cockpit

cockpit copied to clipboard

cockpit-machines and subscription-manager-cockpit currently have outdated code copies of these. It makes more sense to move them to test/common, so that they can be shared and versioned similar to testlib.py.

Right now it's awkward to do that, as these can only be checked out to test/verify/ with our current mechanism, and that's not where every other project keeps their tests.

This touches the same files as PR #17645 -- I hope that they won't collide (as git is fairly bad with tracking renames), otherwise it's an easy enough rebase.

@skobyda , @ptoscano : Once this lands, you can add *lib.py to Makefile's COCKPIT_REPO_FILES and drop the code copies.

@ptoscano : Once this lands, you can add *lib.py to Makefile's COCKPIT_REPO_FILES and drop the code copies.

sub-man-cockpit does not use COCKPIT_REPO_FILES; that said, as long as it is in test/common/, it is enough for sub-man-cockpit.

@ptoscano : Ah right, that's another commit to backport :grin:

@ptoscano : Ah right, that's another commit to backport :grin:

Possibly, although it requires first updating the cockpit library to a newer version, which most likely means adapting to newer APIs to cockpit components (and I know for sure at least one change needed). Even if it's something I'd like to, it is definitely a bigger task.

@allisonkarlitskaya : Bah, this c8s failure is fallout from https://github.com/cockpit-project/bots/pull/3709 , sorry about that. It's not immediately clear to me what happens -- .write() definitively uses input=, and the value is "[Log]\nFatal = criticals\n", i. e. not even an empty string or similar corner case. There is a certain asymmetry:

stdin=None if input is not None else subprocess.DEVNULL,
input=input.encode("UTF-8") if input else None,

In the second line it should be if input is not None, but that only matters for input=="" and such. In this case I'd expect `stdin=None. I don't have time any more to debug this, smells like some Python 3.6 bug in RHEL/CentOS 8?

@allisonkarlitskaya : Bah, this c8s failure is fallout from cockpit-project/bots#3709 , sorry about that. It's not immediately clear to me what happens -- .write() definitively uses input=, and the value is "[Log]\nFatal = criticals\n", i. e. not even an empty string or similar corner case. There is a certain asymmetry:

stdin=None if input is not None else subprocess.DEVNULL,
input=input.encode("UTF-8") if input else None,

In the second line it should be if input is not None, but that only matters for input=="" and such. In this case I'd expect `stdin=None. I don't have time any more to debug this, smells like some Python 3.6 bug in RHEL/CentOS 8?

This looks like https://github.com/python/cpython/issues/79067 to me.

https://github.com/python/cpython/pull/11727/files

Fix is on the way in cockpit-project/bots#3734

@allisonkarlitskaya Friendly review ping? I can rebase, but it seems packit tests are now broken on rawhide, so would just trade one known red against another?