Emit mixin forwarders as ordinary, non-bridge methods again #21890
Conversation
sjrd
force-pushed
the
emit-mixin-forwards-as-non-bridges-again
branch
2 times, most recently
from
0af3b78 to
9a896d5
Compare
This reverts commit 700eea7. Forward port of scala/scala@96f844e
This reverts commit 6d0f9ca. Forward port of most commits in scala/scala#8037 Compensate some of the consequences by adding the `MixedIn` flag. In addition to the use seen in the diff in `BCodeHelpers`, the JS backend has an existing test for `isOneOf(Bridge | MixedIn)` which needs this compensation.
This is a forward port of scala/scala@6fc2202 In Scala 2 that commit was reverting a commit that had stopped emitting Java generic signatures. In dotc we never had that logic before, so this is new code.
sjrd
force-pushed
the
emit-mixin-forwards-as-non-bridges-again
branch
from
9a896d5 to
66f90c6
Compare
|
The only remaining community-build test failure seems to be in scala-parallel-collections in SerializationStabilityTest.scala. After checking the failure, it seems to be a more-or-less expected consequence of this PR. Different generated classifies generated here cause a different SerialVersionUID to be used (which happened even when I modified the PR to only change flags, keeping the types of the methods same as before). The only solution is to stabilize the UID and change the test - thankfully there is a precedent for it: I’ve submitted a PR that fixes this, so we can then bring it into the dotty-staging fork: scala/scala-parallel-collections#303 About the Mima errors, there are two types here:
|
Forward port of scala/scala#8037
Reverts #6352 and #6141.
Fixes #19270.