Fix #10085: check if enum case confirms to scrutinee type

[liufengyun]

Fix #10085: check if enum case confirms to scrutinee type
Fix #9190: infer union types

[liufengyun]

test performance please

[dottybot]

performance test scheduled: 6 job(s) in queue, 1 running.

[dottybot]

Performance test finished successfully:

Visit http://dotty-bench.epfl.ch/10094/ to see the changes.

Benchmarks is based on merging with master (22c23a5)

[smarter]

The latest commit in this PR is:

Fix CI: add facility to infer union and singleton types

• infer singleton types: tests/patmat/i4227.scala
• infer union types: tests/patmat/i9190.scala

• I checked out the previous commit and tried scala3-compiler/testOnly *PatmatExhaustivityTest and no test failed, how can I reproduce the problem?
• tests/patmat/i9190.scala doesn't exist

[liufengyun]

The latest commit in this PR is:

Fix CI: add facility to infer union and singleton types

• infer singleton types: tests/patmat/i4227.scala
• infer union types: tests/patmat/i9190.scala

• I checked out the previous commit and tried scala3-compiler/testOnly *PatmatExhaustivityTest and no test failed, how can I reproduce the problem?

You can try testOnly -- *.patmatExhaustivity.

• tests/patmat/i9190.scala doesn't exist

Good catch, it's now added.

[smarter]

You can try testOnly -- *.patmatExhaustivity.

I think that runs the same set of tests, anyway I tried it and I still don't get any error on commit 149d4de, (I do get an error for i9190 if I add it manually, but nothing for i4227)

[liufengyun]

You can try testOnly -- *.patmatExhaustivity.

I think that runs the same set of tests, anyway I tried it and I still don't get any error on commit 149d4de, (I do get an error for i9190 if I add it manually, but nothing for i4227)

For i4227, we only get an error if we remove the following lines (removed in the 2nd commit):

          // force type inference to infer a narrower type: could be singleton
          // see tests/patmat/i4227.scala
          mt.paramInfos(0) <:< scrutineeTp

The code was there just to make type inference infer a singleton type.

[smarter]

I see, thanks. I was surprised that mt.paramInfos(0) <:< scrutineeTp didn't work for the union case, so I investigated and found that this was caused by a bug for which I made a PR: #10106. Once this is merged we shouldn't need c245d33 (we can separately discuss if this is still useful, but it wouldn't be needed for this PR).

[liufengyun]

I see, thanks. I was surprised that mt.paramInfos(0) <:< scrutineeTp didn't work for the union case, so I investigated and found that this was caused by a bug for which I made a PR: #10106. Once this is merged we shouldn't need c245d33 (we can separately discuss if this is still useful, but it wouldn't be needed for this PR).

Thanks for going deep into the issue. Once that's merged, I'll rebase and drop the 2nd commit for now.

[smarter]

rebase and drop the 2nd commit for now.

Done.

[liufengyun]

rebase and drop the 2nd commit for now.

Done.

Thanks, I also pushed some simplifications from the 2nd commit in 6a9c4ea.