| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
| |
another Symbol::tpe bites the dust (should'be been tpeHK)
|
|
|
|
|
|
|
|
| |
Bringing BigInt and BigDecimal into the club of things which can be
equal to one another and which will have the same hashCode. Fixed
some old and some new bugs associated with equality. Note: not fully
optimized.
|
|
|
|
|
|
| |
Removed everything deprecated in 2.7.3 or earlier except the lower case
primitive type aliases, plus associated fixes.
|
| |
|
| |
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
| |
two underlying problems:
- isAsSpecific did not skolemize lower (left) type in subtyping check (instead used withTypeVar on left and right)
- withTypeVars did not clone the symbols of the type params (so they were not fresh)
(the second fix is not essential due to the first fix, it "improves correctness", but should check whether performance is not impacted too severely)
applied martin's documentation diff
|
|
|
|
|
|
|
| |
Fixed #2606
Fixed #2598
Fixed #1836
|
|
|
|
|
|
| |
avoid illegal forward references by moving synthetics to the beginning
of the statement-list. fixes #2489
|
| |
|
|
|
|
|
| |
Allow protected access to constructor in superclass. Closes #1836
|
| |
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
| |
The preferred way to convert between primitives is to call .toInt etc,
but there were lots of casts for historical reasons. This patch remedies
that.
|
| |
|
|
|
|
|
|
|
|
| |
Fixed one of the issues in the BuildManager mentioned in #2590, where
adding a missing reference wasn't causing compilation of the dependent
files. Still, the solution to the whole ticket requires some changes to
the scala plugin itself.
|
|
|
|
|
|
| |
Fixed issue #451 by updating structural method dispatch logic to work
with new arrays.
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Breaks down the hash function in BoxesRunTime by type so we can do as
much as possible at compile time. Documents various trouble points I've
identified with trying to get the hashCodes aligned. Test case exercises
the hashCode functions.
Deleted all the code associated with previous equality adventures. Don't
worry, I can put anything back if it turns out we have to change course
again, but for now it's noise.
Also, gives return types to the box and unbox methods which are added to
the primitive companions, so e.g. Int.box(5) now returns a j.l.Integer
instead of an Object.
|
|
|
|
|
| |
attempt to fix build -- can't build on my old macbook, so flying blind :-/
|
|
|
|
|
| |
isDefinedAt should /not/ be in Function1
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
fixed bug in implicit resolution that only manifested itself when
multiple implicit arguments needed to be resolved and they were intended
to instantiate type parameters
two problems:
- type parameters that could not be inferred where removed from undetparams erroneously
- the successfully inferred parameters were not propagated to the the implicit arguments on the right (implicit resolution searches for implicit arguments from left to right, fixing type parameters in the process)
this should give the green light for the addition of Zipped to TupleN
|
|
|
|
|
|
| |
added overloaded hashes to Predef.
some small changes.
|
| |
|
|
|
|
|
|
|
|
| |
code that mixes in PartialFunction now have to
define isDefinedAt as override.
Fixes #2225.
|
|
|
|
|
| |
Fixed the optimizer for inlining correctly ScalaRunTime.inlinedEquals
|
|
|
|
|
| |
Equality checks go to ScalaRunTime.inlinedEquals when -optimise is on
|
| |
|
|
|
|
|
| |
Fixed #2517
|
|
|
|
|
| |
Some @experimental cleanups/removals I missed on the last pass.
|
|
|
|
|
|
| |
Fixes issue #1006 by marking structural method caches as `volatile` (and
the array of parameter types as `final`).
|
|
|
|
|
|
| |
Tighter type checking rules for structural types that fix issues #967,
#1004, #1388, #1494, and #1906.
|
| |
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
| |
Broke up foldBinOp logic a little bit looking to reduce the likelihood
of inliner pathology. This also knocked about two minutes off my time to
build optimised locker.
|
|
|
|
|
|
| |
Sped up build times by three minutes with a brilliant optimization I
like to call "comment out offending code."
|
|
|
|
|
|
| |
- Renames GenericRange to NumericRange
- Removes most @experimental annotations
|
|
|
|
|
|
| |
Fixed initial variable binding for method parameters, that caused the
inliner to infer wrong types for tail recursive methods
|
| |
|
| |
|
|
|
|
|
|
| |
BuildManagerTest no longer throws npe when some of the files do not
exist
|
| |
|
| |
|