Re: rust on 32-bit arm

[Havard Eidnes <he%NetBSD.org@localhost>]

> Some bad news, I'm afraid.
>
> I could not get rust 1.79.0 to self-host on my NetBSD 9.0_STABLE
> earmv7hf qemu host.
>
> As soon as the "stage 1 -> stage 3" build started, I get
>
> Uplifting library (stage1 -> stage2)
> WARNING: no codegen-backends config matched the requested path to build a codege
> n backend. HELP: add backend to codegen-backends in config.toml.
> Uplifting rustc (stage1 -> stage3)
> using sysroot /usr/pkgsrc/wip/rust179/work/rustc-1.79.0-src/build/armv7-unknown-
> netbsd-eabihf/stage2
> Building tool cargo (stage2 -> stage3, armv7-unknown-netbsd-eabihf)
> ...
>    Compiling libc v0.2.153
> ...
> rustc exited with signal: 11 (SIGSEGV) (core dumped)
> Did not run successfully: signal: 11 (SIGSEGV) (core dumped)
> ...
> My next attempt is going to be by using pkgsrc gcc 10 in the rust
> 1.79.0 build on 9.0_STABLE earmv7hf, as the next attempt in the
> effort to get this to build on 9.x.

I've re-tried by using both gcc10 and gcc13 as the C and C++
compiler for the rebuilds, and in both cases the result was
essentially the same as above: on 32-bit armv7hf on NetBSD 9.x,
there appears to be something other than the C++ compiler which
makes this combination fail, since if I do basically the same on
a 10.0 armv7hf system, it *can* self-host.

This already has the use of thread-local storage turned off, so
that should not be the issue either.

I'm nearing my wit's end here, sadly.

Anyone else have any brilliant (or silly) ideas about how to get
around this problem?


On the somewhat brighter side, rust-bin 1.79 on armv7 NetBSD 9.x
appears to work on this combination, at least sufficiently to
build a working dua-cli application, so there is that...


Regards,

- Håvard