[Wasm RyuJit]: don't skip top-level catchrets when marking try resumption by AndyAyersMS · Pull Request #129205 · dotnet/runtime

AndyAyersMS · 2026-06-10T01:50:42Z

In FlowGraphTryRegions::Build, the BBJ_EHCATCHRET check sat below the early bailout on hasTryIndex, so the catchret of an outer catch was never marking its try region as requiring runtime resumption.

Also: ensure loops that partially contain trys have their wasm-interval lengths extended to encompass any throw helpers associated with the try.

Also: fix some random extra argument added to a the overflow helper in some cases. This is all platform and has been in the code forever. On native it is somewhat harmless to pass unexpected extra args (especially if register args). On Wasm it is not.

Reminder that for Wasm control flow we can always move the end of a Wasm loop later; the extent just delineates where you can branch back to the loop head, not the actual extent of the loop. Likewise we can always move the start of a block earlier.

In FlowGraphTryRegions::Build, the BBJ_EHCATCHRET check sat below the early bailout on hasTryIndex, so the catchret of an outer catch was never marking its try region as requiring runtime resumption. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

AndyAyersMS · 2026-06-10T01:52:47Z

@dotnet/jit-contrib PTAL
@davidwrighton @dotnet/wasm-contrib FYI

Fixes a dumb logic bug.

With this and #129197 I'm able to resume after catch in simple R2R methods. Haven't tested more widely yet.

Copilot

Pull request overview

Adjusts FlowGraphTryRegions::Build so BBJ_EHCATCHRET blocks always mark their associated “dispatching try” region as requiring runtime resumption, even when the catchret block itself has no TryIndex (e.g., outermost catch).

Changes:

Move the BBJ_EHCATCHRET handling before the early !block->hasTryIndex() bailout.
Preserve existing behavior for try-region membership/entry-edge discovery while ensuring top-level catchrets are no longer skipped.

AndyAyersMS · 2026-06-10T13:36:26Z

SPC is not crossgenning cleanly with this fix. Failure is clearly related.

throw helpers associated with the try. Remove an unneded arg on a folded overflow throw helper call. This is harmless on native targets but causes a sig mismatch issue for Wasm.

JulieLeeMSFT

LGTM. The fix extends loop intervals to the right to fully enclose try region's throw helper blocks (prevents loop/try straddle).

AndyAyersMS · 2026-06-11T01:34:58Z

/ba-g unrelated legs timing out

…tion (dotnet#129205) In FlowGraphTryRegions::Build, the BBJ_EHCATCHRET check sat below the early bailout on hasTryIndex, so the catchret of an outer catch was never marking its try region as requiring runtime resumption. Also: ensure loops that partially contain trys have their wasm-interval lengths extended to encompass any throw helpers associated with the try. Also: fix some random extra argument added to a the overflow helper in some cases. This is all platform and has been in the code forever. On native it is somewhat harmless to pass unexpected extra args (especially if register args). On Wasm it is not. Reminder that for Wasm control flow we can always move the end of a Wasm loop later; the extent just delineates where you can branch back to the loop head, not the actual extent of the loop. Likewise we can always move the start of a block earlier. --------- Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

…tion (#129205) In FlowGraphTryRegions::Build, the BBJ_EHCATCHRET check sat below the early bailout on hasTryIndex, so the catchret of an outer catch was never marking its try region as requiring runtime resumption. Also: ensure loops that partially contain trys have their wasm-interval lengths extended to encompass any throw helpers associated with the try. Also: fix some random extra argument added to a the overflow helper in some cases. This is all platform and has been in the code forever. On native it is somewhat harmless to pass unexpected extra args (especially if register args). On Wasm it is not. Reminder that for Wasm control flow we can always move the end of a Wasm loop later; the extent just delineates where you can branch back to the loop head, not the actual extent of the loop. Likewise we can always move the start of a block earlier. --------- Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

AndyAyersMS · 2026-06-10T01:50:42Z

In FlowGraphTryRegions::Build, the BBJ_EHCATCHRET check sat below the early bailout on hasTryIndex, so the catchret of an outer catch was never marking its try region as requiring runtime resumption.

Also: ensure loops that partially contain trys have their wasm-interval lengths extended to encompass any throw helpers associated with the try.

Also: fix some random extra argument added to a the overflow helper in some cases. This is all platform and has been in the code forever. On native it is somewhat harmless to pass unexpected extra args (especially if register args). On Wasm it is not.

Reminder that for Wasm control flow we can always move the end of a Wasm loop later; the extent just delineates where you can branch back to the loop head, not the actual extent of the loop. Likewise we can always move the start of a block earlier.

In FlowGraphTryRegions::Build, the BBJ_EHCATCHRET check sat below the early bailout on hasTryIndex, so the catchret of an outer catch was never marking its try region as requiring runtime resumption. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

AndyAyersMS · 2026-06-10T01:52:47Z

@dotnet/jit-contrib PTAL
@davidwrighton @dotnet/wasm-contrib FYI

Fixes a dumb logic bug.

With this and #129197 I'm able to resume after catch in simple R2R methods. Haven't tested more widely yet.

Copilot

Pull request overview

Adjusts FlowGraphTryRegions::Build so BBJ_EHCATCHRET blocks always mark their associated “dispatching try” region as requiring runtime resumption, even when the catchret block itself has no TryIndex (e.g., outermost catch).

Changes:

Move the BBJ_EHCATCHRET handling before the early !block->hasTryIndex() bailout.
Preserve existing behavior for try-region membership/entry-edge discovery while ensuring top-level catchrets are no longer skipped.

AndyAyersMS · 2026-06-10T13:36:26Z

SPC is not crossgenning cleanly with this fix. Failure is clearly related.

throw helpers associated with the try. Remove an unneded arg on a folded overflow throw helper call. This is harmless on native targets but causes a sig mismatch issue for Wasm.

JulieLeeMSFT

LGTM. The fix extends loop intervals to the right to fully enclose try region's throw helper blocks (prevents loop/try straddle).

AndyAyersMS · 2026-06-11T01:34:58Z

/ba-g unrelated legs timing out

…tion (dotnet#129205) In FlowGraphTryRegions::Build, the BBJ_EHCATCHRET check sat below the early bailout on hasTryIndex, so the catchret of an outer catch was never marking its try region as requiring runtime resumption. Also: ensure loops that partially contain trys have their wasm-interval lengths extended to encompass any throw helpers associated with the try. Also: fix some random extra argument added to a the overflow helper in some cases. This is all platform and has been in the code forever. On native it is somewhat harmless to pass unexpected extra args (especially if register args). On Wasm it is not. Reminder that for Wasm control flow we can always move the end of a Wasm loop later; the extent just delineates where you can branch back to the loop head, not the actual extent of the loop. Likewise we can always move the start of a block earlier. --------- Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

…tion (#129205) In FlowGraphTryRegions::Build, the BBJ_EHCATCHRET check sat below the early bailout on hasTryIndex, so the catchret of an outer catch was never marking its try region as requiring runtime resumption. Also: ensure loops that partially contain trys have their wasm-interval lengths extended to encompass any throw helpers associated with the try. Also: fix some random extra argument added to a the overflow helper in some cases. This is all platform and has been in the code forever. On native it is somewhat harmless to pass unexpected extra args (especially if register args). On Wasm it is not. Reminder that for Wasm control flow we can always move the end of a Wasm loop later; the extent just delineates where you can branch back to the loop head, not the actual extent of the loop. Likewise we can always move the start of a block earlier. --------- Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

AndyAyersMS · 2026-06-10T01:50:42Z

In FlowGraphTryRegions::Build, the BBJ_EHCATCHRET check sat below the early bailout on hasTryIndex, so the catchret of an outer catch was never marking its try region as requiring runtime resumption.

Also: ensure loops that partially contain trys have their wasm-interval lengths extended to encompass any throw helpers associated with the try.

Also: fix some random extra argument added to a the overflow helper in some cases. This is all platform and has been in the code forever. On native it is somewhat harmless to pass unexpected extra args (especially if register args). On Wasm it is not.

Reminder that for Wasm control flow we can always move the end of a Wasm loop later; the extent just delineates where you can branch back to the loop head, not the actual extent of the loop. Likewise we can always move the start of a block earlier.

In FlowGraphTryRegions::Build, the BBJ_EHCATCHRET check sat below the early bailout on hasTryIndex, so the catchret of an outer catch was never marking its try region as requiring runtime resumption. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

AndyAyersMS · 2026-06-10T01:52:47Z

@dotnet/jit-contrib PTAL
@davidwrighton @dotnet/wasm-contrib FYI

Fixes a dumb logic bug.

With this and #129197 I'm able to resume after catch in simple R2R methods. Haven't tested more widely yet.

Copilot

Pull request overview

Adjusts FlowGraphTryRegions::Build so BBJ_EHCATCHRET blocks always mark their associated “dispatching try” region as requiring runtime resumption, even when the catchret block itself has no TryIndex (e.g., outermost catch).

Changes:

Move the BBJ_EHCATCHRET handling before the early !block->hasTryIndex() bailout.
Preserve existing behavior for try-region membership/entry-edge discovery while ensuring top-level catchrets are no longer skipped.

AndyAyersMS · 2026-06-10T13:36:26Z

SPC is not crossgenning cleanly with this fix. Failure is clearly related.

throw helpers associated with the try. Remove an unneded arg on a folded overflow throw helper call. This is harmless on native targets but causes a sig mismatch issue for Wasm.

JulieLeeMSFT

LGTM. The fix extends loop intervals to the right to fully enclose try region's throw helper blocks (prevents loop/try straddle).

AndyAyersMS · 2026-06-11T01:34:58Z

/ba-g unrelated legs timing out

…tion (dotnet#129205) In FlowGraphTryRegions::Build, the BBJ_EHCATCHRET check sat below the early bailout on hasTryIndex, so the catchret of an outer catch was never marking its try region as requiring runtime resumption. Also: ensure loops that partially contain trys have their wasm-interval lengths extended to encompass any throw helpers associated with the try. Also: fix some random extra argument added to a the overflow helper in some cases. This is all platform and has been in the code forever. On native it is somewhat harmless to pass unexpected extra args (especially if register args). On Wasm it is not. Reminder that for Wasm control flow we can always move the end of a Wasm loop later; the extent just delineates where you can branch back to the loop head, not the actual extent of the loop. Likewise we can always move the start of a block earlier. --------- Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

…tion (#129205) In FlowGraphTryRegions::Build, the BBJ_EHCATCHRET check sat below the early bailout on hasTryIndex, so the catchret of an outer catch was never marking its try region as requiring runtime resumption. Also: ensure loops that partially contain trys have their wasm-interval lengths extended to encompass any throw helpers associated with the try. Also: fix some random extra argument added to a the overflow helper in some cases. This is all platform and has been in the code forever. On native it is somewhat harmless to pass unexpected extra args (especially if register args). On Wasm it is not. Reminder that for Wasm control flow we can always move the end of a Wasm loop later; the extent just delineates where you can branch back to the loop head, not the actual extent of the loop. Likewise we can always move the start of a block earlier. --------- Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

AndyAyersMS · 2026-06-10T01:50:42Z

In FlowGraphTryRegions::Build, the BBJ_EHCATCHRET check sat below the early bailout on hasTryIndex, so the catchret of an outer catch was never marking its try region as requiring runtime resumption.

Also: ensure loops that partially contain trys have their wasm-interval lengths extended to encompass any throw helpers associated with the try.

Also: fix some random extra argument added to a the overflow helper in some cases. This is all platform and has been in the code forever. On native it is somewhat harmless to pass unexpected extra args (especially if register args). On Wasm it is not.

Reminder that for Wasm control flow we can always move the end of a Wasm loop later; the extent just delineates where you can branch back to the loop head, not the actual extent of the loop. Likewise we can always move the start of a block earlier.

In FlowGraphTryRegions::Build, the BBJ_EHCATCHRET check sat below the early bailout on hasTryIndex, so the catchret of an outer catch was never marking its try region as requiring runtime resumption. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

AndyAyersMS · 2026-06-10T01:52:47Z

@dotnet/jit-contrib PTAL
@davidwrighton @dotnet/wasm-contrib FYI

Fixes a dumb logic bug.

With this and #129197 I'm able to resume after catch in simple R2R methods. Haven't tested more widely yet.

Copilot

Pull request overview

Adjusts FlowGraphTryRegions::Build so BBJ_EHCATCHRET blocks always mark their associated “dispatching try” region as requiring runtime resumption, even when the catchret block itself has no TryIndex (e.g., outermost catch).

Changes:

Move the BBJ_EHCATCHRET handling before the early !block->hasTryIndex() bailout.
Preserve existing behavior for try-region membership/entry-edge discovery while ensuring top-level catchrets are no longer skipped.

AndyAyersMS · 2026-06-10T13:36:26Z

SPC is not crossgenning cleanly with this fix. Failure is clearly related.

throw helpers associated with the try. Remove an unneded arg on a folded overflow throw helper call. This is harmless on native targets but causes a sig mismatch issue for Wasm.

JulieLeeMSFT

LGTM. The fix extends loop intervals to the right to fully enclose try region's throw helper blocks (prevents loop/try straddle).

AndyAyersMS · 2026-06-11T01:34:58Z

/ba-g unrelated legs timing out

…tion (dotnet#129205) In FlowGraphTryRegions::Build, the BBJ_EHCATCHRET check sat below the early bailout on hasTryIndex, so the catchret of an outer catch was never marking its try region as requiring runtime resumption. Also: ensure loops that partially contain trys have their wasm-interval lengths extended to encompass any throw helpers associated with the try. Also: fix some random extra argument added to a the overflow helper in some cases. This is all platform and has been in the code forever. On native it is somewhat harmless to pass unexpected extra args (especially if register args). On Wasm it is not. Reminder that for Wasm control flow we can always move the end of a Wasm loop later; the extent just delineates where you can branch back to the loop head, not the actual extent of the loop. Likewise we can always move the start of a block earlier. --------- Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

…tion (#129205) In FlowGraphTryRegions::Build, the BBJ_EHCATCHRET check sat below the early bailout on hasTryIndex, so the catchret of an outer catch was never marking its try region as requiring runtime resumption. Also: ensure loops that partially contain trys have their wasm-interval lengths extended to encompass any throw helpers associated with the try. Also: fix some random extra argument added to a the overflow helper in some cases. This is all platform and has been in the code forever. On native it is somewhat harmless to pass unexpected extra args (especially if register args). On Wasm it is not. Reminder that for Wasm control flow we can always move the end of a Wasm loop later; the extent just delineates where you can branch back to the loop head, not the actual extent of the loop. Likewise we can always move the start of a block earlier. --------- Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

AndyAyersMS · 2026-06-10T01:50:42Z

In FlowGraphTryRegions::Build, the BBJ_EHCATCHRET check sat below the early bailout on hasTryIndex, so the catchret of an outer catch was never marking its try region as requiring runtime resumption.

Also: ensure loops that partially contain trys have their wasm-interval lengths extended to encompass any throw helpers associated with the try.

Also: fix some random extra argument added to a the overflow helper in some cases. This is all platform and has been in the code forever. On native it is somewhat harmless to pass unexpected extra args (especially if register args). On Wasm it is not.

Reminder that for Wasm control flow we can always move the end of a Wasm loop later; the extent just delineates where you can branch back to the loop head, not the actual extent of the loop. Likewise we can always move the start of a block earlier.

In FlowGraphTryRegions::Build, the BBJ_EHCATCHRET check sat below the early bailout on hasTryIndex, so the catchret of an outer catch was never marking its try region as requiring runtime resumption. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

AndyAyersMS · 2026-06-10T01:52:47Z

@dotnet/jit-contrib PTAL
@davidwrighton @dotnet/wasm-contrib FYI

Fixes a dumb logic bug.

With this and #129197 I'm able to resume after catch in simple R2R methods. Haven't tested more widely yet.

Copilot

Pull request overview

Adjusts FlowGraphTryRegions::Build so BBJ_EHCATCHRET blocks always mark their associated “dispatching try” region as requiring runtime resumption, even when the catchret block itself has no TryIndex (e.g., outermost catch).

Changes:

Move the BBJ_EHCATCHRET handling before the early !block->hasTryIndex() bailout.
Preserve existing behavior for try-region membership/entry-edge discovery while ensuring top-level catchrets are no longer skipped.

AndyAyersMS · 2026-06-10T13:36:26Z

SPC is not crossgenning cleanly with this fix. Failure is clearly related.

throw helpers associated with the try. Remove an unneded arg on a folded overflow throw helper call. This is harmless on native targets but causes a sig mismatch issue for Wasm.

JulieLeeMSFT

LGTM. The fix extends loop intervals to the right to fully enclose try region's throw helper blocks (prevents loop/try straddle).

AndyAyersMS · 2026-06-11T01:34:58Z

/ba-g unrelated legs timing out

…tion (dotnet#129205) In FlowGraphTryRegions::Build, the BBJ_EHCATCHRET check sat below the early bailout on hasTryIndex, so the catchret of an outer catch was never marking its try region as requiring runtime resumption. Also: ensure loops that partially contain trys have their wasm-interval lengths extended to encompass any throw helpers associated with the try. Also: fix some random extra argument added to a the overflow helper in some cases. This is all platform and has been in the code forever. On native it is somewhat harmless to pass unexpected extra args (especially if register args). On Wasm it is not. Reminder that for Wasm control flow we can always move the end of a Wasm loop later; the extent just delineates where you can branch back to the loop head, not the actual extent of the loop. Likewise we can always move the start of a block earlier. --------- Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

…tion (#129205) In FlowGraphTryRegions::Build, the BBJ_EHCATCHRET check sat below the early bailout on hasTryIndex, so the catchret of an outer catch was never marking its try region as requiring runtime resumption. Also: ensure loops that partially contain trys have their wasm-interval lengths extended to encompass any throw helpers associated with the try. Also: fix some random extra argument added to a the overflow helper in some cases. This is all platform and has been in the code forever. On native it is somewhat harmless to pass unexpected extra args (especially if register args). On Wasm it is not. Reminder that for Wasm control flow we can always move the end of a Wasm loop later; the extent just delineates where you can branch back to the loop head, not the actual extent of the loop. Likewise we can always move the start of a block earlier. --------- Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

AndyAyersMS · 2026-06-10T01:50:42Z

In FlowGraphTryRegions::Build, the BBJ_EHCATCHRET check sat below the early bailout on hasTryIndex, so the catchret of an outer catch was never marking its try region as requiring runtime resumption.

Also: ensure loops that partially contain trys have their wasm-interval lengths extended to encompass any throw helpers associated with the try.

Also: fix some random extra argument added to a the overflow helper in some cases. This is all platform and has been in the code forever. On native it is somewhat harmless to pass unexpected extra args (especially if register args). On Wasm it is not.

Reminder that for Wasm control flow we can always move the end of a Wasm loop later; the extent just delineates where you can branch back to the loop head, not the actual extent of the loop. Likewise we can always move the start of a block earlier.

In FlowGraphTryRegions::Build, the BBJ_EHCATCHRET check sat below the early bailout on hasTryIndex, so the catchret of an outer catch was never marking its try region as requiring runtime resumption. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

AndyAyersMS · 2026-06-10T01:52:47Z

@dotnet/jit-contrib PTAL
@davidwrighton @dotnet/wasm-contrib FYI

Fixes a dumb logic bug.

With this and #129197 I'm able to resume after catch in simple R2R methods. Haven't tested more widely yet.

Copilot

Pull request overview

Adjusts FlowGraphTryRegions::Build so BBJ_EHCATCHRET blocks always mark their associated “dispatching try” region as requiring runtime resumption, even when the catchret block itself has no TryIndex (e.g., outermost catch).

Changes:

Move the BBJ_EHCATCHRET handling before the early !block->hasTryIndex() bailout.
Preserve existing behavior for try-region membership/entry-edge discovery while ensuring top-level catchrets are no longer skipped.

AndyAyersMS · 2026-06-10T13:36:26Z

SPC is not crossgenning cleanly with this fix. Failure is clearly related.

throw helpers associated with the try. Remove an unneded arg on a folded overflow throw helper call. This is harmless on native targets but causes a sig mismatch issue for Wasm.

JulieLeeMSFT

LGTM. The fix extends loop intervals to the right to fully enclose try region's throw helper blocks (prevents loop/try straddle).

AndyAyersMS · 2026-06-11T01:34:58Z

/ba-g unrelated legs timing out

…tion (dotnet#129205) In FlowGraphTryRegions::Build, the BBJ_EHCATCHRET check sat below the early bailout on hasTryIndex, so the catchret of an outer catch was never marking its try region as requiring runtime resumption. Also: ensure loops that partially contain trys have their wasm-interval lengths extended to encompass any throw helpers associated with the try. Also: fix some random extra argument added to a the overflow helper in some cases. This is all platform and has been in the code forever. On native it is somewhat harmless to pass unexpected extra args (especially if register args). On Wasm it is not. Reminder that for Wasm control flow we can always move the end of a Wasm loop later; the extent just delineates where you can branch back to the loop head, not the actual extent of the loop. Likewise we can always move the start of a block earlier. --------- Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

…tion (#129205) In FlowGraphTryRegions::Build, the BBJ_EHCATCHRET check sat below the early bailout on hasTryIndex, so the catchret of an outer catch was never marking its try region as requiring runtime resumption. Also: ensure loops that partially contain trys have their wasm-interval lengths extended to encompass any throw helpers associated with the try. Also: fix some random extra argument added to a the overflow helper in some cases. This is all platform and has been in the code forever. On native it is somewhat harmless to pass unexpected extra args (especially if register args). On Wasm it is not. Reminder that for Wasm control flow we can always move the end of a Wasm loop later; the extent just delineates where you can branch back to the loop head, not the actual extent of the loop. Likewise we can always move the start of a block earlier. --------- Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

AndyAyersMS · 2026-06-10T01:50:42Z

In FlowGraphTryRegions::Build, the BBJ_EHCATCHRET check sat below the early bailout on hasTryIndex, so the catchret of an outer catch was never marking its try region as requiring runtime resumption.

Also: ensure loops that partially contain trys have their wasm-interval lengths extended to encompass any throw helpers associated with the try.

Also: fix some random extra argument added to a the overflow helper in some cases. This is all platform and has been in the code forever. On native it is somewhat harmless to pass unexpected extra args (especially if register args). On Wasm it is not.

Reminder that for Wasm control flow we can always move the end of a Wasm loop later; the extent just delineates where you can branch back to the loop head, not the actual extent of the loop. Likewise we can always move the start of a block earlier.

In FlowGraphTryRegions::Build, the BBJ_EHCATCHRET check sat below the early bailout on hasTryIndex, so the catchret of an outer catch was never marking its try region as requiring runtime resumption. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

AndyAyersMS · 2026-06-10T01:52:47Z

@dotnet/jit-contrib PTAL
@davidwrighton @dotnet/wasm-contrib FYI

Fixes a dumb logic bug.

With this and #129197 I'm able to resume after catch in simple R2R methods. Haven't tested more widely yet.

Copilot

Pull request overview

Adjusts FlowGraphTryRegions::Build so BBJ_EHCATCHRET blocks always mark their associated “dispatching try” region as requiring runtime resumption, even when the catchret block itself has no TryIndex (e.g., outermost catch).

Changes:

Move the BBJ_EHCATCHRET handling before the early !block->hasTryIndex() bailout.
Preserve existing behavior for try-region membership/entry-edge discovery while ensuring top-level catchrets are no longer skipped.

AndyAyersMS · 2026-06-10T13:36:26Z

SPC is not crossgenning cleanly with this fix. Failure is clearly related.

throw helpers associated with the try. Remove an unneded arg on a folded overflow throw helper call. This is harmless on native targets but causes a sig mismatch issue for Wasm.

JulieLeeMSFT

LGTM. The fix extends loop intervals to the right to fully enclose try region's throw helper blocks (prevents loop/try straddle).

AndyAyersMS · 2026-06-11T01:34:58Z

/ba-g unrelated legs timing out

…tion (dotnet#129205) In FlowGraphTryRegions::Build, the BBJ_EHCATCHRET check sat below the early bailout on hasTryIndex, so the catchret of an outer catch was never marking its try region as requiring runtime resumption. Also: ensure loops that partially contain trys have their wasm-interval lengths extended to encompass any throw helpers associated with the try. Also: fix some random extra argument added to a the overflow helper in some cases. This is all platform and has been in the code forever. On native it is somewhat harmless to pass unexpected extra args (especially if register args). On Wasm it is not. Reminder that for Wasm control flow we can always move the end of a Wasm loop later; the extent just delineates where you can branch back to the loop head, not the actual extent of the loop. Likewise we can always move the start of a block earlier. --------- Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

…tion (#129205) In FlowGraphTryRegions::Build, the BBJ_EHCATCHRET check sat below the early bailout on hasTryIndex, so the catchret of an outer catch was never marking its try region as requiring runtime resumption. Also: ensure loops that partially contain trys have their wasm-interval lengths extended to encompass any throw helpers associated with the try. Also: fix some random extra argument added to a the overflow helper in some cases. This is all platform and has been in the code forever. On native it is somewhat harmless to pass unexpected extra args (especially if register args). On Wasm it is not. Reminder that for Wasm control flow we can always move the end of a Wasm loop later; the extent just delineates where you can branch back to the loop head, not the actual extent of the loop. Likewise we can always move the start of a block earlier. --------- Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

AndyAyersMS · 2026-06-10T01:50:42Z

In FlowGraphTryRegions::Build, the BBJ_EHCATCHRET check sat below the early bailout on hasTryIndex, so the catchret of an outer catch was never marking its try region as requiring runtime resumption.

Also: ensure loops that partially contain trys have their wasm-interval lengths extended to encompass any throw helpers associated with the try.

Also: fix some random extra argument added to a the overflow helper in some cases. This is all platform and has been in the code forever. On native it is somewhat harmless to pass unexpected extra args (especially if register args). On Wasm it is not.

Reminder that for Wasm control flow we can always move the end of a Wasm loop later; the extent just delineates where you can branch back to the loop head, not the actual extent of the loop. Likewise we can always move the start of a block earlier.

In FlowGraphTryRegions::Build, the BBJ_EHCATCHRET check sat below the early bailout on hasTryIndex, so the catchret of an outer catch was never marking its try region as requiring runtime resumption. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

AndyAyersMS · 2026-06-10T01:52:47Z

@dotnet/jit-contrib PTAL
@davidwrighton @dotnet/wasm-contrib FYI

Fixes a dumb logic bug.

With this and #129197 I'm able to resume after catch in simple R2R methods. Haven't tested more widely yet.

Copilot

Pull request overview

Adjusts FlowGraphTryRegions::Build so BBJ_EHCATCHRET blocks always mark their associated “dispatching try” region as requiring runtime resumption, even when the catchret block itself has no TryIndex (e.g., outermost catch).

Changes:

Move the BBJ_EHCATCHRET handling before the early !block->hasTryIndex() bailout.
Preserve existing behavior for try-region membership/entry-edge discovery while ensuring top-level catchrets are no longer skipped.

AndyAyersMS · 2026-06-10T13:36:26Z

SPC is not crossgenning cleanly with this fix. Failure is clearly related.

throw helpers associated with the try. Remove an unneded arg on a folded overflow throw helper call. This is harmless on native targets but causes a sig mismatch issue for Wasm.

JulieLeeMSFT

LGTM. The fix extends loop intervals to the right to fully enclose try region's throw helper blocks (prevents loop/try straddle).

AndyAyersMS · 2026-06-11T01:34:58Z

/ba-g unrelated legs timing out

…tion (dotnet#129205) In FlowGraphTryRegions::Build, the BBJ_EHCATCHRET check sat below the early bailout on hasTryIndex, so the catchret of an outer catch was never marking its try region as requiring runtime resumption. Also: ensure loops that partially contain trys have their wasm-interval lengths extended to encompass any throw helpers associated with the try. Also: fix some random extra argument added to a the overflow helper in some cases. This is all platform and has been in the code forever. On native it is somewhat harmless to pass unexpected extra args (especially if register args). On Wasm it is not. Reminder that for Wasm control flow we can always move the end of a Wasm loop later; the extent just delineates where you can branch back to the loop head, not the actual extent of the loop. Likewise we can always move the start of a block earlier. --------- Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

…tion (#129205) In FlowGraphTryRegions::Build, the BBJ_EHCATCHRET check sat below the early bailout on hasTryIndex, so the catchret of an outer catch was never marking its try region as requiring runtime resumption. Also: ensure loops that partially contain trys have their wasm-interval lengths extended to encompass any throw helpers associated with the try. Also: fix some random extra argument added to a the overflow helper in some cases. This is all platform and has been in the code forever. On native it is somewhat harmless to pass unexpected extra args (especially if register args). On Wasm it is not. Reminder that for Wasm control flow we can always move the end of a Wasm loop later; the extent just delineates where you can branch back to the loop head, not the actual extent of the loop. Likewise we can always move the start of a block earlier. --------- Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

CopilotAI review requested due to automatic review settings June 10, 2026 01:50

Copilotstarted reviewing on behalf of AndyAyersMS June 10, 2026 01:50 View session

github-actionsBot added the area-CodeGen-coreclr CLR JIT compiler in src/coreclr/src/jit and related components such as SuperPMI label Jun 10, 2026

CopilotAI reviewed Jun 10, 2026
View reviewed changes

Comment threadsrc/coreclr/jit/flowgraph.cpp

kg approved these changes Jun 10, 2026
View reviewed changes

Ensure loop intervals that enclose trys have an extent that covers any
9620316
throw helpers associated with the try. Remove an unneded arg on a folded overflow throw helper call. This is harmless on native targets but causes a sig mismatch issue for Wasm.

JulieLeeMSFT added the arch-wasm WebAssembly architecture label Jun 10, 2026

JulieLeeMSFT approved these changes Jun 10, 2026
View reviewed changes

AndyAyersMS merged commit 7d42228 into dotnet:mainJun 11, 2026
122 of 148 checks passed

dotnet-maestroBot mentioned this pull request Jun 11, 2026
[main] Source code updates from dotnet/runtime dotnet/dotnet#7196
Merged

This was referenced Jun 12, 2026
Wasm / R2R: conv_ovf_i8_i aborts at runtime (try / catch [OverflowException] around conv.ovf.i) #129337
Closed
Wasm / R2R: IsFilterFunclet assertion fires when re-throwing from a finally after a cctor failure (throwinclassconstructor) #129341
Closed

dotnet-milestone-botBot added this to the 11.0-preview6 milestone Jun 17, 2026

github-actionsBot locked and limited conversation to collaborators Jul 18, 2026