Bug#1085704: auto-multiple-choice: race condition in d/t/gui causes sometimes autopkgtest failures

[Emanuele Rocca <ema@debian.org>]

Hi,

On 2024-10-23 11:12, Paul Gevers wrote:
> On 23-10-2024 10:40, Emanuele Rocca wrote:
> > Indeed the simplest repro is just running the following in a sid armhf
> > schroot running on a amd64 host with binfmt qemu-user-static emulation
> > on:
> 
> For avoidance of doubt, on ci.d.n for armhf; we have a bare metal arm64 host
> with multiple arm64 VM's which use lxc and not qemu to get an armhf testbed.

Ah yes, I meant the simples repro *for me*. :-)

Turns out that there's a new libunwind upstream version, 1.8.1. I gave
it a go, and I can't reproduce the segfault anymore with it.

To package the new upstream I just dropped all patches, which seem
included in the new version from a cursory look:

- 0001-Add-a-step-limit-to-Ltest-init-local-signal.c.patch
- 0001-Fix-FTBFS-on-Linux-ppc32.patch
- 0001-Fixed-miscompilation-of-unw_getcontext-on-ARM.patch

The new upstream also fixes the other libunwind RC bug: #1082284.

Adrian, could you please upload 1.8.1 when you have a chance?

Thanks,
  Emanuele