Fix wasm R2R JIT validation issues by AndyAyersMS · Pull Request #129555 · dotnet/runtime

AndyAyersMS · 2026-06-18T01:42:37Z

Use call node type when forming wasm call signatures.
Treat GT_FRAME_SIZE as a wasm leaf during use-edge iteration.
Classify single-register struct call stores by wasm ABI type.
Preserve ABI segment widths for wasm parameter register remaps.
Use the return type for wasm fast tail calls (gtType is overwritten to TYP_VOID).
JIT/wasm: skip Try/ExnRefWrapper in findTargetDepth for plain br targets.
JIT/wasm: emit an explicit br for try-leave even when the target is the contiguous next block.
JIT/wasm: extend the exnref wrapper to the catch-resumption dispatcher.
JIT/wasm: sort intervals by Start before conflict resolution (cross-try-exit Block intervals are backdated to start at the try header, breaking the non-decreasing-start-order invariant).
VM/wasm: use the slow IsFilterFunclet path for wasm R2R (wasm doesn't use personality routines in unwind data). (fixes Wasm / R2R: IsFilterFunclet assertion fires when re-throwing from a finally after a cctor failure (throwinclassconstructor) #129341)
JIT/wasm: skip the emitter-driven LCL_ADDR GC-tracking assert (wasm uses stack-slot GC info, not fixed-register tracking).

- Use call node type when forming wasm call signatures. - Treat GT_FRAME_SIZE as a wasm leaf during use-edge iteration. - Classify single-register struct call stores by wasm ABI type. - Preserve ABI segment widths for wasm parameter register remaps. - Bail out on wasm method/call signatures over the engine limit. Issue: dotnet#129497 Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

AndyAyersMS · 2026-06-18T01:42:55Z

@adamperlin PTAL
fyi @dotnet/wasm-contrib

Copilot

Pull request overview

This PR adjusts WebAssembly (Wasm) JIT/R2R codegen plumbing to produce wasm-valid call/method signatures and to avoid validation/assert failures in several JIT phases when compiling for Wasm.

Changes:

Update Wasm call signature construction and add explicit bailouts when a call/function would exceed the Wasm engine parameter limit (1000).
Treat GT_FRAME_SIZE as a leaf for Wasm during use-edge iteration to avoid consumers assuming it has operands.
Refine Wasm ABI typing in lowering for single-register struct call stores and preserve ABI segment widths when inducing parameter-register locals.

Reviewed changes

Copilot reviewed 6 out of 6 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
src/coreclr/jit/lower.cpp	Adjusts Wasm ABI-based typing for single-reg struct call stores; preserves register-segment type widths for induced parameter locals.
src/coreclr/jit/lclvars.cpp	Adds a Wasm-specific compilation bailout when the function argument count exceeds the Wasm engine limit.
src/coreclr/jit/gentree.cpp	Treats `GT_FRAME_SIZE` as a leaf for Wasm in `GenTreeUseEdgeIterator`.
src/coreclr/jit/error.h	Introduces `WASM_IMPL_LIMITATION` macro (Wasm-only impl-limitation helper).
src/coreclr/jit/compiler.h	Defines `MaxWasmFunctionParameters` constant for Wasm.
src/coreclr/jit/codegenwasm.cpp	Adjusts wasm call signature construction and adds a call-argument-count bailout prior to requesting a type symbol.

For fast tail calls gtType is overwritten to TYP_VOID, so use gtReturnType when forming the wasm call signature. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

…us next block

Cross-try-exit Block intervals are backdated to start at the try header, breaking the non-decreasing-start-order invariant assumed by the resolve loop.

AndyAyersMS · 2026-06-19T02:14:52Z

Not entirely happy with the "backdate" fix for branches out of try interval. Seems like constraint solving ought to be able to handle it, if a block overlaps a try (and since trys are single entry this means the block begins inside the try and ends outside) it should just stretch to encompass the try and (now) the two wrapper intervals.

Some of the other fixes are consequences of codegen inserting code (namely the post-try unreached and exception-capturing store) without having IR; this may also be something to revisit.

Wasm R2R doesn't use personality routines in unwind data, so fall back to the slow EH-clause walk like X86 does.

dotnet-policy-service · 2026-06-19T13:52:10Z

Tagging subscribers to 'arch-wasm': @lewing, @pavelsavara
See info in area-owners.md if you want to be subscribed.

Copilot

Pull request overview

Copilot reviewed 9 out of 9 changed files in this pull request and generated 1 comment.

AndyAyersMS · 2026-06-19T18:06:04Z

Pri-0 R2R sweep impact

	Baseline	After these commits
Tests (merged-runner XML)	699	1299
Pass	561	1161 (+600)
Fail	5	1
Wrapper failures	28	20

A separate Pri-1 sweep on top of these commits shows 1738 / 1909 tests pass (91.0%). No new R2R-specific failure modes surfaced in the broader Pri-1 set; remaining wrapper failures are tracked separately in #129622 and #129337.

Wasm uses stack-slot GC info encoding rather than fixed-register tracking, so the emitter concern this assert guards (live updates for LCL_ADDR DEFs of tracked GC vars) does not apply.

Copilot

Pull request overview

Copilot reviewed 6 out of 6 changed files in this pull request and generated no new comments.

davidwrighton

VM changes look good. JIT looks good too, but I'm not a good reviewer of JIT changes. :)

adamperlin

JIT changes lgtm from what I understand!

Copilot AI review requested due to automatic review settings June 18, 2026 01:42

github-actions Bot added the area-CodeGen-coreclr CLR JIT compiler in src/coreclr/src/jit and related components such as SuperPMI label Jun 18, 2026

Copilot started reviewing on behalf of AndyAyersMS June 18, 2026 01:43 View session

dotnet-policy-service Bot assigned AndyAyersMS Jun 18, 2026

Copilot AI reviewed Jun 18, 2026

View reviewed changes

Comment thread src/coreclr/jit/codegenwasm.cpp Outdated

Comment thread src/coreclr/jit/error.h Outdated

Comment thread src/coreclr/jit/error.h Outdated

build-analysis Bot mentioned this pull request Jun 18, 2026

The Operation will be canceled. The next steps may not contain expected logs. dotnet/dnceng#3008

Open

3 tasks

SingleAccretion reviewed Jun 18, 2026

View reviewed changes

Comment thread src/coreclr/jit/compiler.h Outdated

AndyAyersMS and others added 5 commits June 18, 2026 07:45

Use return type for wasm fast tail calls

eb65324

For fast tail calls gtType is overwritten to TYP_VOID, so use gtReturnType when forming the wasm call signature. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

JIT/wasm: skip Try/ExnRefWrapper in findTargetDepth for plain br targets

0f1e78a

JIT/wasm: emit explicit br for try-leave even when target is contiguo…

1f6c4d4

…us next block

JIT/wasm: extend exnref wrapper to the catch-resumption dispatcher

449fcc0

JIT/wasm: sort intervals by Start before conflict resolution

617c6df

Cross-try-exit Block intervals are backdated to start at the try header, breaking the non-decreasing-start-order invariant assumed by the resolve loop.

VM/wasm: use slow IsFilterFunclet path for wasm R2R

277c663

Wasm R2R doesn't use personality routines in unwind data, so fall back to the slow EH-clause walk like X86 does.

Copilot AI review requested due to automatic review settings June 19, 2026 13:43

Copilot started reviewing on behalf of AndyAyersMS June 19, 2026 13:44 View session

pavelsavara added the arch-wasm WebAssembly architecture label Jun 19, 2026

Copilot AI reviewed Jun 19, 2026

View reviewed changes

Comment thread src/coreclr/jit/lower.cpp

This was referenced Jun 19, 2026

Wasm / R2R: IsFilterFunclet assertion fires when re-throwing from a finally after a cctor failure (throwinclassconstructor) #129341

Closed

[Wasm R2R]: crossgen seems to be missing some thunk signatures #129622

Open

Undo param impl limitation

ce364d6

JIT/wasm: skip emitter-driven LCL_ADDR GC-tracking assert

fd15a98

Wasm uses stack-slot GC info encoding rather than fixed-register tracking, so the emitter concern this assert guards (live updates for LCL_ADDR DEFs of tracked GC vars) does not apply.

Copilot AI review requested due to automatic review settings June 19, 2026 18:40

Copilot started reviewing on behalf of AndyAyersMS June 19, 2026 18:40 View session

Copilot AI reviewed Jun 19, 2026

View reviewed changes

build-analysis Bot mentioned this pull request Jun 19, 2026

slow macOS - "##[error]The job running on agent Azure Pipelines 9 ran longer than the maximum time of 60 minutes." dotnet/dnceng#1883

Open

3 tasks

davidwrighton approved these changes Jun 22, 2026

View reviewed changes

adamperlin approved these changes Jun 22, 2026

View reviewed changes

AndyAyersMS merged commit 72788f0 into dotnet:main Jun 22, 2026
136 of 138 checks passed

dotnet-maestro Bot mentioned this pull request Jun 23, 2026

[main] Source code updates from dotnet/runtime dotnet/dotnet#7320

Merged

dotnet-milestone-bot Bot added this to the 11.0-preview6 milestone Jun 23, 2026

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Fix wasm R2R JIT validation issues#129555

Fix wasm R2R JIT validation issues#129555
AndyAyersMS merged 9 commits into
dotnet:mainfrom
AndyAyersMS:fix-wasm-r2r-jit-validation

AndyAyersMS commented Jun 18, 2026 •

edited

Loading

Uh oh!

AndyAyersMS commented Jun 18, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

AndyAyersMS commented Jun 19, 2026

Uh oh!

dotnet-policy-service Bot commented Jun 19, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Uh oh!

AndyAyersMS commented Jun 19, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

davidwrighton left a comment

Uh oh!

adamperlin left a comment •

edited

Loading

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

6 participants

Uh oh!

Conversation

AndyAyersMS commented Jun 18, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

AndyAyersMS commented Jun 18, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

AndyAyersMS commented Jun 19, 2026

Uh oh!

dotnet-policy-service Bot commented Jun 19, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

AndyAyersMS commented Jun 19, 2026

Pri-0 R2R sweep impact

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

davidwrighton left a comment

Choose a reason for hiding this comment

Uh oh!

adamperlin left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

6 participants

AndyAyersMS commented Jun 18, 2026 •

edited

Loading

adamperlin left a comment •

edited

Loading