Cleanup/interface #218
Closed
Cleanup/interface #218
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Some changes needed so that Flatten can run after LambdaLift
To be combinable with follow-up mini-phases the lift operation needs to handle Thickets specially. This commit factors out the behavior from LambdaLift, so that Flatten can do the same thing.
Cleans up after LambdaLift and Flatten. RestoreScopes exhibited a problem (double definition) when compiling Unpickler. The root of the problem was in Applications.scala. The effect was that arguments woulkd be lifted out, but then the argument expression would be used anyway. That caused a closure to be present twice which caused the double def error much later. -Ycheck did not catch it because the two closure expressions were in non-overlapping scopes.
Threw away list of constructed statements.
Make private methods in traits static, so that we do not need to give a default for them.
Needs adaptations in getters/setters before it can be tested.
Trait getters are no longer needed.
When inheriting from Scala2 traits we sometimes encounter names with a space at the end, denoting a local variable. Drop the space because our translation scheme has no room for it.
Now also handles all supercalls. Seems to do the right thing on pos/traits.scala. But does not pass most tests because the sym transformer forces too many things.
Will print owners of symbols when displaying trees. Requires -Yprint-syms to be set also.
Motivation: Avoid needless forcing of symbols in scope. This is a problem when cloneScope is called in TreeTransforms.
The problem manifests itself when changing the owner of a class field. Being a ValDef, this is classified as a weak owner. But it's parent (the class) should not be owner-changed.
No more trait_setters are called.
Trait initializers do not have outer params; need to avoid passing outer args to them. This problem did not manifest itself before because no trait constructor calls were generated before erasure.
After Mixin, super calls can appear in statemenr sequences. They need to be typechecked and transformed using a special context.
Array types are no longer used after erasure, so the test was redundant.
Previous version could fail where the thisType widened to an explicit self type.
Since we demand that after erasure all TermRefs are SymDenotations we need to assure this when computing the denotations of term refs in new phases.
Otherwise we get binding races, e.g. for pos/t2133.scala after mixin because we create same-named symbols in subclasses, and type refs now can refer to the subclass symbol where previously they refereed to the superclass symbol. To avoid a data race the new term ref will have a fixed symbol.
Need to have Method flag set
This is reasonable because getters are only generated late in the compilation pipeline. No need to hide a getter as a val in error messages.
1. Erase unit results in getters to BoxedUnit. 2. Erase => Unit to ()Unit; was ()BoxedUnit 3. Make sure ValDefs have same type in tpt as in symbol, analogous to DefDefs.
Separate test instead of inline.
Now survives the case where a field is written x_=, dissumulating a setter.
Split into two phases, ResolveSuper before Erasure and Mixin after. Likewise GettersSetters is split into Getters and Memoize. All tests pass, except two tests fail when compiled twice. Will investigate next why.
Needs a weakening of an assertion in assertErased
Need to be treated like TermRefs.
Ig an implicit method has a result type which does not match the expected type, the new case fires. We avoid to diagnose missing arguments, because that's really confusing for an implicit method.
The fix to Types in 9bb3551 (Retract of optimization in computeDenot) allowed the two tests to compile "twice" again.
Changes owners after a phase without copying the tree. This should be more suitable for the changeOwner operations used in the tree transforms so far, which are linear, i.e. no tree duplication is needed.
It functionality is now all implemented in ExplicitOuter.
Previous implementation was confused about the meaning of interface (with default methods
or without?).
Now instead of Interface/JavaInterface we have
PureInterface - all members are abstract methods
NoInits - all members are methods
|
Needs rebasing over rebased mixin. I'll do it now. |
Merged
WojciechMazur
pushed a commit
to WojciechMazur/dotty
that referenced
this pull request
May 8, 2025
Backport "Filter help renders box border" to 3.3 LTS
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Depends on mixin PR. Review by @DarkDimius