Marco Trevisan

Indeed you can hold a version in snap world too, you just have to install the snap with --revision and then the updates can be blocked. Also ideally we could...

We lock our errno implementation, so that's not of our tests doing it, but likely another syscall/libc call overriding it... Having it not parallel is an option though.

Marking a draft until newer libc is released

As update, this landed in libc 0.2.178, but... That also [broke nix tests](https://github.com/rust-lang/libc/pull/4806#issuecomment-3603204960), and so we need to wait for the new one that will contain https://github.com/rust-lang/libc/commit/d5b0f290a1c464fecc9710bb7fb1c8e011c1314a 🤯

See for example: https://github.com/ubuntu/authd/actions/runs/15642453208/job/44072755005 ``` helpers_test.go:328: Could not read artifact "/tmp/TestExecModuleItem_Set_user877767020/001/exec-module.log": open /tmp/TestExecModuleItem_Set_user877767020/001/exec-module.log: no such file or directory ``` The interesting part here is that `exec-module.log` is not there, but...

Oh, maybe that's `testing.go:1399: race detected during execution of test` ?!

My feeling is more related to races of pam.go itself here, maybe we can avoid by some os locking... Let's try

Maybe is something like this https://github.com/ubuntu/authd/actions/runs/15703467996/job/44243650461? Although that fails since the client crashes somehow...

> Is there any risk of this breaking existing core24 snaps using the gnome extension? The snaps no, builds it may... But well at rebuild time adjustments can be done....

> And I can't type 'r' to cancel the request and go back to the provider selection That's normal, once you chose to go through the local way, it's just...