On Fri, Nov 14, 2025 at 7:11 AM Thomas Huth <[email protected]> wrote: > > I'll look at it today; if I can't make heads or tails of it we'll revert it, > > yes. > > Sorry if I missed it, but did you find out anything? ... the test is still > failing for me, which is somewhat annoying ...
It was failing and when I started debugging it, it fixed itself. :( I did pin it: 1) to the _recording_ side. That is, the broken recording cannot be loaded by a good QEMU and a good recording can be loaded by a broken QEMU; 2) to the write of the WAIT system register, which is where the CPU goes to sleep and never gets the event to wake up. The reproducer is ./qemu-system-alpha -display none -vga none -machine clipper -serial stdio -kernel kernel-alpha -append 'printk.time=0 console=ttyS0 -net none -no-reboot -nodefaults' -icount shift=9,rr=replay,rrfile=replay-bad.bin Also, I wrote a patch to add trace events for instructions written to the replay stream, which I will send. Paolo
