| Commit message (Collapse) | Author | Age | Files | Lines |
... | |
|\ \ \ \ \ \ \
| | | | | | | |
| | | | | | | | |
Fix #1747: Improve error message for Scala/Java type mismatch
|
| | | | | | | | |
|
| | | | | | | | |
|
| | | | | | | | |
|
| | | | | | | | |
|
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | | |
We used to "fill-in" with zeroes if variances were missing. I now
think that;'s too error-prone. Better define all variances explicitly.
|
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | | |
Omit the `=>' if a PolyType has a MethodType as result type.
|
|\ \ \ \ \ \ \ \
| |_|_|_|_|/ / /
|/| | | | | | | |
Don't inline when errors are detected
|
| | |_|_|/ / /
| |/| | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
Inlining is only well-defined if the body to inline does not
have any errors. We therefore check for errors before we
perform any transformation of trees related to inlining.
The error check is global, i.e. we stop on any error
not just on errors in the code to be inlined. This is a safe
approximation, of course.
|
|\ \ \ \ \ \ \
| |/ / / / / /
|/| | | | | | |
Fix #1990: Handle inlining where this proxies change types
|
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
It turns out that we simply cannot do reliable outer path
computation that fills in the right hand sides of this-proxies
from the types of these proxies. As-seen-from logic can mangle
the types of proxies enough to scramble the necessary information.
What we now do instead is simply count: We record the number
of outer accesses to an outer this in inlineable code, and do the same number
of outer accesses when computing the proxy.
|
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
They are sorted according to the nesting depth of the classes they
represent. This is no necessarily the same as the nesting level of
the symbols of the proxy classes. i1990a.scala shows an example where
the two differ.
|
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
Interesting that the tests pass even if we always assume outOfContext = true.
So this raises the question why have a flag? It's just that I am not sure the
`outOfContext` behavior is correct in all cases. So I prefer to be conservative
here.
|
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
This shows that the builder pattern can be expressed
with implicit function types.
|
| | |_|_|_|/
| |/| | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
The new situation in the test was that outer of the inlined method
was `A` but it's as seen from type is a subtype `B`.
We need two fixes:
- Ignore outerSelects in TreeChecker. These are treated as having fixed symbols.
- Adapt the outer-path logic to deal with code that's moved to another context.
|
|\ \ \ \ \ \
| | | | | | |
| | | | | | | |
mixed left and right assoc
|
| | | | | | | |
|
| | | | | | | |
|
|/ / / / / / |
|
|\ \ \ \ \ \
| | | | | | |
| | | | | | | |
Remove unused flags
|
| | | | | | | |
|
| | | | | | | |
|
| |/ / / / / |
|
|\ \ \ \ \ \
| |_|_|/ / /
|/| | | | | |
Use enum for error messages IDs.
|
| | | | | | |
|
|\ \ \ \ \ \
| |_|_|_|/ /
|/| | | | | |
Disable tests that require scala-compiler
|
| | |/ / /
| |/| | |
| | | | |
| | | | | |
This is necessary if we ever want to get rid of our dependency on scala-compiler
|
|\ \ \ \ \
| | | | | |
| | | | | | |
TreeMap/TreeAccumulator: proper context for inlined trees
|
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
This was already be done in TreeTraverser but should also be done in
TreeMap and TreeAccumulator for ctx.error(..., tree.pos) to not use
completely incorrect positions inside inlined trees.
|
|\ \ \ \ \ \
| | | | | | |
| | | | | | | |
checkNoPrivateLeaks: Use correct position for errors
|
| | |/ / / /
| |/| | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
Previously we never used the `pos` argument of `checkNoPrivateLeaks` and
instead used `sym.pos`, this makes a difference for calls to
`avoidPrivateLeaks` coming from `TreeUnpickler` where we should use
`tree.pos` instead.
|
|\ \ \ \ \ \
| | | | | | |
| | | | | | | |
Fix #1991: Use classtag where available in unapply
|
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
We cannot assume that the untyped rhs of the bind is a `Typed` tree,
with extractors it might be an `Apply` node, and in general it might
also be a `Parens` node.
|
| |/ / / / / |
|
|\ \ \ \ \ \
| |/ / / / /
|/| | | | | |
Bootstrapped dotty should not depend on non-bootstrapped dotty-library
|
|/ / / / / |
|
|\ \ \ \ \
| |_|/ / /
|/| | | | |
Factor out logic for scala functions.
|
| |/ / /
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
* Add `isSyntheticFunction` checks for synthetic functions such as FuntionN
for N > 22 and ImplicitFunctionN for N >= 0.
* Add `erasedFunctionClass` to get the erased verion of synthetic functions.
* Change the semantics of `isFunctionClass` to return true if it is any kind of
FunctionN or ImplicitFunctionN.
|
|\ \ \ \
| | | | |
| | | | | |
Change '... expected but found ...' to Message
|
|/ / / / |
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
This is a gross hack to support the need for a $scope
binding when XML literals are created. It should go away
once we phase out XML literals. A library-based
solution should use implicits instead.
|
|\ \ \ \
| | | | |
| | | | | |
Fix #1975: Align valdefs and for expressions for patterns
|
| | |/ /
| |/| |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
val definitions and for expressions both distinguish whether
something is a pattern or a variable binding. They no do it
the same way: `ident` or an `ident: type` is a variable binding,
everything else is a pattern. Previously, capitalized idents
were considered as bindings in valdefs but as pattern in fors.
|
| | | |
| | | |
| | | |
| | | | |
The non-determinism is anonying, disable for now until we
have a better solution.
|
|\ \ \ \
| |/ / /
|/| | | |
increase command wait time to 1s
|
|/ / /
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
This change helps to reduce non-determinism in the test. The problem
can happen when two consecute commands interfere:
[expect] [jdb]
cmd1 ---->
sleep 0.5 <--- part 1 of rep1
read & match buffer
clear buffer
cmd2 ---->
<--- part 2 of rep1
|
|/ /
| |
| |
| |
| |
| |
| |
| |
| |
| | |
* Change 'forward reference extending over the definition' to Message
* Change 'forward reference extending over the definition' to Message
* pesky file should not be included
* Change 'forward reference extending over the definition' to Message (test case)
|
|\ \
| | |
| | | |
Fix #1501 - Check trait inheritance condition
|
| | |
| | |
| | |
| | | |
It's java.io, not java.lang.
|
| | |
| | |
| | |
| | |
| | | |
Excepted are only Serializable and Comparable. This follows
scalac's behavior.
|