Fix #4758: Revert "New phase for entering annotations" #4776
Conversation
This reverts commit 2b12868. Since we do not plan to run annotation macros during Namer, there's no need to annotate before completing. This allows again to define an annotation in the same scope as an annotated definition. So i3702.scala goes from neg to pos.
| @@ -822,6 +746,22 @@ class Namer { typer: Typer => | |||
| else completeInCreationContext(denot) | |||
| } | |||
|
|
|||
| protected def addAnnotations(sym: Symbol): Unit = original match { | |||
| case original: untpd.MemberDef => | |||
| var hasInlineAnnot = false | |||
| protected def addAnnotations(sym: Symbol): Unit = original match { | ||
| case original: untpd.MemberDef => | ||
| var hasInlineAnnot = false | ||
| lazy val annotCtx = |
There was a problem hiding this comment.
While you're at it, you may want to make an annotCtx def that can be used both here and for https://github.com/lampepfl/dotty/blob/64bd0fe5210cacd7c17ab285c80712944157c212/compiler/src/dotty/tools/dotc/typer/Typer.scala#L1345 :)
|
Should we also revert #1249? |
Good point to check this also. I just had a close look and think it's better to leave it as is. With #1249, we process annotations before marking the symbol as completed. That prevents some races we would otherwise have to worry about. We know then that, once a symbol was completed, it will have up-to-date annotations. |
odersky
force-pushed
the
undo-annotations-phase
branch
from
5e47371 to
0c35b8f
Compare
The last commit caused double definitions of the Annotations class, so the context was too far out. I don't have the time to dig deeper now, so for the moment I am reverting. |
This reverts commit 2b12868.
Since we do not plan to run annotation macros during Namer, there's no need
to annotate before completing. This allows again to define an annotation in the
same scope as an annotated definition. So i3702.scala goes from neg to pos.