bug: --target windows: stack-map rewriter refuses COFF alias symbol records (`.def`/`.scl`/`.endef`) after .llvm_stackmaps for $pshape method aliases
Maintainers usually reply within 1 day
Nobody has claimed this yet.
Assessment
- Difficulty
- 2/5
- Estimated time
- 1-3 hours
- Newbie friendliness
- 82/100
- Issue type
- Bug
- Clarity
- Clearly specified
- Activity status
- Active
- Tech stack
- rust, typescript
Research direction
Start in crates/perry-codegen/src/gc_map.rs and read is_zero_width_directive, which the issue identifies as the place to handle COFF symbol records. Check the existing directive handling and the stack-map rewriter tests, then reproduce with the provided TypeScript example and Windows compile command. Done means the listed COFF directives no longer trigger the parse error and stack-map rewriting still preserves the GC roots.
Written by the indexing model from the issue text.
Description
Summary
On --target windows (COFF), a class whose method's $pshape clone has the same body as the ordinary method fails to compile. The error is "unrecognised directive .def inside the stack-map block".
What happens: inprocess/equivalent_shape_bodies.rs replaces the $pshape clone with an external LLVM alias of the method. On COFF, LLVM prints the alias's symbol record after .llvm_stackmaps:
.globl <sym>
.def <sym>;
.scl 2;
.type 32;
.endef
<sym> = <method>
The compact stack-map rewriter in crates/perry-codegen/src/gc_map.rs accepts .type and sym = expr, but not .def / .scl / .endef, so it refuses the module. This is the COFF counterpart of the Mach-O .alt_entry alias case (#12736).
Reproduction
coff_pshape_alias.ts:
// A method whose `$pshape` capability body equals the ordinary body becomes
// an LLVM alias after optimization. On COFF its symbol record
// (`.def`/`.scl`/`.type`/`.endef`) prints after the stack map; the
// Windows-target compile regression assembles this module.
export class Img {
private template = false;
setTemplateImage(option: boolean): void {
this.template = option;
}
isTemplateImage(): boolean {
return this.template;
}
}
const image = new Img();
image.setTemplateImage(true);
console.log(image.isTemplateImage());
perry compile coff_pshape_alias.ts --target windows --no-link -o c.o
perry: this module emits an LLVM stack map that the compact-map rewriter could not parse, so its GC roots would be invisible to the collector (the runtime reads only the compact section). Refusing to emit a binary that would lose roots silently.
reason: line 5440: unrecognised directive `.def` inside the stack-map block (`.def perry_method_coff_pshape_alias_ts__Img__setTemplateImage$pshape;`). [...] Add it to `directive_width` (with its width) or to `is_zero_width_directive`.
No special settings are needed. It fails the same way with the default settings, with PERRY_LL_OPT_LEVEL=0, with PERRY_LLVM_INPROCESS=1, and with both.
Suggested fix
Treat .def, .scl, .endef (and .safeseh) as zero-width in is_zero_width_directive. They only build COFF symbol-table records and emit no section bytes, so they can be carried through verbatim, as .type already is.
Environment
- PerryTS/perry main 86988bec62 (pristine, current head). Linux host build (
cargo build --release -p perry --no-default-features --features dev-cliplus the runtime/stdlib static archives), cross-compiling with--target windows. - Linux x86_64, Ubuntu 24.04.
- Dominant language
- Rust
- Stars
- 4.9k
- Forks
- 166
- Avg merge
- 4h 54m
- Merged PRs (30d)
- 909
Getting set up
- Ships a Dockerfile or Docker Compose file
- Has a pull request template
- Read the contributing guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
More from PerryTS/perry
-
area:compiler effort:S impact:med type:bug
Difficulty 2/5 1-3 hours Newbie friendliness 78/100
Maintainers usually reply within 1 day
-
area:runtime-gc effort:S impact:low platform:linux type:bug
Difficulty 2/5 1-3 hours Newbie friendliness 73/100
PerryTS/perry#14789 · 1 comment ·
Maintainers usually reply within 1 day
-
Difficulty 2/5 1-3 hours Newbie friendliness 70/100
Maintainers usually reply within 1 day
-
area:compiler effort:S impact:med type:bug
Difficulty 4/5 3-5 days Newbie friendliness 56/100
Maintainers usually reply within 1 day
-
area:runtime-gc effort:L impact:med type:perf
Difficulty 4/5 3-5 days Newbie friendliness 45/100
Maintainers usually reply within 1 day
Similar issues
-
<input type="number"> misses text-entry leaf sizing in layoutPossibly taken @GoCoder7 claimed this today. Open
Difficulty 1/5 Under an hour Newbie friendliness 88/100
DioxusLabs/blitz#1675 ·
Maintainers usually reply within 1 day
-
bug
Difficulty 2/5 1-3 hours Newbie friendliness 65/100
metasequoiaime/msime#11666 ·
Maintainers usually reply within 1 day
-
Difficulty 2/5 1-3 hours Newbie friendliness 77/100
Maintainers usually reply within 1 day
-
bug drawings help wanted ui
Difficulty 2/5 1-3 hours Newbie friendliness 82/100
userFRM/egui-charts#3479 ·
Maintainers usually reply within 1 day
-
Difficulty 2/5 1-3 hours Newbie friendliness 72/100
s2-streamstore/cachey#225 ·
Maintainers usually reply within 1 day