fix(rust): the Zig-shaped array [N]T discards its length (+4, -1 stated) - #3247
Merged
Merged
Conversation
compiler.rs:8322 maps "str" | "string" to []const u8 for Zig and
compiler.rs:18248 maps it to const char* for C. compiler.rs:24426 mapped it to
the owned String for Rust, and String is not constructible in a `const` item,
which is where the corpus mostly uses it:
pub const DEFAULT_NOTEBOOK: String = "t27-QUEEN-BRAIN";
error[E0308]: expected `String`, found `&str`
48 of the 75 specs whose first error was E0308 carried that mismatch.
Measured by name over all 650 corpus specs: rustc accepts 288 -> 314, +26, zero
regressions.
The judgement is stated rather than hidden. `&'static str` is narrower than
String -- a field of that type holds only a string living for the program -- and
that narrowing is what both siblings already chose, neither []const u8 nor
const char* owning its bytes. Zero regressions says nothing in the corpus needed
the owned form.
Also records the boundary my predictor from #3219 was missing. That rule ties a
fix's yield to the stub-bodied share of its class; this class is 3% stub-bodied
and I predicted +2..+8 against a measured +26. The rule holds for a fix repairing
a SIGNATURE while leaving a body, and not for one repairing a COMPLETE
DECLARATION: a const is correct or not on its own.
Seal and 92 drifted seals move in the same commit.
Closes #3239
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
…ositions (+1) A spec field may be named `type`, `ref` or `match`. Those are Rust keywords and the emitter wrote them bare, so the struct did not parse. `r#type` is the exact spelling and needs no judgement. The name is written in four places -- the field declaration, a field access, a function parameter and a struct literal -- and repairing the first two changed nothing beyond one spec. Repairing all four changed nothing further: the remaining specs of the class carry other defects behind the keyword. Measured by name over 650 specs: 314 -> 315, +1, zero regressions. The honest measure here is closure, not yield: `found keyword `type`` as a first error goes 7 -> 1, and the residue is a different cause (keywords in TYPE position, from #3225's list-valued declaration). crate/self/Self/super are deliberately not escaped: `r#` is invalid for them, so escaping would swap one parse error for another. Also records a measurement flaw that has misled me three times: taking the LAST numbered line of rustc's first error block reads back the SUGGESTION, since suggestions render as numbered lines and come last. It produced a phantom `pub struct ListNode<T>` earlier and made two specs here look already-repaired. Closes #3241 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The mapper has both array branches and the Zig-shaped one throws the size away: `[4]u8` becomes Vec<u8> while `[u8; 4]` becomes `[u8; 4]`. The right answer was already forty lines above in the same function, whose comment records that only the Zig spelling used to be handled. The two swapped places and one was left behind. Measured against a binary pinned at /tmp/t27c-pinned: 315 -> 318, +4, -1. THE TRADE IS STATED RATHER THAN BURIED. The regression is octree.t27, whose field is `children : [8]?OctNode` -- a fixed array of the struct it lives in. As Vec<Option<OctNode>> it compiled because a Vec supplies indirection by accident while discarding the length; as [Option<OctNode>; 8] it is honestly infinitely sized. Neither is right: the correct Rust is [Option<Box<OctNode>>; 8], which needs the enclosing type name inside a static function with 18 call sites. A second regression was caught before shipping: reading everything in the brackets as a size produced `[T; * as usize]` from `[*]T`, the many-item pointer. Guarded -- contents that are not a digit string or a plain identifier keep the unsized lowering. Both regressions came from one omission: I enumerated the branches of the bracket syntax and not the contents of the brackets. Provenance: this cause was surfaced by a fan-out audit whose measurements I then invalidated by rebuilding the compiler underneath it. One agent reported "THE RULER MOVED UNDER ME" and was right; the workflow was stopped and its counts discarded. Everything above is re-derived by hand against the pinned copy. Closes #3246 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
gHashTag
enabled auto-merge (squash)
September 5, 2026 03:47
This was referenced Sep 5, 2026
fix(rust): lower Zig's undefined to Default::default() (+2) — NOT auto-merged, needs your call
#3223
Closed
This was referenced Sep 5, 2026
Closed
Contributor
|
📓 NotebookLM Notebook linked to this PR
This notebook contains session context, decisions, and artifacts for this work. |
Contributor
Contributor
|
📓 NotebookLM Notebook linked to this PR
This notebook contains session context, decisions, and artifacts for this work. |
Contributor
PR DashboardGenerated at: 2026-09-05 04:28:54 UTC
Summary
Seal Status
|
This was referenced Sep 5, 2026
…-length # Conflicts: # .trinity/seals/ApbBridge.json # .trinity/seals/BoardMinimalXC7A100T.json # .trinity/seals/EmitterXDC.json # .trinity/seals/File.json # .trinity/seals/Git.json # .trinity/seals/GitDiff.json # .trinity/seals/GitOperations.json # .trinity/seals/Hir.json # .trinity/seals/Lexing.json # .trinity/seals/Linking.json # .trinity/seals/MAC_Testbench.json # .trinity/seals/NotebookLM.json # .trinity/seals/PhiRatio.json # .trinity/seals/PhiSplitOptimality.json # .trinity/seals/PhiUniversalAttractor.json # .trinity/seals/PinsIR.json # .trinity/seals/Project.json # .trinity/seals/SacredVerification.json # .trinity/seals/Session.json # .trinity/seals/Stdlib.json # .trinity/seals/TernaryIsa.json # .trinity/seals/Top_Level_Testbench.json # .trinity/seals/VcdConformanceCompare.json # .trinity/seals/Zamolodchikov4DConjecture.json # .trinity/seals/base-ring-32.json # .trinity/seals/base_base-ring-32.json # .trinity/seals/boards_BoardMinimalXC7A100T.json # .trinity/seals/cloud-railway-deploy.json # .trinity/seals/cloud_cloud-railway-deploy.json # .trinity/seals/compiler_Lexing.json # .trinity/seals/compiler_Linking.json # .trinity/seals/compiler_Stdlib.json # .trinity/seals/file_File.json # .trinity/seals/fpga_ApbBridge.json # .trinity/seals/fpga_FpgaStdlib.json # .trinity/seals/fpga_Hir.json # .trinity/seals/fpga_Memory.json # .trinity/seals/fpga_Stdlib.json # .trinity/seals/fpga_TernaryIsa.json # .trinity/seals/fpga_VcdConformanceCompare.json # .trinity/seals/git_Git.json # .trinity/seals/git_GitDiff.json # .trinity/seals/git_GitOperations.json # .trinity/seals/github::comments.json # .trinity/seals/github::issues.json # .trinity/seals/github::prs.json # .trinity/seals/github_github::comments.json # .trinity/seals/github_github::issues.json # .trinity/seals/github_github::prs.json # .trinity/seals/math_PhiSplitOptimality.json # .trinity/seals/math_PhiUniversalAttractor.json # .trinity/seals/memory_NotebookLM.json # .trinity/seals/nn_GatedLinearAttention.json # .trinity/seals/numeric_PhiRatio.json # .trinity/seals/physics_SacredVerification.json # .trinity/seals/physics_Zamolodchikov4DConjecture.json # .trinity/seals/pins_EmitterXDC.json # .trinity/seals/pins_PinsIR.json # .trinity/seals/race_igla-race-formal.json # .trinity/seals/server_Project.json # .trinity/seals/server_Session.json # .trinity/seals/testbench_MAC_Testbench.json # .trinity/seals/testbench_Top_Level_Testbench.json # bootstrap/stage0/FROZEN_HASH
Contributor
PR DashboardGenerated at: 2026-09-05 06:32:00 UTC
Summary
Seal Status
|
Contributor
|
📓 NotebookLM Notebook linked to this PR
This notebook contains session context, decisions, and artifacts for this work. |
…-length # Conflicts: # .trinity/seals/Hir.json # .trinity/seals/bus-schema.json # .trinity/seals/bus_bus-schema.json # .trinity/seals/config-paths.json # .trinity/seals/config_config-paths.json # .trinity/seals/fpga_Hir.json # .trinity/seals/parser_trilexer.json # .trinity/seals/provider-transform.json # .trinity/seals/provider_provider-transform.json # .trinity/seals/server-mdns.json # .trinity/seals/server_server-mdns.json # .trinity/seals/trilexer.json # bootstrap/src/compiler.rs # bootstrap/stage0/FROZEN_HASH
Contributor
|
📓 NotebookLM Notebook linked to this PR
This notebook contains session context, decisions, and artifacts for this work. |
Contributor
PR DashboardGenerated at: 2026-09-05 07:17:18 UTC
Summary
Seal Status
|
Contributor
|
📓 NotebookLM Notebook linked to this PR
This notebook contains session context, decisions, and artifacts for this work. |
Contributor
PR DashboardGenerated at: 2026-09-05 07:29:12 UTC
Summary
Seal Status
|
This was referenced Sep 6, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The mapper has both array branches and the Zig-shaped one throws the size away:
The answer was already forty lines above in the same function, and its comment records
the history: "Previously only the Zig-style
[N]Tform was handled". The two swappedplaces and one was left behind.
Measured against a pinned binary
The trade is stated in the open rather than buried. One spec regresses and it is
specs/tri/trees/octree.t27, whose field ischildren : [8]?OctNode— a fixed arrayof the struct it lives in. As
Vec<Option<OctNode>>it compiled, because aVecsupplies indirection by accident while discarding the length; as
[Option<OctNode>; 8]it is honestly infinitely sized and rustc says so.Neither lowering is right. The correct Rust is
[Option<Box<OctNode>>; 8], which keepsboth the length and the indirection, and that is recorded here for whoever takes it —
it needs the enclosing type's name inside
t27_type_to_rust, which is a staticfunction with 18 call sites, so it is not a mapper tweak.
The siblings do not settle it either: the C output for that spec is already invalid
(
?OctNode* children;— the optional leaked), and Zig's acceptance is the shallowbuild-objreading that does not analyse unreferenced declarations.A regression this change caused, and the measurement caught
The first version read everything inside the brackets as a size and produced
[T; * as usize]from[*]T— the source language's many-item pointer, where the*is not a length. That cost a second regression until the guard was added: anythinginside the brackets that is not a digit string or a plain identifier keeps the previous
unsized lowering.
Both regressions came from the same omission — I enumerated the branches of the
bracket syntax and not the contents of the brackets.
On provenance
This cause was surfaced by a fan-out audit whose measurements I then invalidated by
rebuilding the compiler underneath it — one agent reported "THE RULER MOVED UNDER
ME", and it was right. The workflow was stopped and its counts discarded. This finding
was re-derived by hand against a binary pinned at
/tmp/t27c-pinned, and every numberabove comes from that pinned copy.
Closes #3246