cmd/compile: "fatal error: found pointer to free object" on arm64 #71932
Comments
gopherbot
added
the
compiler/runtime
|
For added context, the test does pass on Go 1.23, and as noted above, I did a bisect down to 98e719f, but unfortunately that's where things went over my head. 🙇 ❤️ cc @randall77 ❤️ |
|
Thanks for the detailed report. |
|
@gopherbot please open a backport issue for 1.24. |
|
Backport issue(s) opened: #71938 (for 1.24). Remember to create the cherry-pick CL(s) as soon as the patch is submitted to master, according to https://go.dev/wiki/MinorReleases. |
|
Change https://go.dev/cl/652078 mentions this issue: |
randall77
added
the
Critical
|
Thanks @tianon ! And thanks @randall77 for picking this up ❤️ 🙈 not sure if it's just noise, or "related"; we got a report in docker/docker from a packager trying to use go1.24, and they ran into a SEGFAULT. We have not yet bisected that (or tried to reproduce), as we're currently on go1.23 ourselves, but could there be any relation? From moby/moby#49513 (comment) (but I think there's some other traces in the ticket) |
|
We hit this too: #71451 (comment) Thanks @tianon for reporting this! I attempted to do the same, but clearly I went to the wrong issue :) @randall77 although I wonder if that earlier issue could be related? The platform is different, but the panic looks very similar. |
|
Directly related, no. This issue is arm64 only. |
dmitshur
added
the
NeedsFix
|
Sorry, I meant to test this before it merged, but FWIW I've confirmed that the fix works in the codebase I was testing against. ❤️ Thank you! |
Thanks! Yes, fully understandable. It was really a bit of a drive-by comment from me, in hopes that possibly it would ring a bell (from the little information available) or if not, to ever so slightly spread awareness. We have not yet opened a ticket (and I don't think the original reporter has), as we currently don't have a small and consistent reproducer, and "build this giant project with obscure bash scripts (for |
Go version
go version go1.24.0 linux/arm64Output of
go envin your module/workspace:Details:
What did you do?
I wish I had a smaller reproducer, but even this one only reproduces sometimes so a bisect to 98e719f is the best I've been able to do (notaryproject/notary#1708 (comment)).
With a checkout of https://github.com/notaryproject/notary, running
GOARCH=arm64 go test -count 100 -run TestPublishDelegations ./client(on native arm64 hardware, emulated hardware, or even QEMU user-space emulation) should fail pretty reliably. I have not been able to reproduce a failure on any other architecture.What did you see happen?
notaryproject/notary#1708
(where most of the entries in the runtime stack are not related to the code being tested, and the ones that are don't have any obvious issues I see, and don't fail on any other architectures)
Full Log:
What did you expect to see?
The test passing. 😅
The text was updated successfully, but these errors were encountered: