Fix #3259: Generate overriding bridges in synthesized SAM wrappers. #3260
Conversation
|
Thanks! I was going to poke at this over the weekend in case you were busy, but this is better ;) |
sjrd
force-pushed
the
fix-sams-with-overriding-bridges
branch
from
4871252 to
ae1a0af
Compare
|
Refer to this link for build results (access rights to CI server needed): https://scala-webapps.epfl.ch/jenkins/job/scalajs-pr/4498/ |
|
Refer to this link for build results (access rights to CI server needed): https://scala-webapps.epfl.ch/jenkins/job/scalajs-pr/4501/ |
|
Dear Jenkins, retest this please |
sjrd
force-pushed
the
fix-sams-with-overriding-bridges
branch
from
ae1a0af to
4748ffc
Compare
|
Rebased. |
sjrd
changed the title
|
Refer to this link for build results (access rights to CI server needed): https://scala-webapps.epfl.ch/jenkins/job/scalajs-pr/4503/ |
project/Build.scala
Outdated
| (if (isJSTest) ((testDir / "require-sam") ** "*.scala").get else Nil) | ||
| } | ||
|
|
||
| val hasBugWithOverridenMethods = |
| } | ||
|
|
||
| for (sam <- samInfo.sam :: samBridges) { | ||
| if (seenEncodedNames.add(encodeMethodSym(sam).name)) |
There was a problem hiding this comment.
Because the JVM backend does something similar (they call .disctinct here). I guess this would happen if a SAM type extends several interfaces, each defining the same method, which would then collapse as the same bridge. The most reliable way for us to detect duplicates at this point is to test uniqueness of the encodedName.
There was a problem hiding this comment.
Hmmm... OK. Not very nice, but fair enough.
…ppers. Just like Scala/JVM needs to instruct the `LambdaMetaFactory` to generate bridges for methods overridden by the SAM, we need to directly generate appropriate bridges in our synthesized SAM wrappers. Instead of generating actual *bridges*, in the sense that they call the canonical implementation of the method, we generate bodies for all the bridges just like for the canonical method. This is fine because the bodies are just calling the closure stored in a field of the instance anyway.
sjrd
force-pushed
the
fix-sams-with-overriding-bridges
branch
from
4748ffc to
b6be908
Compare
|
Updated. |
|
Refer to this link for build results (access rights to CI server needed): https://scala-webapps.epfl.ch/jenkins/job/scalajs-pr/4504/ |
Just like Scala/JVM needs to instruct the
LambdaMetaFactoryto generate bridges for methods overridden by the SAM, we need to directly generate appropriate bridges in our synthesized SAM wrappers.Instead of generating actual bridges, in the sense that they call the canonical implementation of the method, we generate bodies for all the bridges just like for the canonical method. This is fine because the bodies are just calling the closure stored in a field of the instance anyway.
Only the last commit is part of this PR. It will have to be rebased on top of #3258.
Tested locally with: