On 9/2/20 4:11 PM, Richard Henderson wrote: > On 9/2/20 12:09 PM, Thomas Huth wrote: >> On 01/09/2020 17.20, Richard Henderson wrote: >>> Version 2. Serves me right for not testing 32-bit host >>> when I knew there was a patch that mattered. >> >> Hi Richard, >> >> I'm afraid, but I think this PR broke the >> tests/acceptance/replay_kernel.py:ReplayKernel.test_microblaze_s3adsp1800 >> acceptance >> test: >> >> https://gitlab.com/qemu-project/qemu/-/jobs/716158589#L176 >> >> Could you please have a look? > > That's odd. > > Doesn't replay rely on migration (or at least VMStateDescription) to do its > job? Microblaze doesn't implement that, so how could this have worked before? > > I admit that I don't often run check-acceptance, because I rarely see a clean > bill of health, even for stuff I'm not working on. > > Looking into it...
Ah, in this case it's all about the icount setting. So the log recorded by replay=record is indeed useless for input to replay=replay, but not relevant. r~
