Robert O'Callahan

This test records with `-c100`. The last event recorded was `rrcall_init_buffers` on the non-main thread in a buffered call to `gettimeofday()`, after which the assert triggers at a syscall in...

If the signal was received and stashed any time before the exit from `rrcall_init_buffers` then `prepare_to_inject_signal` should have tried to inject it. `is_safe_to_deliver_signal` should have returned true since `rrcall_init_buffers` is...

Looks like we stopped at 0xf79b7 here in glibc: ``` 00000000000f79b0 : f79b0: b8 1c 00 00 00 mov $0x1c,%eax f79b5: 0f 05 syscall f79b7: 48 3d 01 f0 ff...

I suppose we could save extra data in `StashedSignal` so the assertion can report more information about when/why the signal was stashed.

d0254a93a4b094b8c74da759a3c749ac2c3371c2 collects a bit more information if we hit this again.

I haven't published more about remix, no. Pernosco doesn't have any plans to open-source remix at this time. If you're using it for work we might be able license it...

I haven't touched a Mac for years and neither has Kyle AFAIK so some Mac user will have to investigate this. To be clear, which hypervisor are you talking about?

Looks like there is no VDSO present on this system, hence the vsyscall was selected. That is very old stuff that I don't want to worry about supporting ... though...

As for the actual patch here, this is very tricky code. What is supposed to happen if the task is already exited, is that waitpid will return ECHILD, in which...

> I don't know more about why it works that way, just have recognized that with the old code the actual process stood at "defunc" and rr waiting "forever" (I...