Assertion error: when pthread returns an error, this library crashes the entire program

[lovasoa]

Hello !
I work on SQLPage.
Stacker was recently introduced as a transitive dependency (through sqlparser).
I just received a bug report from an user that involves stacker: sqlpage/SQLPage#814

The problem seems to be in https://github.com/rust-lang/stacker/blob/04d334ca6ad8ef942fef80e173c6e34bbb905a6a/src/lib.rs#L418C30-L418C51 , where pthread_attr_getstack returns 2 (ENOENT).

I am not sure how to proceed, but wanted to report the issue here. Do you have an idea what may be going on and how to fix it ?

[lovasoa]

Ok, I can reproduce it locally now !

sudo mount -t tmpfs none /proc

./sqlpage.bin

WIth sqlpage.bin from https://github.com/sqlpage/SQLPage/releases/tag/v0.33.0

[nagisa]

yeah, I don't think glibc targets where procfs isn't mounted at /proc is something that's going to be supported like ever. But if you think it should be, that's a report for glibc.

[nagisa]

stacker could conceptually try to have some fallback like it does for systems that don't have a way to query stack size, but I don't think this corner case is particularly interesting for stacker to support… at that point you'd be better off using psm and creating your stacks yourself. Or perhaps corosensei.

[lovasoa]

Would it be hard to implement proper error handling? Crashing the entire program on startup, when the stack doesn't even need to grow, is not ideal...

I ended up with this crate as a sub dependency in a large program that worked without issues in this environment before...

[lovasoa]

@nagisa , I made a proposition to add proper error handling in https://github.com/rust-lang/stacker/pull/116/files

Let me know what you think !

[lovasoa]

That you very much for merging ! Could you publish a new version so that I can update my dependencies ?