You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
When making a change to a submodule CI ran and it looks like it degenerated by now :/. CI here and in smmap is failing, but GitDB still works for some reason. Probably it's something small and obvious to you, so I didn't look into it yet as I have no clue.
Happy new year!
It looks like some things have broken on CI due to:
1. The ubuntu-24.04 images (which ubuntu-latest points to) not having
Python 3.7. This has been fixed in gitdb already but still needs to be
fixed in smmap and GitPython. In smmap the changes are analogous to those
that have been made to gitdb and we can add Python 3.13 at the same time
(as was done there). I'll open a PR on smmap shortly.
2. The Docker tests with Alpine Linux show an error in which attempting to
use `python -m pip` commands to install packages fails because the command
is running globally. This looks like it is due to some software being at a
higher version, but I am not sure what software. I think this shouldn't be
happening, since we are deliberately running the tests as a non-root user
in Docker, but it could be that I am misremembering how we set that up. In
any case, it should be possible to solve it by using a venv or by examining
why a venv is not being used if it persists even when attempting to use
one. This might relate to changes in user IDs in the test runner (outside
Docker), or to versions of Docker, of Alpine Linux, of Python, or of the
Python package in Alpine Linux separately from the Alpine Linux version. I
suspect it will not be too hard to solve.
3. Unrelatedly, the Alpine Linux image for WSL no longer successfully
downloading. This is only marginally faster than the Debian image we were
using for that before, so if updating the setup-wsl action doesn't fix it,
then the change from Debian to Alpine can be reverted.
4. Something going on with the number of lines read in a test on Cygwin. I
don't know what's going on with that, and maybe it's even random rather
than being related to anything that has durably changed. If it's not
random, then this might be the trickiest issue to solve.
5. Possibly other things. I haven't comprehensively examined everything
yet, but also, since some jobs are canceled due to fail-fast behavior,
there could be other failures that aren't evidenced yet.
Another interesting thing, though not likely related to any of the test
failures, is that I cannot load the comment I am replying to in a web
browser when logged in to GitHub. I can view it when logged out, but not
when logged in. I've tested in two Chromium-based browsers as well as
Firefox. I received a notification of the comment, as well as an email with
its text, but the comment itself is forever loading on the page, if I am
logged in to GitHub. To work around that, I am replying by email. (I hardly
ever comment by email on GitHub and I imagine I may not be getting the
formatting entirely right.)
Regarding the inability to use the web-frontend to reply, that sounds like a strange GitHub bug to me. Replying here seems to work for me at least, not that it means anything.
3 commit comments
Byron commentedon Jan 2, 2025
Hi @EliahKagan, all the best for the new year!
When making a change to a submodule CI ran and it looks like it degenerated by now :/. CI here and in
smmapis failing, but GitDB still works for some reason. Probably it's something small and obvious to you, so I didn't look into it yet as I have no clue.EliahKagan commentedon Jan 2, 2025
Byron commentedon Jan 2, 2025
Thanks again for your help!
Regarding the inability to use the web-frontend to reply, that sounds like a strange GitHub bug to me. Replying here seems to work for me at least, not that it means anything.