> … > This is wrong at least for QEMU_CLOCK_HOST. > … > Reading the host clock here is not protected by the checkpoint. > Therefore it may incur the inconsistency when replaying the execution.
That's why I didn't like idea of this patch and asked for any possible side effects beforehand. So, here is one. I have no idea how to fix it properly and would like to recall this particular patch from these series (remember, it doesn't match series goal). Lets leave this patch/issue for separate resolution, although I'm not very interested in it.