Closed romaricpascal closed 9 months ago
One lead could be the checkout
action using main
instead of the actual branch the PR relates to: https://github.com/actions/checkout/issues/1418.
Started a little branch for investigation, logging the SHA that are checked out in the different scripts.
Looking at the logs (expanding "Checking out the ref" in the "Checkout code" step), the stats job checks out the same ref as the ref that the build checks out and the rest of out actions (the SHA referenced by github.sha
which is the merge of the PR head with the branch it'll be merged into).
To be continued...
Looks like our workflows are only building the npm package and not the files for our release archive. That would explain the sameness of the stats, as well as the 10K difference between local and CI.
Two potential fixes:
Updating the files being built and the cache paths so we build dist
as well.
npm run build --workspaces --if-present --ignore-scripts
npm run build:release --workspaces --if-present --ignore-scripts
The comment of the file stats reported the same file sizes on the PR that disabled mangling of our class and function names and the last one merged on main.
Running build:release on main and my branch locally does give me different file sizes (and a 10K smaller number than on the stats PR):