Experimental test workflow with devcontainer#585
Conversation
|
the |
|
ok, the dream of "ci/devcontainers works like vscode devcontainers" is unfortunately not realistic; I may take another stab at this but re-evaluating the strategy might make sense as well |
we are now running the vast majority of the tests, several of the skips are legacy mongodb, I will review the actual failures but we now have real test coverage again (and locally reproducible 1:1) |
|
needs a bunch of cleanup + running other workflows in devcontainer then will mark as ready for review |
821b458 to
65fbf8b
Compare
|
ok, I fixed the failing DB related tests/underlying behavior which was revealed by re-enabling the DB backends, migrating remaining actions to use the devcontainer |
|
most of the functionality is working great, I'm waiting to merge #584 to avoid too much divergence and we can get this wrapped up |
- AND takes precedence over OR in MySQL (https://dev.mysql.com/doc/refman/8.4/en/operator-precedence.html) which does not appear to be the intended logic
baa23d5 to
7fee74d
Compare
|
closing in favor of #586 586 |
MOTIVATION
the workflows in this repo were historically based on PL templates/work patterns, with complex matrix builds, multi-container orchestration, and a huge number of action dependencies
this resulted in repeated breakage due to actions changing, versions not being pinned fully/going out of sync, new golang releases changing linting rules, etc
numerous efforts were made to get that CI workflow working reliably (#443, #436, #467, #581, etc) but ultimately this all feels like threading water at best
to make matters worse, testing the workflows locally is a giant headache because even
actdoesn't provide an isomorphic environment to what happens on the runners and is not feature complete; normalmake testoutput diverges heavily between local dev and CIPROPOSAL
NOTE: we're using https://github.com/parkan/github-actions to run a rootless podman devcontainer because: