[ci] Run python-so-copying.yml on Linux and Mac both [WIP]
#2208
Conversation
Codecov Report
Additional details and impacted files@@ Coverage Diff @@
## main #2208 +/- ##
==========================================
- Coverage 77.68% 72.13% -5.56%
==========================================
Files 137 103 -34
Lines 10679 6875 -3804
Branches 214 215 +1
==========================================
- Hits 8296 4959 -3337
+ Misses 2297 1818 -479
- Partials 86 98 +12
Flags with carried forward coverage won't be shown. Click here to find out more.
|
| container: | ||
| image: ubuntu:22.04 | ||
| image: ${{ matrix.os }} |
There was a problem hiding this comment.
This is the tricky part. This needs to be a Docker image, not the name of the GitHub Actions runner.
The linux job runs inside of a Docker container (this made it easier for me to test locally). Thus it's not as easy as adding macos-12 to the build matrix. We need to either 1) make the container directive only apply to the ubuntu job (probably possible, but not immediately clear to me how), or 2) create a second job that runs similar code on macOS (which might be worth it given all the changes to accommodate running the same steps on linux and macOS)
There was a problem hiding this comment.
@jdblischak I'm totally happy with well-documented code-duplication here.
There was a problem hiding this comment.
I did find some workarounds to create a build matrix that builds natively on macOS but in Docker on Ubuntu. There is no if, so the key is to pass null to container for the macOS job. Overall the readability is not great though.
Also, I realized that we'll also need to download a different pre-built libtiledb binary for the macOS build, which makes the option to have a completely separate macOS build all the more appealing.
johnkerl
changed the title
python-so-copying.yml on Linux and Mac bothpython-so-copying.yml on Linux and Mac both [WIP]
Issue and/or context: My current theory on TileDB-Inc/tiledbsoma-feedstock#95 and TileDB-Inc/tiledbsoma-feedstock#91 is that this code path, not being regression-tested on MacOS, is perhaps faulty on MacOS. Regardless, the logic which was developed on #1937 needs testing on MacOS, and it was an oversight on #1937 for us to have only tested on Linux.
See also #2210 for tracking.
See also #2219 for tracking.
Changes:
Notes for Reviewer: