Feature request: Catch segfault and show stacktrace in result

[roblabla]

Motivation

When writing unit tests for FFI or unsafe code, it can often be the case that the test will fail not with a panic, but with a segfault. In this case, the logs don't provide much information, only giving a single line "(test aborted with signal 11: SIGSEGV)", requiring more in-depth investigation to debug the issue. It'd be nice if cargo nextest could give a bit more information to allow a first level of investigation.

Proposal

It'd be nice for cargo-nextest to catch sigsegv signals that happen in the child process, and print a stacktrace of the crashing thread.

Alternatives

No response

Additional context

No response

[sunshowers]

Thanks. This is somewhat tricky -- is your use case CI or local?

I think the first step here would be writing a little script or library which does this (on Linux/glibc and macOS this is probably an LD_PRELOAD shim that injects a signal handler), and passing it in as either a wrapper script or --debugger on the command line. That might give you a sense of how difficult this is. We're probably not going to have this feature by default but I'm open to being able to automatically configure this as long as it is suitably non-invasive and lightweight. Is that something you feel like you can do?

I had Claude build a prototype that seems to work on Linux, though I haven't looked at the code at all; feel free to take it from there or start from scratch. Anything that eventually ships with nextest will probably need to be pure Rust so cargo install continues to work -- I think a little build.rs that compiles a cdylib is a promising approach.