On 2023-09-05 18:35:30, Jonas Vidra wrote: > On 2023-09-05 18:14:01, Jonas Vidra wrote: >> I will try building a vanilla 6.1.46 with the patch by John David Anglin >> applied and I'll report back. It might take me a while, though, as the >> PA-RISC system in question is not available 24/7. > > I did a dirty rebuild (no `make clean`) with the patch and it fixes the > issue as well. I was too hasty. The failure is still present on patched 6.1.46, although it is now intermittent. With a vanilla 6.1.46, I hit the issue perhaps 10 times, and it seemed that the problematic command fails every single time. On a patched 6.1.46, the problematic command ran 3 times but only failed twice – it finished fine on the first run. It might be just pure chance, though, if the actual failure frequency is something like 4/5 – it is possible that both versions are identically buggy. I have no idea if the Helge's tree I tested previously is similarly affected – the failure rate there is 0/4, which too might be due to just exceptionally good luck. But I still hope it is fixed there. > Is a minimal reproducer of the issue still desired? I can try to create > something from the strace output, but I'm not sure whether it's still > necessary. I'm now assuming it's really needed. I will prepare it, hopefully in the next few days.
