tests crashes on NetBSD sparc64

[0323pin]

Hi,

NetBSD package maintainer here.

Got the following report, https://mail-index.netbsd.org/pkgsrc-bugs/2024/09/22/msg073230.html from an user. Unfortunately, I have no clue on how to fix this.

Any thoughts?

[snshn]

Hi there,

thank you for reporting this issue, and thank you for maintaining the NetBSD package!

I'm investigating this right now, suspecting the latest version of reqwest to be the cause. Might downgrade it back to what it was before the breakage.

[0323pin]

@snshn Thank you!

I'm investigating this right now, suspecting the latest version of reqwest to be the cause. Might downgrade it back to what it was before the breakage.

If this turns out to be the case, it might be good to report it upstream, so it can be fixed in the reqwest crate.

[als-git]

One data point:

• amd64 is little endian byte order and relaxed memory access rules (i.e. unaligned memory access "only" costs performance)
• sparc64 is big endian byte order and strict memory access rules (i.e. unaligned memory access causes a CPU fault that is usually signaled to user space as SIGBUS, causing program abort)

At least for C code naively written for Intel and then built on SPARC, these two are a common source of bugs. I'm not familiar with Rust though, so no idea how that effects this language.

[snshn]

@0323pin could you please let me know if it's just sparc64 that crashes? I tested it on NetBSD with --no-default-features (non-static OpenSSL linkage), it seems to be doing fine, but it's x86_64, I suspect @als-git is correct about it being problem specific to sparc64 due to its endianness

[0323pin]

@snshn I've switched the package to use the static linkage on one of the last updates. I only have x86_64 hardware myself and, that you already know.

I haven't seen other problem reports on the subject.

If you still want me to try on x86_64, just let me know.