Stream: git-wasmtime

Topic: wasmtime / PR #10973 Test out s390x inline assembly chang...

Wasmtime GitHub notifications bot (Jun 06 2025 at 23:32):

alexcrichton opened PR #10973 from alexcrichton:test-s390x to bytecodealliance:main:

Extracted from #10960

prtest:linux-s390x

<!--
Please make sure you include the following information:

• If this work has been discussed elsewhere, please include a link to that
  conversation. If it was discussed in an issue, just mention "issue #...".

• Explain why this change is needed. If the details are in an issue already,
  this can be brief.

Our development process is documented in the Wasmtime book:
https://docs.wasmtime.dev/contributing-development-process.html

Please ensure all communication follows the code of conduct:
https://github.com/bytecodealliance/wasmtime/blob/main/CODE_OF_CONDUCT.md
-->

Wasmtime GitHub notifications bot (Jun 06 2025 at 23:59):

cfallin commented on PR #10973:

I'm pretty perplexed staring at this one too -- for reference here are Godbolt links for before (working) and after (apparently segfaulting). The clobber-saves come in different order wrt loading values into payload (r6/r7) but the dataflow from arguments to PC/SP (r15)/payload (r6, r7) looks completely equivalent...

Wasmtime GitHub notifications bot (Jun 07 2025 at 00:16):

alexcrichton commented on PR #10973:

cc @uweigand would you be able to help dig in to what's going on here? (we can provide more background context if necessary, but the general gist is that this PR in theory is a refactoring that shouldn't do anything but it causes tests to segfault)

Wasmtime GitHub notifications bot (Jul 02 2025 at 17:31):

uweigand commented on PR #10973:

I was finally able to reproduce this. Turns out the crash only happens when building with CARGO_INCREMENTAL=0.

Wasmtime GitHub notifications bot (Jul 02 2025 at 17:33):

uweigand submitted PR review.

Wasmtime GitHub notifications bot (Jul 02 2025 at 17:33):

uweigand created PR review comment:

Here's the problem. The br instruction requires an address register operand (i.e. GPR 1-15, but not GPR 0). In the failing case, the compiler chose to use GPR 0 to hold pc. To fix this, you'll need to use the reg_addr register class instead of reg for pc.

Wasmtime GitHub notifications bot (Jul 02 2025 at 19:16):

alexcrichton updated PR #10973.

Wasmtime GitHub notifications bot (Jul 02 2025 at 20:36):

alexcrichton updated PR #10973.

Last updated: Dec 06 2025 at 06:05 UTC