Skip to content

Emit generic signature for static forwarders to nullary methods #12710

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 2 commits into from
Jun 15, 2021
Merged

Emit generic signature for static forwarders to nullary methods #12710

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
43ca070
Emit generic signature for static forwarders to nullary methods
lrytz Jun 4, 2021
5c8188a
Fix generic signatures for static forwarders involving value classes
lrytz Jun 4, 2021
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion compiler/src/dotty/tools/backend/jvm/BCodeHelpers.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -926,7 +926,7 @@ trait BCodeHelpers extends BCodeIdiomatic with BytecodeWriters {
// (one that doesn't erase to the actual signature). See run/t3452b for a test case.

val memberTpe = atPhase(erasurePhase) { moduleClass.denot.thisType.memberInfo(sym) }
val erasedMemberType = TypeErasure.fullErasure(memberTpe)
val erasedMemberType = TypeErasure.transformInfo(sym, memberTpe)
if (erasedMemberType =:= sym.denot.info)
getGenericSignatureHelper(sym, moduleClass, memberTpe).orNull
else null
Expand Down
2 changes: 1 addition & 1 deletion compiler/src/dotty/tools/dotc/core/TypeErasure.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -697,7 +697,7 @@ class TypeErasure(sourceLanguage: SourceLanguage, semiEraseVCs: Boolean, isConst
// See doc comment for ElimByName for speculation how we could improve this.
else
MethodType(Nil, Nil,
eraseResult(sym.info.finalResultType.translateFromRepeated(toArray = sourceLanguage.isJava)))
eraseResult(rt.translateFromRepeated(toArray = sourceLanguage.isJava)))
Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This was needed for test run/t3452g

The backend calls TypeErasure.transformInfo(<Symbol TraversableLike.tail>, <Type ExprType(AbstractTrav[String])>).

Erasure should, I assume, continue to work on AbstractTrav[String] (which is rt here). Calling sym.info.finalResultType gives Repr which erases differently.

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Apparently that was intentionally changed in 76af6c5

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Tests passed...

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Sure, but at the very least we need to carefully check if this breaks binary compatibility.

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Very true. How could I do that? Compile the community build twice and do a jardiff?

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I ran dotty.communitybuild.CommunityBuildTestA twice; once with this PR, once with the line here reverted. This resulted in 14081 identical classfiles.

I also added a check to see when the old and new erasure are different and it only happened when called from the backend:

+          val a = eraseResult(rt.translateFromRepeated(toArray = sourceLanguage.isJava))
+          val b = eraseResult(sym.info.finalResultType.translateFromRepeated(toArray = sourceLanguage.isJava))
+          if (!(a =:= b) && !(new Exception).getStackTrace.exists(_.toString.contains("getStaticForwarderGenericSignature"))) then
+            throw new Exception(s"[ERASURE]\n- $a\n- $b\n- $tp\n- ${sym.showFullName}")
           MethodType(Nil, Nil,
-            eraseResult(rt.translateFromRepeated(toArray = sourceLanguage.isJava)))
+            a)

I ran the sbt test and CommunityBuildTestA with this in place and did't see the exception showing up.

@smarter any other ideas?

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks, I've run a similar test on the whole CI and it also didn't show any situation where the two types differed, this makes sense since the tp passed to eraseInfo should always be the symbol info at that point (it's only called from transformInfo which is only called from Erasure#transform, which always uses the symbol info except for java.lang.Object which is treated specially), so LGTM 🎇

case tp1: PolyType =>
eraseResult(tp1.resultType) match
case rt: MethodType => rt
Expand Down
4 changes: 4 additions & 0 deletions tests/pos/i10347/A_1.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,4 @@
trait L[+T] { def head: T }
object A {
def foo: L[String] = ???
}
3 changes: 3 additions & 0 deletions tests/pos/i10347/C_2.java
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,3 @@
public class C_2 {
String hi = A.foo().head();
}