| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |\
| |
| | |
Small documentation fixes & small fix to Scaladoc @see formatting
|
| | | |
|
| | | |
|
| | | |
|
| | | |
|
| | | |
|
| | | |
|
| | |
| |
| |
| | |
Culling accumulated unnecessary code.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
maxId is not really the "maximum id", but rather one past the max:
scala> object Foo extends Enumeration {
| val x = Value
| val y = Value
| }
defined module Foo
scala> Foo.maxId
res0: Int = 2
scala> Foo(2)
java.util.NoSuchElementException: key not found: 2
at scala.collection.MapLike$class.default(MapLike.scala:225)
...
scala> Foo(1)
res2: Foo.Value = y
|
| |/ |
|
| | |
|
| |
|
|
| |
And undeprecated Positional.
|
| |
|
|
|
| |
Fixed a comment and simplified a mysterious "if (x) y else y" method
thanks to the helpful jibe received in comments.
|
| |\ \ \ \ \
| | | | | |
| | | | | |
| | | | | | |
'refs/pull/549/head'; commit 'refs/pull/550/head'; commit 'refs/pull/551/head' into develop
|
| | | | | | | |
|
| | | | | |/ |
|
| | | | | |
| | | | |
| | | | |
| | | | | |
make TreeInfo recognize constructor calls after named arguments transformation.
|
| | | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
SubTypePair's equals method calls =:= on the involved types,
which mutates the TypeVars contained in them
this is undesirable since we're simply checking whether a subtype test is pending
in addition to making subtyping "more correct" for type vars,
it should avoid the stackoverflow that's been plaguing us
(https://groups.google.com/d/topic/scala-internals/2gHzNjtB4xA/discussion)
SubTypePair's equals method method is only called when subtype checking
hits a recursion threshold (subsametypeRecursions >= LogPendingSubTypesThreshold)
|
| | |/ / /
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Previously, specialized lazy vals would not work at all when used in
specialized classes, and would just return an uninitialized value.
After this patch, they work in the same way as other specialized fields
do (i.e. a new specialized field is created, and the specialized class
uses that instead of the base class' field).
Note that there are still known problems with specialized lazy vals
(for instance SI-4717) but it seemed to me that this was better than
nothing.
|
| |/ / / |
|
| | | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
- Match @param/@tparam names to the actual parameter name
- Use @tparam for type parameters
- Whitespace is required between `*` and `@`
- Fix incorrect references to @define macros.
- Use of monospace `` and {{{}}} (much more needed)
- Remove `@param p1 ...` stubs, which appear in the generated docss.
- But, retainsed `@param p1` stubs, assuming they will be filtered from
the generated docs by SI-5795.
- Avoid use of the shorthand `@param doc for the solitary param`
(which works, but isn't recognized by the code inspection in IntelliJ
I used to sweep through the problems)
The remaining warnings from `ant docs` seem spurious, I suspect they are
an unintended consequence of documenting extension methods.
[scaladoc] /Users/jason/code/scala/src/library/scala/collection/TraversableOnce.scala:181: warning: Variable coll undefined in comment for method reduceOption in class Tuple2Zipped
[scaladoc] def reduceOption[A1 >: A](op: (A1, A1) => A1): Option[A1] = reduceLeftOption(op)
[scaladoc] ^
|
| | | |
| | |
| | |
| | |
| | | |
Relative symlinks which start with .. are then run with an
absolute path. It's robustness city.
|
| | | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
All this stuff. I believe most of this is uncontroversial.
scala/util/Marshal.scala
scala/util/automata/BaseBerrySethi.scala
scala/util/automata/DetWordAutom.scala
scala/util/automata/Inclusion.scala
scala/util/automata/NondetWordAutom.scala
scala/util/automata/SubsetConstruction.scala
scala/util/automata/WordBerrySethi.scala
scala/util/grammar/HedgeRHS.scala
scala/util/grammar/TreeRHS.scala
scala/util/parsing/ast/AbstractSyntax.scala
scala/util/parsing/ast/Binders.scala
scala/util/parsing/combinator/testing/RegexTest.scala
scala/util/parsing/combinator/testing/Tester.scala
scala/util/parsing/input/Positional.scala
scala/util/regexp/Base.scala
scala/util/regexp/PointedHedgeExp.scala
scala/util/regexp/SyntaxError.scala
scala/util/regexp/WordExp.scala
I'd have deprecated much of it long ago if the compiler didn't
still depend on it due to xml/dtd/something. And it still does,
but it's time to deprecate them anyway.
|
| |\| | |
|
| | | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Don't print trees under -Xprint:all if they're identical to the
tree printed at the previous phase. It only works for a single
compilation unit but that is a huge step forward for us debuggers.
For instance this file:
trait Foo { def f = 5 }
used to produce 332 lines of output and now produces 92, with
zero loss of information. It ends with:
[[syntax trees at end of cleanup]] // a.scala: tree is unchanged since mixin
[[syntax trees at end of icode]] // a.scala: tree is unchanged since mixin
[[syntax trees at end of inliner]] // a.scala: tree is unchanged since mixin
[[syntax trees at end of inlineExceptionHandlers]] // a.scala: tree is unchanged since mixin
[[syntax trees at end of closelim]] // a.scala: tree is unchanged since mixin
[[syntax trees at end of dce]] // a.scala: tree is unchanged since mixin
[[syntax trees at end of jvm]] // a.scala: tree is unchanged since mixin
|
| | | | | |
| | \ \ | |
| | \ \ | |
| | \ \ | |
| | \ \ | |
| | \ \ | |
| | \ \ | |
| | \ \ | |
| |\ \ \ \ \ \ \
| |_|_|_|_|/ /
|/| | | | | |
| | | | | | | |
'scalamacros/topic/macrocherrypick', 'lrytz/t5626', 'lrytz/t5009', 'retronym/ticket/5029', 'retronym/ticket/4025', 'retronym/topic/quieter-nsdhnao' and 'retronym/ticket/1133' into develop
|
| | | |_|_|/ /
| |/| | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
This is done for all crashers now. Furthermore, that error report
makes it pretty clear that it's a compiler bug, not user error, so
"this is a bug" seem like needless words.
Resisted the nostalgic temptation to remove the "an".
|
| | | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
Example: Given
case class C(a: Int)(b: Int)
if you call (new C(1)(2)).copy(a = 10)), you get a function (f: Int => C) such
that (f.apply(20)) yields a (new C(10)(20)).
|
| | | | |/ /
| | | | |
| | | | |
| | | | | |
By not replacing 'CaseClass.apply()' factor by 'new CaseClass()' when the class type 'CaseClass' is not accessible.
|
| | | | | | |
|
| | | | | | | |
| | \ \ \ \ | |
| | \ \ \ \ | |
| | \ \ \ \ | |
| | \ \ \ \ | |
| | \ \ \ \ | |
| |\ \ \ \| | | |
| | | | | | | |
| | | | | | | |
| | | | | | | | |
'refs/pull/532/head'; commit 'refs/pull/533/head'; commit 'refs/pull/534/head' into develop
|
| | | | | | |/ /
| | | | |/| |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
Fixed up one of the mismatches between how java source is modeled
and how java bytecode is modeled. We should get the rest of them
too. Closes SI-2764.
|
| | | | |/ / / |
|
| | | |/ / /
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
Mirror classes (with static forwarders) for top-level
objects would get unnecessary InnerClasses entries
in some cases. See the following comment I introduced
to the place I fixed:
We need to make sure that we do not confuse inner classes
of the class we mirror with inner classes of the class
we are mirroring. These two sets can be different as
seen in this case:
```scala
class A {
class B
def b: B = new B
}
object C extends A
```
Here mirror class of C has a static forwarder for
(inherited) method `b` therefore it refers to class `B`
and needs InnerClasses entry. However, the real class
for `C` (named `C$`) is empty and does not refer to `B`
thus does not need InnerClasses entry it.
NOTE: This logic has been refactored in GenASM and
everything is implemented in a much cleaner way by
having two separate buffers.
|
| | | | | | | |
| | | | \ \ | |
| | | |\ \ \ \
| | | | | | |
| | | | | | |
| | | | | | | |
'refs/pull/529/head' into develop
|
| | | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
More details as code comment and in the bug database.
|
| | | | | | | | |
|
| | | |/ / / /
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
Prevents the wheels falling off during later compiler phases,
or, worse, during bytecode verification.
Closes SI-4098.
|
| | | |\ \ \ \ |
|
| | | | | | | | |
|
| | | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
They're not traits.
|
| | | | | | | | |
|
| | | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
If there are no primitives, use the regular boxy one.
|
| | | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
If you can't get your hands on something which crashes
scalac - I know, I know - you can try this ready-made crasher.
% cat test/pending/pos/t4717.scala
trait Bounds[@specialized A] {
// okay without `>: A`
def x[B >: A]: Unit = new Bounds[B] {
lazy val it = ??? // def or val okay
it
}
}
% scalac -d /tmp test/pending/pos/t4717.scala
error:
while compiling: test/pending/pos/t4717.scala
during phase: specialize
library version: version 2.10.0-20120510-134429-ce1d68ed19
compiler version: version 2.10.0-20120510-152646-ba4dfd1e63
reconstructed args: -d /tmp
last tree to typer: Select(This(trait Bounds$mcZ$sp), x$mcZ$sp)
symbol: method x$mcZ$sp in trait Bounds$mcZ$sp (flags: override <method> <specialized>)
symbol definition: override def x$mcZ$sp[B >: Boolean](): Unit
tpe: [B >: Boolean]()Unit
symbol owners: method x$mcZ$sp -> trait Bounds$mcZ$sp -> package <empty>
context owners: value it -> anonymous class $anon -> method x$mcZ$sp -> trait Bounds$mcZ$sp -> package <empty>
== Enclosing template or block ==
Block(
Assign(
$anon.this."it "
Apply( // def ???(): Nothing in object Predef, tree.tpe=Nothing
scala.this."Predef"."$qmark$qmark$qmark" // def ???(): Nothing in object Predef, tree.tpe=()Nothing
Nil
)
)
$anon.this."it " // lazy private[this] var it: Nothing, tree.tpe=Nothing
)
== Expanded type of tree ==
PolyType(
typeParams = List(TypeParam(B >: Boolean))
resultType = NullaryMethodType(
resultType = TypeRef(TypeSymbol(final class Unit extends AnyVal))
)
)
// And then the usual stack trace
|
| | | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
Stuff written during the era of "one symbol table" is
surprised to find two.
|
| | | | | | | | |
|
| | |/ / / / /
|/| | | | | |
|
| |\ \ \ \ \ \
| |_|/ / / /
|/| | | | | |
|
| | | | | | | |
|
| |\ \ \ \ \ \ |
|
