Update release flow for BCK prerelease and final release #12453
Conversation
Update BCK static files with latest.
Build Artifacts
|
| name: ${{ needs.whl.outputs.whl-file-name }} | ||
| path: dist | ||
| - name: Zip whl file | ||
| run: zip -j dist/kolibri.zip dist/${{ needs.whl.outputs.whl-file-name }} |
There was a problem hiding this comment.
Assuming that the filename looks like it does in the PR comment (kolibri-0.17.0b3.dev0+git.3.gb6d89f4b-py2.py3-none-any.whl), this looks like it'd install just fine, yep. Link to relevant lines in the CD dockerfile
| path: 'dist/kolibri.zip' | ||
| destination: '${{ secrets.BCK_PRERELEASE_BUILD_ARTIFACT_GCS_BUCKET }}' |
There was a problem hiding this comment.
Was initially concerned by this, but if the docs are telling the truth, should be uploaded flat. And so this may be alright.
| unzip -j dist/${{ needs.whl.outputs.whl-file-name }} 'kolibri/core/content/static/*' -d static | ||
| rm static/*.file_size | ||
| # Ungzip all .gz files in the static folder | ||
| for f in static/*.gz; do gunzip -f $f; done |
There was a problem hiding this comment.
yeesh you fancy with your bash functions
There was a problem hiding this comment.
This all looks fine to me, though you're doing more than the old buildkite step was doing; makes sense, given you're pulling from a WHL rather than the source code — do you not have access to the source code, or is it just safer to use the artifact we know successfully built?
There was a problem hiding this comment.
By this point in the step, the built source has been gzipped and truncated to keep the overall whl file smaller, so we have to do this to reinflate it back to uncompressed sources.
And yes, definitely safer to use the files we know have built successfully into the whl file.
| uses: 'google-github-actions/upload-cloud-storage@v2' | ||
| with: | ||
| path: 'static' | ||
| destination: '${{ secrets.STUDIO_BCK_CONTENT_STATIC_BUCKET }}' |
There was a problem hiding this comment.
The the bajillion examples of the not-immediately-intuitive behaviour on upload described in the docs has be squinting suspiciously at this. But if my understanding is right, all the extracted folders should start at the root of the bucket. Which is what I believe Kolibri-modified-to-point-at-a-bucket-for-content is expecting
There was a problem hiding this comment.
Yes - the docs are highly confusing, but this is also why I unzipped all the static files into the root of the current directory, so that I could just say "upload this folder". I am fully ready for my having misunderstood the docs again, again, again.
Summary
References
Fixes #12240
Reviewer guidance
The only part of this I was really able to test was the unzipping and ungzipping of the assets into a folder, everything else I have had to do by thinking I got it right.
Particular things to pay attention to:
needsjob in anifconditional for the whole job? (I don't know, and Github Actions wouldn't tell me)Testing checklist
PR process
Reviewer checklist
yarnandpip)