Quick lunchtime fix. This probably breaks overriding a method of literal type. (Edit: that works after all. I was wondering about the asymmetry of the fix.)

@som-snytt is this one you're thinking of reviving...?

@SethTisue yes, although like all old relationships, I've forgotten what was so attractive about it.

The bug fix is to handle FoldableConstantType correctly in reflect.internal.transform.Erasure.

The rest of the commit is style adjustments while reading code.

It is difficult to see the history above, but the first swing was about matching members; lrytz hinted it's just an erasure problem; and at some point I made that tweak.

A duplicate ticket from a decade ago shows that the symptom has been noticed for some time; that ticket said specialized, so may have been neglected on that basis; it is unknown whether mysterious runtime errors were due to this behavior; OTOH, how often does one override with a constant result.

Edit: maybe this was what gave me pause:

[error] ## Exception when compiling 26 sources to /home/runner/work/scala/scala/build/quick/classes/sbt-bridge
[error] java.lang.ClassFormatError: Method HASH_SIZE in class scala/reflect/internal/Names has illegal modifiers: 0x402

Edit: that was constructors detecting whether to drop RHS, resulting in abstract member.

Ran the sbt test locally, thanks to instructions in build.sbt.

[info] [info] compiling 1 Scala source to /tmp/sbt_6b9da7a3/target/classes ...
[info] [error] /tmp/sbt_6b9da7a3/A.scala:3:20: type mismatch;
[info] [error]  found   : Int
[info] [error]  required: String
[info] [error]  val z: String = B.y
[info] [error]                    ^
[info] [error] one error found
[info] [error] (Compile / compileIncremental) Compilation failed
[info] [error] Total time: 0 s, completed Apr 28, 2025, 1:53:32 PM
[error] x source-dependencies/false-error [PENDING]
[error]  Cause of test exception: {line 7}  Command failed: compile failed

It's supposed to recompile and notice B.y is now a String. From the test name, it looks like erroring was the old behavior.

Or is it supposed to be skipped as "pending" not "test"? What's going on? (That is, I have to learn something new again?)

All the PENDING are failures.

The repording for those pending tests is very confusing :-/

[error] x source-dependencies/false-error [PENDING]
[error]  Cause of test exception: {line 7}  Command failed: compile failed

is its way of saying "I ran source-dependencies/false-error and it failed, as expected, since it's marked pending".

So the scripted tests look clean.

The real error seems to be

[error] Test scala.reflect.ClassOfTest.classOfUnitConstant failed: java.lang.ClassFormatError: Duplicate method name "f" with signature "()Ljava.lang.Class;" in class file __wrapper$4$8c32535ff2d0445a8346dfeff9e261a4/__wrapper$4$8c32535ff2d0445a8346dfeff9e261a4$C$1, took 0.094 sec

D'oh, I didn't scroll back far enough. It's so hard to keep up with these important computering skills.

The other important skill is when the comment says "special case", to preserve the special case.

My other style preference that has evolved is to avoid long lines where the important bit is pushed too far to the right. Less important side effects can be pushed right, speaking typographically not politically. Then vertical alignment becomes a matter of aligning important bits, as we learned when writing out addition with a pencil.