Rust code segfaults with stack overflow

[Volker-Weissmann]

If I'm not mistaken, writing an infinite recursing function, is guaranteed to produce this message:

thread '' has overflowed its stack
fatal runtime error: stack overflow

I recently accidentally wrote an infinite recursing function (src/ir/item.rs:982) and I got this message instead:

process didn't exit successfully: `/home/volker/rust-bindgen/target/debug/deps/tests-ec5d637d33ad84bf stack_overflow` (signal: 11, SIGSEGV: invalid memory reference)

To be fair, rust_bindgen does contain unsafe code, but I still suspect that this SIGSEGV is a bug in rust's handling with segmentation fault, because the call stack of the crash looks like this:

Standard library
Functions that do not contain unsafe code
A few thousand times the `is_constified_enum_module` function
Functions that contain unsafe code
start

How to reproduce:

$ git clone https://github.com/Volker-Weissmann/rust-bindgen --branch stack_overflow --depth 1
$ cd rust-bindgen
$ cargo test stack_overflow

I'm using rustc 1.50.0-nightly (d32c320 2020-12-10)

Note: There is also a forum post about this:
https://users.rust-lang.org/t/is-rust-guaranteed-to-detect-stack-overflows/52593/12

[nagisa]

is guaranteed to produce this message

We do not guarantee this. We only guarantee (on targets that have stack probes available) that overflowing the stack will touch a guard page, crashing the process (on x86_64 linux with SIGSEGV). Detection of whether a crash is an actual stack overflow is done on a best-effort basis. There are a number of prerequisites for the detection code to work, and therefore plenty of valid reasons this detection could fail.

I'm going to close this out. If you think it should remain open regardless, please let me know and I'll reopen.

[Volker-Weissmann]

Ok, then the guy in the forum was wrong.