| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
| |
Don't force the owner info.
|
| |\
| |
| | |
Fixes SI-6150 - backport to 2.10.x branch.
|
| | | |
|
| |\ \
| | |
| | | |
SI-5330, SI-6014 deal with existential self-type
|
| | |/
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
This has been broken since https://github.com/scala/scala/commit/b7b81ca2#L0L567.
The existential rash is treated in a similar manner as in fc24db4c.
Conceptually, the fix would be `def selfTypeSkolemized =
widen.skolemizeExistential.narrow`, but simply widening before
narrowing achieves the same thing. Since we're in existential voodoo
territory, let's go for the minimal fix: replacing `this.narrow` by
`widen.narrow`.
--
Original patch by @retronym in #1074, refined by @paulp to
only perform widen.narrow incantation if there are
existentials present in the widened type, as
narrowing is expensive when the type is not a singleton.
The result is that compiling the entirety of quick, that
code path is hit only 143 times. All the other calls hit
.narrow directly as before. It looks like the definition
of negligible in the diff of -Ystatistics when compiling
src/library/scala/collection:
< #symbols : 306315
---
> #symbols : 306320
12c13
< #unique types : 293859
---
> #unique types : 293865
I'm assuming based on the 2/1000ths of a percent increase
in symbol and type creation that wall clock is manageable,
but I didn't measure it.
|
| |\ \
| |/
|/| |
sane printing of renamed imports
|
| | |
| |
| |
| |
| | |
Having a select named "foo" with an underlying symbol named "bar"
and trying to make sense of all that by prettyprinting is very confusing
|
| |\ \
| | |
| | | |
SI-6539 Annotation for methods unfit for post-typer ASTs
|
| | | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
- Don't default the message, and show it exclusively.
- Fix cut-and-pasto in the @since tag
- Be tolerant if the annotaion class is missing, as seems to
have been the case compiling the continuations plugin.
- s/\t/ / in the test file to show the errors are positioned
correctly.
- Use defensive getOrElse
|
| | | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Motivated by the `.value` method in the SBT task-syntax branch,
which should only be called within the context of the argument
to a setting initialization macro.
The facility is akin to a fatal deprecation.
|
| | |/
|/|
| |
| |
| | |
Don't give up before you try tupling. Who knows what
someone might be doing with a Unit.
|
| |/
|
|
| |
Inspired by https://issues.scala-lang.org/browse/SI-6649
|
| |\ |
|
| | |\
| | |
| | | |
Fix for SI-6600, regression with ScalaNumber.
|
| | | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Not much in the end; I divided ScalaNumericConversions
into two traits such that the ScalaNumericAnyConversions can
be used in value classes, and ScalaNumericConversions can
override methods in ScalaNumber (since one trait cannot do
both those things.)
The fact that ScalaNumber is privileged for equality but a) extends
java.lang.Number and therefore b) cannot be a value class is something
we will want to revisit real soon.
|
| | |/
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
e6b4204604 moved access widenings from ExplicitOuter to
SuperAccessors to reflect them in pickled signatures so
that the inliner can take advantage of them under separate
compilation.
The followup discussion [1] determined that this wasn't
the right solution: while it enabled new separate compilation
inlinings, it failed to widen access of outer pointers and
hence prevented certain inlinings.
A better solution was proposed: modify the inliner to know
that access widening is guaranteed to have happened in
ExplicitOuter for any field accessed by an @inline-d method
body, rather than relying solely on the pickled types.
But this hasn't happened yet. In the meantime 07f94297 / #1121
reinstated the access widening to SuperAccessors, but took a
slightly different approach, using `Symbol#enclMethod` rather
than `closestEnclMethod`. That deviation triggers SI-6562.
This commit goes back to `closestEnclMethod`.
[1] https://groups.google.com/forum/#!topic/scala-internals/iPkMCygzws4
|
| | |\
| | |
| | | |
New take on SI-6534, value classes.
|
| | | |
| | |
| | |
| | |
| | |
| | | |
Only exclude hashCode and equals from being overridden in
value classes, not other synthetics which may turn up such
as case class methods.
|
| | | |
| | |
| | |
| | |
| | | |
Don't prohibit equals and hashCode in universal traits;
instead, always override them in value classes.
|
| | |\ \
| | | |
| | | | |
Fixes SI-6500 by making erasure more regular.
|
| | | |/
| | |
| | |
| | |
| | |
| | | |
With the introduction of value classes, erasure uses specialErasure where a value class C with underlying type T is unboxed to an ErasedValueType. ErasedValue types are eliminated on phase later, in post-erasure. This was done everywhere, except in the parameter types of bridge methods. That was a mistale, because that way bridge methods could not do the boxing/unboxing logic triggered by ErasedValueTypes.
Note: there is one remaining use of erasure (not specialErasure) in Erasure.scala. I put in a comment why that is OK.
|
| | |\ \
| | | |
| | | | |
SI-6556 no assert for surprising ctor result type
|
| | | |/
| | |
| | |
| | |
| | |
| | |
| | | |
Previous fix to value classes uncovered some questionable cases in the backend where result types of constructor signatures are surprising. It's not a big deal because these types will be ignored afterwards anyway. But
the method uncovered some questionable situations which we should follow up on. However, breaking 2.9 code because of this is way too harsh. That's why the asserts were converted to warnings.
review by @paulp, @adriaanm
|
| | | |
| | |
| | |
| | |
| | |
| | |
| | | |
* Removes actors-migration hooks from partest
* Removes actors-migration code
* removes actors-migration tests
* removes actors-migration distribution packaging.
|
| | |\ \ |
|
| | | |\ \
| | | | |
| | | | | |
SI-6581 fixed by inlining `Actor.self`.
|
| | | | |/
| | | |
| | | |
| | | |
| | | |
| | | | |
This avoids the necessary type cast that was preventing leakage of internal migration classes.
Review by @phaller
|
| | | |\ \
| | | | |
| | | | | |
Fixes SI-5031 for separate compilation scenario.
|
| | | | |/
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
When you have a conflicting member in package object and normal package that share the same namespace we remove the latter ClassSymbol from the scope. Now, this has an unpleasant consequence that companionClass/companionModule/companionSymbol no longer work correctly as they rely on finding the correspondent symbol using decls of the owner.
This fixes the problem of SI-5031 for separate compilation. Why the above change matters for finding foo.bar.Foo? Because when parsing the class we needed information about the static module (and we have the correct module symbol when completing the info). It's just that 043ce6d0565c9d5d960 relied on no longer valid assumptions. So we were getting NoSymbol and sym.exist was failing.
Obviously a more complete solution would be better if we didn't rely on the scope but that's too big to change for 2.10.0.
|
| | | |/
| |/|
| | |
| | |
| | |
| | |
| | | |
* Removes actors-migration hooks from partest
* Removes actors-migration code
* removes actors-migration tests
* removes actors-migration distribution packaging.
|
| | | |
| | |
| | |
| | |
| | |
| | |
| | | |
It seems to me like every call to scope.lookup in the
compiler is a latent bug. If a symbol is overloaded, you
get one at random. (See the FIXME comment in f5c336d5660
for more on this.)
|
| |\ \ \
| | | |
| | | | |
Merge 2.10.0-wip into 2.10.x.
|
| | |\ \ \
| | | |/
| | |/|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
# By Jason Zaugg (5) and others
# Via Josh Suereth (5) and others
* origin/2.10.0-wip:
Use Typed rather than .setType
Wider use and a new variant of typedPos.
SI-6575 Plug inference leak of AbstractPartialFun
Disabled generation of _1, _2, etc. methods.
SI-6526 Additional test case.
Fix SI-6552, regression with self types.
avoid single-art assert where harmful in duration-tck
Fix for SI-6537, inaccurate unchecked warning.
SI-6526 Tail call elimination should descend deeper.
Changes Tree and Type members from vals to defs.
Fixes SI-6170: issue with dragging scaladoc splitter over central iframe
|
| | | |/
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Usually this isn't needed, as partial functions can only be
defined with an expected type. But if that expected type is
and inherited method return type, the actual type of the partial
function literal is used, and the implementation detail of
`AbstractPartialFunction[A, B] with Serializable` leaks out.
After this change, the inferred types match those from
Scala 2.9.2.
ticket/6575 ~/code/scala scalac29 -Xprint:typer test/files/pos/t6575a.scala | grep def > 29.txt
ticket/6575 ~/code/scala squalac -Xprint:typer test/files/pos/t6575a.scala | grep def > 210.txt
ticket/6575 ~/code/scala diff -u 29.txt 210.txt
--- 29.txt 2012-10-28 13:51:07.000000000 +0100
+++ 210.txt 2012-10-28 13:51:20.000000000 +0100
@@ -1,7 +1,16 @@
def foo: PartialFunction[Int,Int]
def /*Y*/$init$(): Unit = {
- absoverride def foo: PartialFunction[Int,Int] = ((x0$1: Int) => x0$1 match {
+ absoverride def foo: PartialFunction[Int,Int] = {
+ def <init>(): anonymous class $anonfun = {
+ final override def applyOrElse[A1 >: Nothing <: Int, B1 >: Int <: Any](x$1: A1, default: A1 => B1): B1 = (x$1: A1 @unchecked) match {
+ final def isDefinedAt(x$1: Int): Boolean = (x$1: Int @unchecked) match {
def /*Z*/$init$(): Unit = {
- absoverride def foo: PartialFunction[Int,Int] = ((x0$2: Int) => x0$2 match {
+ absoverride def foo: PartialFunction[Int,Int] = {
+ def <init>(): anonymous class $anonfun = {
+ final override def applyOrElse[A1 >: Nothing <: Int, B1 >: Int <: Any](x$1: A1, default: A1 => B1): B1 = (x$1: A1 @unchecked) match {
+ final def isDefinedAt(x$1: Int): Boolean = (x$1: Int @unchecked) match {
def /*Comb*/$init$(): Unit = {
- absoverride def foo: PartialFunction[Int,Int] = ((x0$3: Int) => x0$3 match {
+ absoverride def foo: PartialFunction[Int,Int] = {
+ def <init>(): anonymous class $anonfun = {
+ final override def applyOrElse[A1 >: Nothing <: Int, B1 >: Int <: Any](x$1: A1, default: A1 => B1): B1 = (x$1: A1 @unchecked) match {
+ final def isDefinedAt(x$1: Int): Boolean = (x$1: Int @unchecked) match {
|
| | | |\
| | | |
| | | | |
Disabled generation of _1, _2, etc. methods.
|
| | | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
This was part of the introduction of ProductN, which had
to go back into pandora's box because of issues with cycles
during typing. These should have been reverted along
with it.
|
| | | |\ \
| | | | |
| | | | | |
SI-6526 Tail call elimination should descend deeper.
|
| | | | | | |
|
| | | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
It wasn't traversing into Select nodes nor into the receiver of
a tail call.
|
| | | | |/
| | |/|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
In 6eb55d4b7a we put in a remedy for an old issue SI-4560 which
had accumulated a number of sketchy partial remedies which carried
no tests to illustrate their necessity. Looks like at least one of
those was doing something useful. Here's to reversion-reversion.
This reverts commit c8bdf199, which itself reverted cb4fd6582.
|
| | | |\ \
| | | | |
| | | | | |
Fix for SI-6537, inaccurate unchecked warning.
|
| | | | |/
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
I found a more direct expression of the unchecked logic,
which should be much easier for others to verify. But the
bug being fixed here is that the unchecked checking happens
too early, and the sealed children of a symbol are not yet
visible if it is being simultaneously compiled.
|
| | | |/ |
|
| |\ \ \
| | | |
| | | | |
SI-6488: Fix for race with open I/O fds
|
| | |/ / |
|
| |/ /
| |
| |
| |
| | |
As reported by Curtis Stanford, with indication of what to fix. standardInterpolator was not correctly
calling the passed in process function, so raw strings were not really raw.
|
| |\ \
| | |
| | | |
Scaladoc knows the package structure of the libraries,
|
| | | |
| | |
| | |
| | | |
so don't include them in external documentation setting.
|
| | | | |
|
| | |/
|/|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Until now lazy accessors were handled somehow special because their symbol was created in typers but the corresponding tree was only added in Refchecks. This irregularity caused serious problems for value classes. Also it now looks just better when lazy value is treated in a similar way as other fields.
I needed to adapt reifier so that it handles the new implementation correctly. Previously it had to recreate lazy val only by removing defdef and renaming. Now we basically need to recreate lazy val from scratch.
There is one minor change to cps plugin but that is still fine because lazy vals were never really part of the transformation.
Some range positions needed to be fixed manually. We could do it at the creation time but that would require a lot more "if (symbol.isLazy)" conditions for MethodSyntheis and Symbol/Tree creation and would just unnecessary complicate api. If someone has a better idea, please speak up. Range positions changes were necessary because previously accessors were created at refchecks and they weren't checked by validator (even though they were wrong).
This commit removes lazy val implementation restriction introduced for 2.10.0.
(cherry-picked from 981424b)
|
