviewed-on: https://cl.tvl.fyi/c/depot/+/6501 Tested-by: BuildkiteCI Reviewed-by: sterni <sternenseemann@systemli.org>
2022-09-11 r/4813 fix(tvix/eval): reduce scope depth in scope moduleVincent Ambo2-2/+2
Change-Id: If32f9e4c9212f20a39baa15d479ff1387c17570d Reviewed-on: https://cl.tvl.fyi/c/depot/+/6500 Tested-by: BuildkiteCI Reviewed-by: sterni <sternenseemann@systemli.org>
2022-09-11 r/4811 refactor(tvix/eval): refactor methods for parsing static identsVincent Ambo1-22/+39
Refactors the methods used for determining whether an identifier in a binding (i.e. an `rnix::Attr` node) is a static string, and extracting it. Previously all uses of this logic were for `let`-expressions, where dynamic attributes are always an error. However, we need the same logic to properly implement the phase separation of attribute set compilation. To facilitate this, the actual core logic of these methods now return `Option`, and are only converted to errors in cases where the errors are actually required. Change-Id: Iad7826eff2cb428182521c6f92276310edeae1eb Reviewed-on: https://cl.tvl.fyi/c/depot/+/6498 Tested-by: BuildkiteCI Reviewed-by: sterni <sternenseemann@systemli.org>
2022-09-11 r/4810 refactor(tvix/eval): move attrset-related code to compiler::attrsVincent Ambo2-202/+209
There's about to be a lot more code for attrsets (hopefully temporarily as part of an expand&contract cycle), while nested attribute logic is being refactored in preparation for recursive attribute sets. This does not change any functionality. Change-Id: I667565cd810ca7d9046120d1721c2ceb9095550b Reviewed-on: https://cl.tvl.fyi/c/depot/+/6497 Tested-by: BuildkiteCI Reviewed-by: sterni <sternenseemann@systemli.org>
2022-09-11 r/4809 fix(tvix/eval): place plain inherits in correct stack slotssterni1-27/+41
We need to make sure that we compile all plain inherits in a let expression before declaring any other locals. Plain inherits are special in the sense that they can never be recursive, instead resolving to a higher scope. Thus we need to compile their value, before declaring them. If we don't do that, before any other local can be declared, we cause a situation where the plain inherits' values are placed into other locals' stack slots. Note that we can't integrate the plain inherit compilation into the regular 2-3 phase model where we defer the compilation of the value or we'd compile `let inherit x; in …` as `let x = x; in …`. Change-Id: I951d5df3c9661a054e12401546875f4685b5bf08 Reviewed-on: https://cl.tvl.fyi/c/depot/+/6496 Tested-by: BuildkiteCI Reviewed-by: tazjin <tazjin@tvl.su> Reviewed-by: sterni <sternenseemann@systemli.org>
2022-09-11 r/4807 fix(tvix/eval): wrap asserts in a thunksterni1-1/+3
As the new test case demonstrates, asserts need to be evaluated lazily. Change-Id: I808046722c5a504e9497855ca5026d255c7a4c34 Reviewed-on: https://cl.tvl.fyi/c/depot/+/6494 Tested-by: BuildkiteCI Reviewed-by: tazjin <tazjin@tvl.su> Reviewed-by: sterni <sternenseemann@systemli.org>
2022-09-11 r/4805 fix(tvix/eval): declare let inherit (from) locals before compilingsterni1-47/+49
The recent change that split declaration of let based locals and the compilation of their values did not touch locals bound by inherit in let. These were previously declared and compiled immediately before starting to work on the other locals introduced in a let. In the case of plain inherits, this behavior is kept in this change, because there's nothing wrong with it: The value of a plain inherit will always resolve to a higher scope, either statically or dynamically. Since inherit (from) expression might refer to other locals bound in the same let, we need to handle them in the same three steps as ordinary let based locals: 1. We need to declare the (uninitialised) locals. 2. We need to compile the expression that obtains their value. For this, we create a new thunk, since the from expression may very well return a thunk which we need to force before selecting the value we are interested in. 3. Thunks need to be finalised. For 1., we create an extra pass over the inherits that already declares and initialises plain inherits and notes inherit (from) expressions in the entries vector after declaring them. 2. only needs a bit of adapting to create the thunks for selecting if appropriate, the rest of the existing code can be reused. Change-Id: Ie4ac1c0f9ffcbf7c07c452036aa8e577443af773 Reviewed-on: https://cl.tvl.fyi/c/depot/+/6490 Tested-by: BuildkiteCI Reviewed-by: sterni <sternenseemann@systemli.org> Reviewed-by: tazjin <tazjin@tvl.su>
2022-09-11 r/4802 fix(tvix/eval): thunk all uses of `with`Vincent Ambo1-1/+3
With this all other "weird scope" logic starts working for `with` as well. Change-Id: I0ea1d8c5fbd9cec5084bd574224f77b71ff2b487 Reviewed-on: https://cl.tvl.fyi/c/depot/+/6487 Tested-by: BuildkiteCI Reviewed-by: sterni <sternenseemann@systemli.org>
2022-09-11 r/4801 refactor(tvix/eval): capture entire with_stack in upvaluesVincent Ambo2-98/+66
This completely rewrites the handling of "dynamic upvalues" to, instead of resolving them at thunk/closure instantiation time (which forces some values too early), capture the entire with stack of parent contexts if it exists. There are a couple of things in here that could be written more efficiently, but I'm first working through this to get to a bug related to with + recursion and the code complexity of some of the optimisations is distracting. Change-Id: Ia538e06c9146e3bf8decb9adf02dd726d2c651cf Reviewed-on: https://cl.tvl.fyi/c/depot/+/6486 Tested-by: BuildkiteCI Reviewed-by: sterni <sternenseemann@systemli.org>
2022-09-11 r/4798 fix(tvix/eval): correctly account for slots during list constructionVincent Ambo1-1/+18
Similarly to attribute sets, list elements can be arbitrary expressions and their (temporary) stack slots during construction must be accounted for by the compiler. Change-Id: I3b6f7927860627fd867c64d0cab9104fd636d4f5 Reviewed-on: https://cl.tvl.fyi/c/depot/+/6470 Tested-by: BuildkiteCI Reviewed-by: sterni <sternenseemann@systemli.org>
2022-09-11 r/4797 refactor(tvix/eval): cut down one iteration over locals arrayVincent Ambo1-1/+2
Caught by sterni in cl/6339 Change-Id: I2f2cd746114f14854cf599a7223a42a3e8ebe4fc Reviewed-on: https://cl.tvl.fyi/c/depot/+/6469 Tested-by: BuildkiteCI Reviewed-by: sterni <sternenseemann@systemli.org>
2022-09-11 r/4796 fix(tvix/eval): account for attrset temporaries during constructionVincent Ambo1-2/+47
The temporaries left on the stack as operands to `OpAttrs` must be accounted for in the locals array in order for operations within them to receive correct slots. Some test cases that were previously broken have been added. Change-Id: Ib52b629bbdf7931f63fd45a45af1073022da923c Reviewed-on: https://cl.tvl.fyi/c/depot/+/6468 Tested-by: BuildkiteCI Reviewed-by: sterni <sternenseemann@systemli.org>
2022-09-11 r/4795 refactor(tvix/eval): add `initialised` arg to declare_phantomVincent Ambo2-9/+8
There are more upcomming uses of declare_phantom where this will come in handy to avoid some code bloat. Change-Id: I75cad8caf14511c519ab2f56e87e99bcbf0a082e Reviewed-on: https://cl.tvl.fyi/c/depot/+/6467 Tested-by: BuildkiteCI Reviewed-by: sterni <sternenseemann@systemli.org>
2022-09-11 r/4794 refactor(tvix/eval): encapsulate scope cleanup logic in moduleVincent Ambo2-37/+55
Moves the logic for removing tracked locals from a given scope from the compiler's locals list, and leaves only the actual compiler-related stuff (emitting warnings, cleaning up locals at runtime) in the compiler itself. Change-Id: I9da6eb54967f0a7775f624d602fe11be4c7ed5c4 Reviewed-on: https://cl.tvl.fyi/c/depot/+/6466 Tested-by: BuildkiteCI Reviewed-by: sterni <sternenseemann@systemli.org>
2022-09-11 r/4793 fix(tvix/eval): avoid forcing with-target until absolutely necessaryVincent Ambo1-1/+0
Change-Id: I00efbbb8b9d3d22f32becf0919c6adf1be8b4b69 Reviewed-on: https://cl.tvl.fyi/c/depot/+/6465 Tested-by: BuildkiteCI Reviewed-by: sterni <sternenseemann@systemli.org>
2022-09-11 r/4792 refactor(tvix/eval): extract attribute set inherit into helperVincent Ambo1-18/+31
This will be re-used between the code paths for recursive/non-recursive sets, and it might even be possible to unify it with the logic for compiling `let inherit ...`. Change-Id: I960a061048ac583a6e932e11ff6e642d9fc3093e Reviewed-on: https://cl.tvl.fyi/c/depot/+/6464 Tested-by: BuildkiteCI Reviewed-by: sterni <sternenseemann@systemli.org>
2022-09-11 r/4790 feat(tvix/eval): implement "formals" function parametersVincent Ambo1-1/+97
The comment explains how this works fairly well. Note that this does not yet have the ability to check "closed formals", i.e. without an ellipsis Tvix will *NOT* fail if unexpected attribute set keys are provided. Change-Id: I0d2b77e893243093d2789baa57f876d35d0a32ff Reviewed-on: https://cl.tvl.fyi/c/depot/+/6463 Reviewed-by: sterni <sternenseemann@systemli.org> Tested-by: BuildkiteCI
2022-09-10 r/4786 fix(tvix/eval): fix doc comment syntax where applicableVincent Ambo2-53/+56
As pointed out by grfn on cl/6091 Change-Id: I28308577b7cf99dffb4a4fd3cc8783eb9ab4d0d6 Reviewed-on: https://cl.tvl.fyi/c/depot/+/6460 Tested-by: BuildkiteCI Reviewed-by: sterni <sternenseemann@systemli.org>
2022-09-10 r/4784 feat(tvix/eval): optimise tail calls in emitted chunksVincent Ambo1-2/+20
When the last instruction in a chunk is OpCall, make it an OpTailCall instead. Change-Id: I2c80a06ee85e4abf545887b1a79b6d8b5e6123e9 Reviewed-on: https://cl.tvl.fyi/c/depot/+/6458 Tested-by: BuildkiteCI Reviewed-by: sterni <sternenseemann@systemli.org>
2022-09-09 r/4781 chore(tvix/eval): clean up a stale commentVincent Ambo1-4/+0
Change-Id: If1b02fe1c78398387ea98490e5b099f1ff1b4164 Reviewed-on: https://cl.tvl.fyi/c/depot/+/6455 Reviewed-by: sterni <sternenseemann@systemli.org> Tested-by: BuildkiteCI
2022-09-09 r/4775 feat(tvix/eval): implement DisassemblingObserver for compilerVincent Ambo1-41/+18
This type implements an observer that is called whenever the compiler emits a chunk (after the toplevel, thunks, or lambdas) and prints the output of the disassembler to its internal writer. This replaces half of the uses of the `disassembler` feature, which has been removed from the Cargo configuration. Note that at this commit runtime tracing is not yet implemented as an observer. Change-Id: I7894ca1ba445761aba4ad51d98e4a7b6445f1aea Reviewed-on: https://cl.tvl.fyi/c/depot/+/6449 Reviewed-by: sterni <sternenseemann@systemli.org> Tested-by: BuildkiteCI
2022-09-08 r/4770 feat(tvix/eval): thunk binary operations and select expressionsVincent Ambo1-2/+14
With this, most cases of `fix` in attribute sets will work correctly. A simple test exercising both has been added. Change-Id: I70fd431177bb6e48ecb33a87518b050c4c3d1c09 Reviewed-on: https://cl.tvl.fyi/c/depot/+/6437 Reviewed-by: sterni <sternenseemann@systemli.org> Tested-by: BuildkiteCI
2022-09-08 r/4768 feat(tvix/eval): print lambda memory adresses in disassemblerVincent Ambo1-15/+10
This makes it easier to track exactly which lambda is which when inspecting e.g. the concrete representation of a thunk. At runtime all lambdas live in an Rc. To make this print the right address, the construction of these Rcs had to be moved up right to the point where the lambda is first emitted (and disassembled). Change-Id: I6070e6c8ac55f0bd697966c4e7c5565c20d19106 Reviewed-on: https://cl.tvl.fyi/c/depot/+/6435 Reviewed-by: sterni <sternenseemann@systemli.org> Tested-by: BuildkiteCI
2022-09-08 r/4766 feat(tvix/eval): thunk function applicationsVincent Ambo1-1/+4
Change-Id: I18065ed234ec104ac74d0e1c2d0937c2d78ca7db Reviewed-on: https://cl.tvl.fyi/c/depot/+/6433 Reviewed-by: sterni <sternenseemann@systemli.org> Tested-by: BuildkiteCI
2022-09-08 r/4765 feat(tvix/eval): thunk creation of listsVincent Ambo1-1/+3
Change-Id: I84b68c5d002ec613d278315bbf49e9839f0fe8e8 Reviewed-on: https://cl.tvl.fyi/c/depot/+/6432 Reviewed-by: sterni <sternenseemann@systemli.org> Tested-by: BuildkiteCI
2022-09-08 r/4763 fix(tvix/eval): address current clippy & grfn lintsVincent Ambo2-7/+10
Change-Id: I65c6feb9f817b5b367d37204a1f57acfe4100d97 Reviewed-on: https://cl.tvl.fyi/c/depot/+/6430 Reviewed-by: sterni <sternenseemann@systemli.org> Tested-by: BuildkiteCI
2022-09-08 r/4762 fix(tvix/eval): consider local depth when deciding to deferVincent Ambo1-5/+8
Deferred local upvalues can *only* occur at the same depth as the thing that is closing over them, but there are various situations with scope nesting where the actual stack indexes of the local and the closer look like a deferred value is being accessed. To fix this, simply compare the depth as well. Change-Id: Ice77424cc87ab0a2c4f01379e68d4399a917b12b Reviewed-on: https://cl.tvl.fyi/c/depot/+/6429 Reviewed-by: sterni <sternenseemann@systemli.org> Tested-by: BuildkiteCI
2022-09-08 r/4760 refactor(tvix/eval): clean up logic in Compiler::end_scopeVincent Ambo1-5/+3
The condition here was extremely hard to read prior to this change. As the locals vector is now guaranteed to never be empty (there is always at least a phantom for the current chunk's root expression), the logic here can be simplified to just dropping tailing locals entries while their depth matches that of the scope being closed. Change-Id: I24973e23bc2ad25e62ece64ab4d8624e6e274c16 Reviewed-on: https://cl.tvl.fyi/c/depot/+/6427 Reviewed-by: sterni <sternenseemann@systemli.org> Tested-by: BuildkiteCI
2022-09-08 r/4759 fix(tvix/eval): set up root stack slot in closures & thunksVincent Ambo1-4/+8
Similar to setting up a phantom slot when compiling the root value of a file, closures and thunks need to have a phantom stack slot for the root of the expression yielded by their thunk to make all accounting work correctly. The tricky thing here is that closures & thunks *escape* their inner lambda context (that's the point!), so the functions emitting them need to know both the *inner* slot (to resolve everything correctly while compiling the slot) and the *outer* slot (to correctly emit instructions for closing over upvalues). Change-Id: I62ac58e2f639c4b9e09cc702bdbfd2373e985d7f Reviewed-on: https://cl.tvl.fyi/c/depot/+/6426 Reviewed-by: sterni <sternenseemann@systemli.org> Tested-by: BuildkiteCI
2022-09-08 r/4758 fix(tvix/eval): only pop initialised locals when closing scopesVincent Ambo1-5/+10
This avoids emitting OpPop instructions for locals that only existed virtually (as uninitialised phantoms). Change-Id: I8105afcca80c3f7b7ef93ce5e2f0d08a93f4ad27 Reviewed-on: https://cl.tvl.fyi/c/depot/+/6425 Reviewed-by: sterni <sternenseemann@systemli.org> Tested-by: BuildkiteCI
2022-09-08 r/4757 fix(tvix/eval): ensure that root stack slot actually existsVincent Ambo2-5/+3
Instead of using a sentinel LocalIdx which potentially points to a value in the locals stack that does not actually exist, set up an initial uninitialised phantom value representing the result of the root expression. Change-Id: I82ea774daab83168020a3850bed57d35ab25c7df Reviewed-on: https://cl.tvl.fyi/c/depot/+/6424 Reviewed-by: sterni <sternenseemann@systemli.org> Tested-by: BuildkiteCI
2022-09-08 r/4756 fix(tvix/eval): compare *stack* slots when deciding whether to deferVincent Ambo1-1/+2
When deciding whether an upvalue needs to have a deferred resolution step, the *stack* indexes should be compared - not the locals indexes. The results are almost always the same, but there are tricky situations where this can cause errors. It's difficult to reproduce these errors in isolation, as they depend on other scope behaviour, so this is one in a series of commits to address the combination of issues which will gain some tests at the end. Change-Id: Iaa400b8d9500af58f493ab10e4f95022f3b5dd21 Reviewed-on: https://cl.tvl.fyi/c/depot/+/6423 Reviewed-by: sterni <sternenseemann@systemli.org> Tested-by: BuildkiteCI
2022-09-08 r/4755 fix(tvix/eval): propagate scope depth when nesting scopesVincent Ambo1-0/+1
Change-Id: Id441646db550f6195c2e247a0afbb5c9d91da8a0 Reviewed-on: https://cl.tvl.fyi/c/depot/+/6422 Reviewed-by: sterni <sternenseemann@systemli.org> Tested-by: BuildkiteCI
2022-09-08 r/4754 refactor(tvix/eval): refactor locals to use an enum for phantomsVincent Ambo2-22/+48
Instead of using sentinel values and an additional bool, this tracks the identifier of a local as an enum that is either a statically known name, or a phantom. To make this work correctly some more locals related logic has been encapsulated in the `scope` module, which is a good thing (that's the goal). Phantom values are now not initialised by default, but the only current call site of phantoms (`with` expression compilation) performs the initialisation right away. This commit changes no actual functionality right now, but paves the way for fixing an issue related to `let` bodies. Change-Id: I679f93a59a4daeacfe40f4012263cfb7bc05034e Reviewed-on: https://cl.tvl.fyi/c/depot/+/6421 Reviewed-by: sterni <sternenseemann@systemli.org> Tested-by: BuildkiteCI
generated by cgit-pink 1.4.1 (git 2.44.2) at 2024-12-18T04·13+0000