Send SIGSTOP to linux/android processes before doing other procfs/ptrace things. #108
Conversation
things. If this fails, we continue as we used to. This is an attempt to get a consistent/static process state. Closes rust-minidump#28.
Not sure this is needed, but better to allow users to customize this rather than rely on a hardcoded value
|
I had a couple of comments about this but wasn't fast enough, I'll put them in a separate issue. |
|
Yeah later last night I also realized that |
|
Oh right, I thought about something like starting if the bool was true or checking of the process has reached a stopped state in the interim, but I assume that sigcont is idempotent ? |
|
It also might not hurt to just always send SIGCONT, but it'd be cleaner to be symmetric. I think as long as the SIGSTOP goes out, there's no reason to think the process won't stop (it's unblockable after all) even if we don't observe that. |
|
Given the processes will frequently be in the same process group it might be possible to |
Do you want to add yourself to https://github.com/rust-minidump/minidump-writer/blob/main/.github/CODEOWNERS so that you get auto-added as a reviewer? Then you can just remove the request if you don't have time/etc. |
Yes, thank you! |
If this fails, we continue as we used to. This is an attempt to get a consistent/static process state.
Closes #28.
Needs to be tested in all sorts of scenarios (but due to the timeout and the fact that it continues on failure, one would hope it'd only be an improvement).