[the8472]

View all comments

The workaround applies on kernels >= 6.9 according to the manpage.
It doesn't work on GHA runners, I suspect virtualization issues, maybe a faulty virtio implementation.
But both on my local machine and on the dev-desktop it works.

running test_append_write_at under strace:

strace: Process 460033 attached
[pid 460033] uname({sysname="Linux", nodename="dev-desktop-eu-1.infra.rust-lang.org", ...}) = 0
[pid 460033] openat(AT_FDCWD, "/tmp/rust-2231989334/file_test_append_write_at.txt", O_WRONLY|O_CREAT|O_TRUNC|O_CLOEXEC, 0666) = 3
[pid 460033] write(3, "it's not working!", 17) = 17
[pid 460033] openat(AT_FDCWD, "/tmp/rust-2231989334/file_test_append_write_at.txt", O_WRONLY|O_APPEND|O_CLOEXEC) = 3
[pid 460033] pwritev2(3, [{iov_base="   ", iov_len=3}], 1, 5, 0x20 /* RWF_??? */) = 3
[pid 460033] openat(AT_FDCWD, "/tmp/rust-2231989334/file_test_append_write_at.txt", O_RDONLY|O_CLOEXEC) = 4
[pid 460033] openat(AT_FDCWD, "/tmp/rust-2231989334", O_RDONLY|O_NOFOLLOW|O_CLOEXEC|O_DIRECTORY) = 3
[pid 460033] +++ exited with 0 +++
[pid 460032] write(1, "{ \"type\": \"test\", \"name\": \"os::u"..., 92) = 92
[pid 460032] write(1, "{ \"type\": \"suite\", \"event\": \"ok\""..., 137) = 137
[pid 459907] write(1, ".", 1.)           = 1
[pid 459907] write(1, "\n", 1
)          = 1
[pid 459907] write(1, "\n", 1
)          = 1
[pid 459907] write(1, "test result: ok. 1 passed; 0 fai"..., 101test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 477 filtered out; finished in 794.28µs

somewhat fixes #140867

[rustbot]

r? @thomcc

rustbot has assigned @thomcc.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

[the8472]

Weird, the kernel should be new enough.

[bors]

☔ The latest upstream changes (presumably #143294) made this pull request unmergeable. Please resolve the merge conflicts.

[the8472]

At this point I'm fairly sure this is a kernel bug in the runner. Tests pass locally, the pwritev2 also happens in CI and doesn't return an error, meaning the flag should be supported.

Filed actions/runner-images#12530

[rustbot]

This PR was rebased onto a different main commit. Here's a range-diff highlighting what actually changed.

Rebasing is a normal part of keeping PRs up to date, so no action is needed—this note is just to help reviewers.

[the8472]

Still broken on ubuntu 26.06. I now suspect something is wrong with GHA, not the ubuntu kernel.

I'll just ignore those tests when detecting an azure kernel.

[the8472]

@rustbot ready