run cleanupRelay only once in TerminalIO and StandardIO

[elijah-wright]

possibly fixes #124

[elijah-wright]

@adityaramani I think the formatting issues have been fixed

[adityaramani]

Im gonna pull in the changes locally and test it out as well

[elijah-wright]

thank you, I've been depending on the workflow tests and it's not fair to you that they're not passing

[adityaramani]

No worries at all!

[adityaramani]

Tried running the tests in a loop and it failed with this, and there werent any other logs from the guest

2025-06-18T14:50:23-0700 error com.apple.containerization : [containerization_integration] ❌ test multiple concurrent processes with output stress failed: internalError: "failed to start process" (cause: "internal error (13): startProcess: failed to start process: Error Domain=NSCocoaErrorDomain Code=256 "(null)"UserInfo={NSUnderlyingError=Error Domain=NSPOSIXErrorDomain Code=9 "Bad file descriptor"}")

I'll try combining your change with what I was testing out with the writeFd being deinited

[elijah-wright]

dang, I really thought that was going to solve it. I'm going to look at this again when I get home, I got caught in a literal sandpit 💀

[adityaramani]

Thanks! I think your changes look good - and I tested it after adding in a couple of other fixes (see here https://github.com/adityaramani/containerization/tree/vminitd-crash)

Although I do think there maybe some other issues in the process handling flow, this change makes it more stable.

Will wait for the others to review

[elijah-wright]

@adityaramani do you want me to cherry pick your commit? it's only partially signed so I don't know

[adityaramani]

Yeah go for it. Take that commit in. If the signature are a problem just apply the diff

[elijah-wright]

ok done, but it says unverified

[adityaramani]

Yeah. Maybe try removing me as author?

[elijah-wright]

@adityaramani if I could ask, what are the specs of the machine you're testing on?

[adityaramani]

It's an M2 Pro with 64 gb of ram. Are you asking cause you don't see the issue on your end?

I need to run the test on a loop and see it failing once in like 15 times

[elijah-wright]

I was just wondering, I think it has to be a race condition then. increasing system load (stress-ng) might make it more predictable for testing if that's true

[dcantah]

Great catch 😆

[dcantah]

This looks good to me (and sorry for delay), but I'll wait for either @adityaramani or @crosbymichael to give it a second thumbs up.

[adityaramani]

LGTM! I think Michael should take a look as well