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
I'm slightly surprised/saddened at the size of cc1bc6f (from #40176), modifying the one build script to make the output between the two scripts more similar, and changing 1502 generated files, given my old PR managed to merge the two scripts with only 138 generated files changed in total. Maybe by the point that it landed it was impossible to have a smaller diff, but it still feels very unfortunate that we went from merging the scripts being something that was done and didn't change many generated files at all, to something where even reducing the diff between the two scripts has involved changing an order of magnitude more generated files.
cc @graveljp, who seems to have done most of the updates since my early 2022 PR.
The text was updated successfully, but these errors were encountered:
I'm sorry I didn't see #33613 before. Things have changed a lot since 2022. It is indeed unfortunate that these scripts got forked in the first place. I originally tried to maintain them in sync as I improved them, but this was like pulling a dead weight since our goal was to delete the old generator. In hindsight, merging them would indeed have been a better idea, but it's too late for that, the new script generator has seen a lot of improvements and new features in the last years.
The old test generator is almost unused now. The only tests that it still generates are the ones in html/canvas/tools/yaml/element/meta.yaml and .../offscreen/meta.yaml. I migrated one of these meta test in 215c5cc. I can try to find some time soon to complete that migration and delete the old generator.
Regarding the number of files changed, we have not been concerned with the impact on generated files. The whole point of a code generator is to be able to generate or modify lots of redundant code with a minimal configuration. The fact that the generated files are submitted in the repository is a mistake in my opinion. It's causing lots of problems, like noise in code review or generated files getting out of sync with the generator. It would be better if file were generated as part of the build or by the test runner.
I previously did this in #33613, but lacking review, this never landed. Since then, the two scripts have been substantially modified (https://github.com/web-platform-tests/wpt/commits/master/html/canvas/tools/gentestutils.py, https://github.com/web-platform-tests/wpt/commits/master/html/canvas/tools/gentestutilsunion.py).
I'm slightly surprised/saddened at the size of cc1bc6f (from #40176), modifying the one build script to make the output between the two scripts more similar, and changing 1502 generated files, given my old PR managed to merge the two scripts with only 138 generated files changed in total. Maybe by the point that it landed it was impossible to have a smaller diff, but it still feels very unfortunate that we went from merging the scripts being something that was done and didn't change many generated files at all, to something where even reducing the diff between the two scripts has involved changing an order of magnitude more generated files.
cc @graveljp, who seems to have done most of the updates since my early 2022 PR.
The text was updated successfully, but these errors were encountered: