Workaround for Scala 3 regression

[raquo]

Heyo! It turns out that newer Scala versions (verified in 3.8.4, 3.9.0, and 3.10-nightly) have a regression that makes Tuplez 0.5.0-M* throw unwarranted "unreachable case" compiler warnings when using in typical Airstream patterns – simple ones, nothing weird. I haven't noticed until now because 3.3.8 is unaffected.

Scala has a number of unresolved issues about "unreachable case" false positives, I'm not sure which one of them (if any) applies. I'll try to come up with a zero-dependency Scala reproduction later, but I'm out of tokens for now to polish it.

I do have a self-contained reproduction in Tuplez though. If you check out the "Add regression test for scala 3 bug" commit, it's just 0.5.0-M3. Run the test with sbt tuplezScala3Bug/compile, and you'll see it throw warnings on perfectly valid code:

sbt:tuplez> tuplezScala3Bug/compile
[info] compiling 1 Scala source to /Users/raquo/code/scala/tuplez/modules/shared/.jvm/target/scala-3.3.7/classes ...
[info] compiling 1 Scala source to /Users/raquo/code/scala/tuplez/modules/scala3-bug/target/scala-3.9.0/classes ...
[warn] -- [E030] Match case Unreachable Warning: /Users/raquo/code/scala/tuplez/modules/scala3-bug/src/main/scala/app/tulz/tuplez/bug/CompileBugRepro.scala:67:13
[warn] 67 |        case (n, preset, Some(a), Some(b)) => s"$n-$preset-$a-$b"
[warn]    |             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[warn]    |             Unreachable case
[warn] -- [E121] Pattern Match Warning: /Users/raquo/code/scala/tuplez/modules/scala3-bug/src/main/scala/app/tulz/tuplez/bug/CompileBugRepro.scala:68:13
[warn] 68 |        case _                             => "none"
[warn]    |             ^
[warn]    |Unreachable case except for null (if this is intentional, consider writing case null => instead).
[warn] -- [E030] Match case Unreachable Warning: /Users/raquo/code/scala/tuplez/modules/scala3-bug/src/main/scala/app/tulz/tuplez/bug/CompileBugRepro.scala:75:22
[warn] 75 |      .collect { case (ev, Some(x), Some(y), l, Some(z)) => (ev, x, y, l, z) }
[warn]    |                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[warn]    |                      Unreachable case
[error] No warnings can be incurred under -Werror
[warn] three warnings found
[error] one error found
[error] (tuplezScala3Bug / Compile / compileIncremental) Compilation failed
[error] Total time: 4 s, completed Sep 4, 2026, 2:56:31 a.m.

In the next commit, I have replaced Tuplez Scala 3 implementation with its Scala 2 implementation. If you run sbt tuplezScala3Bug/compile on that, the test will pass.

The setup of the test as a separate module is weird – that's because Claude determined that we need a compilation boundary to trigger the bug. I'll see if I can bypass that requirement later, but in practice, users of tuplez would be affected either way.

The downside of using Scala 2 implementation is that it's quite a bit more verbose – 4K lines. Claude proposed an alternative Scala 3 implementation that was approx 300 lines – using *: instead of Tuple.Concat, but even though it passed existing tests, I think it wasn't quite finished and required additional testing for edge cases. I haven't submitted that, I don't know if you have time to review all that. The Scala 2 implementation is at least proven.

I'm getting ready to release the next and probably final pre-release version of Laminar v18, and ran into this blocker.

I have verified that this branch fixed the compilation warnings on my codebase. I had 4 warnings in our pretty large codebase. The pattern was as in the test – a tuplez-using Airstream operator like combineWith followed by map or collect with a case .... argument.

If you want to test this yourself, remember that URL-DSL also depends on Tuplez. I checked and its master compiles fine with this branch, although I assume it's not binary compatible, so would need a publishLocal if you use it.

My suggested plan of action is to release this version of Tuplez, and later file a bug with Scala, and bring back your original Scala 3 implementation when the Scala bug is eventually fixed – I assume that'll take a while seeing that they've just released 3.9.0.

Also, please don't bump Tuplez Scala 3 version to 3.9.0 just yet, I haven't confirmed that all Laminar stuff supports it yet.

---

The "Add sbt-buildkit-dynver plugin" commit is not related to any of this. I just used it to automatically get reasonable dev version (0.5.0-M4-SNAPSHOT) when publishLocal-ing. Feel free to discard that commit or I can remove it if you'd rather not introduce a dependency for it. I'm adding it to all my projects as I got tired of copy-pasting PlayFramework's VersionHelper.scala file into every repo.

[raquo]

Oh, CI fails because githubWorkflowGenerate added the new bug-repro module's target directory to the list of directories to archive, but the test command does not compile that module (so its target dir does not get created), I made a separate workflow .yml for that. Will sort out later.