Compiler crash when annotated types contain certain trees #17939
Assignees
Labels
Comments
mbovel
added
itype:bug
itype:crash
stat:needs triage
mbovel
changed the title
|
Output when adding |
|
Could the culprit be the following? Should we map symbols in this case? |
|
Just to see if it changes something, I tried to always map symbols to fresh ones in diff --git a/compiler/src/dotty/tools/dotc/transform/PostTyper.scala b/compiler/src/dotty/tools/dotc/transform/PostTyper.scala
index ac3dc15092a..43409e5125b 100644
--- a/compiler/src/dotty/tools/dotc/transform/PostTyper.scala
+++ b/compiler/src/dotty/tools/dotc/transform/PostTyper.scala
@@ -19,6 +19,7 @@ import config.Feature
import util.SrcPos
import reporting._
import NameKinds.WildcardParamName
+import dotty.tools.dotc.ast.TreeTypeMap
object PostTyper {
val name: String = "posttyper"
@@ -84,6 +85,12 @@ class PostTyper extends MacroTransform with IdentityDenotTransformer { thisPhase
val synthMbr: SyntheticMembers = new SyntheticMembers(thisPhase)
val beanProps: BeanProperties = new BeanProperties(thisPhase)
+ def withFreshSymbols(tree: Tree)(using Context): Tree =
+ val tm = new TreeTypeMap():
+ override def withMappedSyms(syms: List[Symbol]): TreeTypeMap =
+ withMappedSyms(syms, mapSymbols(syms, this, mapAlways = true))
+ tm(tree)
+
private def newPart(tree: Tree): Option[New] = methPart(tree) match {
case Select(nu: New, _) => Some(nu)
case _ => None
@@ -135,7 +142,9 @@ class PostTyper extends MacroTransform with IdentityDenotTransformer { thisPhase
val saved = inJavaAnnot
inJavaAnnot = annot.symbol.is(JavaDefined)
if (inJavaAnnot) checkValidJavaAnnotation(annot)
- try transform(annot)
+ try
+ val res = transform(annot)
+ if res != annot then withFreshSymbols(res) else annot
finally inJavaAnnot = saved
}With this change, the file compiles, but the pickling test fails: $ diff before-pickling.txt after-pickling.txt
50,52c50,51
< x: scala.Int@tests/pos/tasty-no-address.scala<149..152>@tests/pos/tasty-no-address.scala[146..152]@@(x=
< tests/pos/tasty-no-address.scala:<146..146>)
< ): scala.Boolean =
---
> x: scala.Int@tests/pos/tasty-no-address.scala<149..152>@tests/pos/tasty-no-address.scala<149..152>@@(x=?))
> : scala.Boolean =
64,66c63,65
< tests/pos/tasty-no-address.scala<157..163>
< .x:scala.Int>@tests/pos/tasty-no-address.scala<157..165>
< .==:((x: scala.Int): scala.Boolean)>@tests/pos/tasty-no-address.scala<157..168>
---
> tests/pos/tasty-no-address.scala<157..162>
> .x:scala.Int>@tests/pos/tasty-no-address.scala<157..162>
> .==:((x: scala.Int): scala.Boolean)>@tests/pos/tasty-no-address.scala<157..162>
69,71c68,70
< @tests/pos/tasty-no-address.scala[145..154..170]@@($anonfun=tests/pos/tasty-no-address.scala:<154..154>)
< <closure(<$anonfun:(x: scala.Int): scala.Boolean>@tests/pos/tasty-no-address.scala<170..170>):scala.Int =>
< scala.Boolean>@tests/pos/tasty-no-address.scala<170..170>
---
> @tests/pos/tasty-no-address.scala<149..152>@@($anonfun=?)
> <closure(<$anonfun:(x: scala.Int): scala.Boolean>@tests/pos/tasty-no-address.scala<152..152>):scala.Int =>
> scala.Boolean>@tests/pos/tasty-no-address.scala<152..152>
73c72
< :scala.Int => scala.Boolean>@tests/pos/tasty-no-address.scala<145..170>
---
> :scala.Int => scala.Boolean>@tests/pos/tasty-no-address.scala<149..152> |
|
The Maybe we should special case the |
|
Minimizing a different failing case, I found the following, which I think is related: import scala.annotation.Annotation
class myRefined(f: ? => Boolean) extends Annotation
def test(axes: Int) = true
trait Tensor:
def mean(axes: Int): Int @myRefined(_ => test(axes))
class TensorImpl() extends Tensor:
def mean(axes: Int) = ???Both the lambda and the subclassing appear to be necessary Here is the error: Click to expand
|
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Compiler version
348729e
Minimized code
Output (click arrow to expand)
Trees with IDs after
posttyperLook at the blocks in the annotations in
A: one has ID#1182and one has the ID#1202, but the innerdef $anonfunand associated symbol are the same, hence the problem. The typesString @myRefinedare also different.In
A2however, both blocks have ID#1032and bothString @myRefinedhave id#1232.The text was updated successfully, but these errors were encountered: