[wasm] Model class-init helper as value-returning in the JIT by lewing · Pull Request #130813 · dotnet/runtime

lewing · 2026-07-15T19:25:17Z

Summary

On wasm, the class-init JIT helpers (CORINFO_HELP_INITCLASS / CORINFO_HELP_INITINSTCLASS) are dispatched through portable entry points to the managed InitClass / InitInstantiatedClass helpers (InitHelpers.cs). Those helpers were changed to return void* (a discarded dummy value) so the compiled call_indirect signature matches the interpreter's value-returning portable-entry-point convention — but the JIT caller side was never updated and still typed these helpers as TYP_VOID.

As a result, R2R-compiled wasm code emits a (...)->() call_indirect for the first class-init of a type, while the runtime dispatches the cold class-init through the i32-returning thunk. The wasm engine traps with function signature mismatch.

This change models the class-init helpers as value-returning (TYP_I_IMPL) on wasm at the five sites that create them:

fgGetStaticsCCtorHelper (the shared-cctor path used by non-generic methods)
impInitClass (runtime-lookup path)
the three fgInitThisClass generic-lookup cases (THISOBJ, CLASSPARAM, METHODPARAM)

The helper result is pushed and then discarded via the existing unused-value drop in wasm codegen. The change is guarded by #ifdef TARGET_WASM; other targets are unaffected.

⚠️ Dependency

This is the JIT caller-side half of the class-init-helper return-arity change. It depends on the corelib half — InitClass / InitInstantiatedClass returning void* in InitHelpers.cs — which is not yet on main (it is part of the in-progress wasm R2R / SPC bring-up work, e.g. AndyAyersMS:next-blocker-spc). This PR is not correct in isolation without that corelib change: the JIT would model the helper as i32-returning while the helper still returned void. The two halves must land together (or this should be folded into the SPC bring-up set).

Opening as draft for human review of the JIT modeling change. There is currently no wasm-R2R CI leg, so validation below is from a local sweep.

Validation (local, browser wasm R2R)

Built the fixed cross-JIT, re-crossgen'd, and ran under corerun.js with DOTNET_ReadyToRun=1:

Test	Before	After
`conv_ovf`, `127931`, `indexMinusOne` (interpreted corelib)	trap (exit 1)	exit 100
`conv_ovf` (R2R corelib)	trap (exit 1)	exit 100
`b491215`, `GitHub_13561` (type-load)	pass	still pass

This resolves the residual overflow bucket of #129622 and removes the class-init signature-mismatch trap in #130634 (that repro now progresses into managed execution and surfaces a separate, unrelated NullReferenceException).

Contributes to #129622 / #130634.

Note

This PR was created with assistance from GitHub Copilot (AI).

azure-pipelines · 2026-07-15T19:26:01Z

Azure Pipelines: Successfully started running 5 pipeline(s). 10 pipeline(s) were filtered out due to trigger conditions. There may be pipelines that require an authorized user to comment /azp run to run.

dotnet-policy-service · 2026-07-15T19:27:40Z

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

Copilot

Pull request overview

This PR updates CoreCLR JIT IR modeling on TARGET_WASM so the class-init helpers (CORINFO_HELP_INITCLASS / CORINFO_HELP_INITINSTCLASS) are treated as value-returning (TYP_I_IMPL) rather than TYP_VOID, to align wasm indirect-call signature expectations with the helper’s return arity.

Changes:

Model CORINFO_HELP_INITCLASS as TYP_I_IMPL on wasm in fgGetStaticsCCtorHelper.
Model CORINFO_HELP_INITCLASS / CORINFO_HELP_INITINSTCLASS as TYP_I_IMPL on wasm at helper creation sites in impInitClass and fgInitThisClass.
Add explanatory comments describing the wasm-specific return-arity rationale.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 2 comments.

File	Description
src/coreclr/jit/flowgraph.cpp	Changes wasm helper return type modeling for `CORINFO_HELP_INITCLASS` in `fgGetStaticsCCtorHelper`, with added rationale comments.
src/coreclr/jit/importer.cpp	Models the runtime-lookup `InitClass` helper call as value-returning on wasm in `impInitClass`.
src/coreclr/jit/morph.cpp	Models the three generic-lookup `fgInitThisClass` helper call cases as value-returning on wasm.

…the JIT Completes gate #2 (91d9e36). That change made corelib InitClass/ InitInstantiatedClass (InitHelpers.cs) return void* so the compiled call_indirect signature matches the interpreter's value-returning portable entry point convention, but the JIT caller side still typed CORINFO_HELP_INITCLASS/INITINSTCLASS as TYP_VOID. On wasm the R2R caller therefore emitted a (...)->() call_indirect while the runtime dispatches the cold class-init through the i32-returning thunk, trapping with "function signature mismatch" on the first R2R class-init. Model the class-init helpers as value-returning (TYP_I_IMPL) on wasm at the five sites that create them (fgGetStaticsCCtorHelper, impInitClass runtime lookup, and the three fgInitThisClass generic-lookup cases). The helper result is pushed and discarded via the existing unused-value drop in wasm codegen. Fixes the residual overflow bucket of dotnet#129622 (conv_ovf, 127931, indexMinusOne now pass under R2R in both interpreted- and R2R-corelib configs) and removes the class-init signature-mismatch trap in dotnet#130634. No regression in the previously passing type-load tests. Prototype-only; for the future WASI R2R PR. Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com> Copilot-Session: 2902374a-f352-4c46-bcec-8a35c97733ae

lewing · 2026-07-15T19:35:22Z

Thanks — both points addressed in the pushed update (also refactored the per-site #ifdefs into a single Compiler::HelperInitClassRetType constant):

Comment referenced a nonexistent genCodeForCall. Fixed — the comment now points at the actual unused-value drop in CodeGen::WasmProduceReg (codegenwasm.cpp).
Depends on the CoreLib InitHelpers signature. Correct, and intentional — this is the JIT caller-side half of the class-init helper return-arity change; it only works together with InitClass/InitInstantiatedClass returning void* (that CoreLib change is part of the in-progress wasm R2R / SPC bring-up and is not on main yet). This is called out in the PR description's ⚠️ Dependency section, and I've now reinforced it in the HelperInitClassRetType comment. That coupling is exactly why this is a draft — it must land with (or fold into) the CoreLib change, not on its own.

Locally validated with both halves present: the residual overflow bucket of #129622 (conv_ovf, 127931, indexMinusOne) passes under R2R in both interpreted- and R2R-corelib configs, with no regression in the type-load tests.

Note

This comment was generated by GitHub Copilot (AI).

Copilot

Pull request overview

Copilot reviewed 4 out of 4 changed files in this pull request and generated 1 comment.

AndyAyersMS · 2026-07-15T20:03:26Z

My speculative "get SPC working" change set fixed this on the helper side, by retyping InitClass and InitInitialized class to return void *.

src/coreclr/System.Private.CoreLib/src/System/Runtime/CompilerServices/InitHelpers.cs

Doing that seems a bit simpler to me?

lewing · 2026-07-15T20:16:00Z

My speculative "get SPC working" change set fixed this on the helper side, by retyping InitClass and InitInitialized class to return void *.

src/coreclr/System.Private.CoreLib/src/System/Runtime/CompilerServices/InitHelpers.cs

Doing that seems a bit simpler to me?

This pr is meant to stack on top of your branch. The diagnosis is that your callee side change requires this caller side change to avoid the call_indirect type mismatch. Will double check.

lewing · 2026-07-15T20:18:51Z

Thanks Andy — your void* retype is the callee half and this PR is the caller half; they're the two matching sides of the same signature, not alternatives.

I ran a clean A/B on conv_ovf with your void* InitClass/InitInstantiatedClass compiled into the corelib in both runs, toggling only the JIT change:

	caller `call_indirect`	result
gate-#2 `void*` only	`(i32,i32,i32)->()`	❌ `function signature mismatch`
gate-#2 `void*` + this PR	`(i32,i32,i32,i32)->i32`	✅ pass

The reason the helper-side change alone isn't enough: the JIT models helper return types explicitly — fgGetStaticsCCtorHelper hardcodes CORINFO_HELP_INITCLASS → TYP_VOID, and helper-call return types aren't derived from the managed method — so the compiled call_indirect keeps declaring ->void until the JIT side is updated to match the now-void* helper.

Note

This comment was generated by GitHub Copilot (AI).

lewing · 2026-07-15T20:22:51Z

With this on top of your speculative branch I get past the call_indirect problems and land on some exception handling issues we're looking into.

lewing · 2026-07-15T22:15:48Z

With this on top of your speculative branch I get past the call_indirect problems and land on some exception handling issues we're looking into.

the eh issue is looking like a non-composite cross module r2r eh bug or an incomplete reloc pass

Per review feedback, condense the HelperInitClassRetType comment to the essential point (on wasm these helpers return void*, so model them as value-returning to match the callee) plus the CoreLib-dependency note. Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com> Copilot-Session: 2902374a-f352-4c46-bcec-8a35c97733ae

Copilot

Pull request overview

Copilot reviewed 4 out of 4 changed files in this pull request and generated 1 comment.

lewing · 2026-07-16T01:04:03Z

With this on top of your speculative branch I get past the call_indirect problems and land on some exception handling issues we're looking into.

the eh issue is looking like a non-composite cross module r2r eh bug or an incomplete reloc pass

figured out the eh issue, r2r functional on the branch. issue/pr incoming

lewing · 2026-07-21T19:50:49Z

depends on #131119 landing first

davidwrighton · 2026-07-21T23:38:37Z

@lewing #131119 has merged. Could you make this ready for merge?

azure-pipelines · 2026-07-22T00:08:11Z

Azure Pipelines: Successfully started running 5 pipeline(s). 11 pipeline(s) were filtered out due to trigger conditions. There may be pipelines that require an authorized user to comment /azp run to run.

lewing · 2026-07-22T00:09:59Z

@lewing #131119 has merged. Could you make this ready for merge?

done

Copilot

Pull request overview

Copilot reviewed 4 out of 4 changed files in this pull request and generated 1 comment.

Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>

Copilot

Pull request overview

Copilot reviewed 4 out of 4 changed files in this pull request and generated 1 comment.

Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>

Copilot

Pull request overview

Copilot reviewed 4 out of 4 changed files in this pull request and generated no new comments.

Copilot AI review requested due to automatic review settings July 15, 2026 19:25

github-actions Bot added the area-CodeGen-coreclr CLR JIT compiler in src/coreclr/src/jit and related components such as SuperPMI label Jul 15, 2026

lewing requested a review from AndyAyersMS July 15, 2026 19:25

dotnet-policy-service Bot assigned lewing Jul 15, 2026

Copilot started reviewing on behalf of lewing July 15, 2026 19:25 View session

lewing requested a review from pavelsavara July 15, 2026 19:26

pavelsavara added the arch-wasm WebAssembly architecture label Jul 15, 2026

Copilot AI reviewed Jul 15, 2026

View reviewed changes

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

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

Copilot AI review requested due to automatic review settings July 15, 2026 19:34

lewing force-pushed the lewing-wasm-r2r-classinit branch from 6acd714 to 80ed48b Compare July 15, 2026 19:34

Copilot started reviewing on behalf of lewing July 15, 2026 19:35 View session

Copilot AI reviewed Jul 15, 2026

View reviewed changes

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

lewing mentioned this pull request Jul 15, 2026

[wasm][R2R] function signature mismatch on the first (cold) R2R call to a method whose native code is not yet on its portable entry point #130634

Open

AndyAyersMS reviewed Jul 15, 2026

View reviewed changes

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

Copilot AI review requested due to automatic review settings July 16, 2026 00:51

Copilot started reviewing on behalf of lewing July 16, 2026 00:52 View session

Copilot AI reviewed Jul 16, 2026

View reviewed changes

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

This was referenced Jul 16, 2026

MSBuild crashing in the build #92290

Open

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

Open

lewing mentioned this pull request Jul 16, 2026

[wasi] Stand up CoreCLR-WASI library-test leg (per-app corerun relink) #130745

Merged

lewing requested a review from davidwrighton July 21, 2026 00:43

lewing mentioned this pull request Jul 21, 2026

Change InitClass and InitInstantiatedClass to return void* for WASM portable entry point compatibility #131119

Merged

lewing marked this pull request as ready for review July 22, 2026 00:07

Merge branch 'main' into lewing-wasm-r2r-classinit

01cd97e

Copilot AI review requested due to automatic review settings July 22, 2026 00:07

Copilot started reviewing on behalf of lewing July 22, 2026 00:08 View session

lewing enabled auto-merge (squash) July 22, 2026 00:09

Copilot AI reviewed Jul 22, 2026

View reviewed changes

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

Potential fix for pull request finding

637fb5e

Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>

Copilot AI review requested due to automatic review settings July 22, 2026 00:21

Copilot started reviewing on behalf of lewing July 22, 2026 00:22 View session

Copilot AI reviewed Jul 22, 2026

View reviewed changes

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

Potential fix for pull request finding

cc8eacf

Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>

Copilot AI review requested due to automatic review settings July 22, 2026 01:15

Copilot started reviewing on behalf of lewing July 22, 2026 01:16 View session

Copilot AI reviewed Jul 22, 2026

View reviewed changes

Uh oh!

Conversation

lewing commented Jul 15, 2026

Summary

⚠️ Dependency

Validation (local, browser wasm R2R)

Uh oh!

azure-pipelines Bot commented Jul 15, 2026

Uh oh!

dotnet-policy-service Bot commented Jul 15, 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!

lewing commented Jul 15, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

AndyAyersMS commented Jul 15, 2026

Uh oh!

lewing commented Jul 15, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

lewing commented Jul 15, 2026

Uh oh!

lewing commented Jul 15, 2026

Uh oh!

lewing commented Jul 15, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

lewing commented Jul 16, 2026

Uh oh!

lewing commented Jul 21, 2026

Uh oh!

davidwrighton commented Jul 21, 2026

Uh oh!

azure-pipelines Bot commented Jul 22, 2026

Uh oh!

lewing commented Jul 22, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

lewing commented Jul 15, 2026 •

edited

Loading

lewing commented Jul 15, 2026 •

edited

Loading