PtyProcess: add NO_CTTY flag

[petreeftime]

The default behavior of posix_openpt seems to be to replace the controlling terminal for the calling process, and PtyProcess should only change it for child instead. I think this might be a reason why some of the tests were failing non-deterministacally sometimes, although I am not 100% confident in this fix.

Signed-off-by: Petre Eftime [email protected]

[petreeftime]

bors try

[bors]

try

Build succeeded:

• CI

[petreeftime]

@matthiasbeyer so far this passes 100% of the tests.

[petreeftime]

bors try

[petreeftime]

Ah. This failed as well.

[bors]

try

Build failed:

• CI

[petreeftime]

bors try

[bors]

try

Build failed:

• CI

[petreeftime]

bors try

[bors]

try

Build succeeded:

• CI

[petreeftime]

bors try

[bors]

try

Build failed:

• CI

[petreeftime]

In pexpect, a similar issue is called out: https://github.com/pexpect/pexpect/blob/2532721644781543ca660e52d48a35bd93872fc1/doc/commonissues.rst#timing-issue-with-send-and-sendline

[petreeftime]

bors try

[bors]

try

Build succeeded:

• CI

[petreeftime]

bors try

[bors]

try

Build failed:

• CI

[petreeftime]

bors try

[bors]

try

Build failed:

• CI

[matthiasbeyer]

I am fine with the patch per se, if it helps. Only one annotation to simplify a bit of code here.

But as far as I can see, CI still fails? Or am I missing something?

[matthiasbeyer]

I assume the unsafe block is only needed for the ioctl call here, isn't it?
also, we don't need to match here, I think.

Suggested change

	unsafe {
	match ioctl(master_fd.as_raw_fd(), TIOCSCTTY) {
	0 => Ok(()),
	_ => Err(nix::Error::last()),
	}?;
	}
	if 0 != unsafe { ioctl(master_fd.as_raw_fd(), TIOCSCTTY) } {
	return Err(nix::Error::last())
	}

[petreeftime]

Yes, so far I was reading through how pexpect or some other terminal emulators set this up and see if we're missing anything. So far, not luck with getting the issue fixed unfortunately, and no matter what I do I can't seem to be able to reproduce it on my setup.

[matthiasbeyer]

Yeah that also was the case for me... it seems like we only have this issue in CI, which makes it even worse.