And I was the last one who ever touched the JsModifier.scala in mdoc :sweat:
I'll take a look - it might be something to do with using the linker as incremental in mdoc - I tried to reproduce by effectively running mdoc-js via CLI (against some laminar snippets) but couldn't - I'll try again with your minimization
I'll ask Sebastien as well - Mdoc basically mimics incremental compilation of SJS by reusing a lot, as long as classpath remains unchanged.
Worst scenario, some of that stuff can be moved behind a flag and a fresh linker used for each file. It will be slower but I think will 99% mimic what SJS does in batch mode.
I've started fleshing out a repro for @raquo's issue here: scalameta/mdoc#455
Sadly the naive approach doesn't seem to work, tests are still passing
I have another approach of intercepting a CLI command one of my tool makes and trying to make it match the reproducible in the issue.
I think the closest we will get is to get a CLI reproducible: https://scalameta.org/mdoc/docs/js.html#command-line
Admittedly, it's quite hairy and that's why I wanted to reproduce it in tests first.
Is this blocking you from upgrading laminar to 1.3x (asking because I thought you wanted to postpone that)
git release 1.2.4script like this
❯ cat `which git-release` set -eux version=$1 git tag -af "v$version" -m "v$version" && git push -f origin v$version
crossScalaVersionsto cross-build the
docssbt project against multiple versions, but that would give you independent copies of the entire website