packaging: Integrate github workflows to automate the air-gapping process of wave Tour #2199
Conversation
There was a problem hiding this comment.
Thanks @sulhicader. Just 1 small comment from me and should be good to go!
.github/workflows/release-wave.yml
Outdated
| run: | | ||
| echo "VERSION=${{ env.VERSION }}" >> "$GITHUB_OUTPUT" | ||
|
|
||
| - name: Make the folder structure |
There was a problem hiding this comment.
This introduces unwanted code duplication as this code can also be found in make build-apps target. Please refactor so that we have a single make build-tour-app target that can be reused here and also in make build-apps.
There was a problem hiding this comment.
Good Suggestion. Thanks @mturoci
|
@mturoci reason for the workflow failure is "ERROR: No matching distribution found for h2o_wave==1.1.1". Are there any time gap to install the h2o_wave latest release from the PYPI after publishing. As I understood some steps in release job of the workflow push the h2o_wave to PYPI. And the failed job start once the release job is finished. |
|
Yes, seems like the problem is PyPI needs more time between publish and subsequent download to serve the updated file. |
|
One thing We can do is using the wheel file to install the h2o_wave. For docker images We only need the linux version. If it is same for the tour wave bundle publishing to the s3 then We can simply add an extra line to copy the whl file. Or another way came to my mind is, add an extra step with sleep command in failed job to delay the image building task if it is a short and known time to become the PYPI h2o_wave available. |
|
Installing from wheel sounds okay to me (for docker images only). |
The PR fulfills these requirements: (check all the apply)
mainbranch.feat: Add a button #xxx, where "xxx" is the issue number).Closes #xxx, where "xxx" is the issue number.uifolder, unit tests (make test) still pass.Closes #2180