bpo-41818: Fix test_master_read() so that it succeeds on all platforms that either raise OSError or return b"" upon reading from master

[8vasu]

This replaces #23533. Build was failing on Solaris since, like the BSDs and Darwin, Solaris does not raise OSError in test_master_read(). Therefore, now we make test_master_read() succeed on all platforms that either raise OSError [ such as Linux ] or return b"" [ such as BSDs, Darwin, and possibly Solaris ] upon reading from master when the slave is closed. Any platform that does not exhibit such behavior can now be detected using test_master_read(). On any given platform, exiting from pty.spawn()'s copy loop depends on our exact knowledge of the behavior of test_master_read() on that platform.

Signed-off-by: Soumendra Ganguly soumendraganguly@gmail.com

https://bugs.python.org/issue41818

[8vasu]

@asvetlov this follows #23526. My original solution was #23533, but this is much better.

Edit: to be extra safe, I request you to run the buildbots again.

[kulikjak]

With this patch, test_pty on Solaris is green again.

[bedevere-bot]

🤖 New build scheduled with the buildbot fleet by @asvetlov for commit f994d09 🤖

If you want to schedule another build, you need to add the ":hammer: test-with-buildbots" label again.

[asvetlov]

to be extra safe, I request you to run the buildbots again.

Done

[asvetlov]

Thanks!

[8vasu]

You're welcome!

[8vasu]

@asvetlov #23546 is the next one. Of course, you can take your time to review it :D The tty module will now need a documentation update; I have not done that yet. Also, please let me know if I need to make tests for the tty module. After we finish working on that one, and before starting to make changes to the pty module, I will add one more nice test for the pty module which will check if pty.spawn() is capable of adjusting slave window size dynamically [ SIGWINCH ].