Re: irreproducible python31x failures on NetBSD/i386

[Martin Husemann <martin%duskware.de@localhost>]

On Tue, Jan 07, 2025 at 09:38:00PM -0000, Michael van Elst wrote:
> gdt%lexort.com@localhost (Greg Troxel) writes:
> 
> >This is probably a question arising from not understanding, but if a
> >native NetBSD 9 i386 kernel has a facility, then I would expect a NetBSD
> >9 amd64 kernel with compat_netbsd32 to also have that facility.
> 
> It's about bugs in that compat_netbsd32 layer.

The PRs are 58215, 58225 and 58746.
Some pullups happened for 10.1:

sys/compat/netbsd32/netbsd32_compat_80.c        1.7,1.8 (patch)

        compat_netbsd32_80: add missing _90 dependency.
        [rin, ticket #987]

sys/compat/netbsd32/netbsd32_compat_16.c        1.5,1.6
sys/compat/netbsd32/netbsd32_signal.c           1.54   
        
        COMPAT_NETBSD32: fixes for signal handling.
        [rin, ticket #989]

But the whole thing is still more a hack (including what we have in
-current) than a proper solution, and also a quite confusing maze of
small issues that need to be untangled properly.

Martin