| Commit message (Collapse) | Author | Age | Files | Lines |
|\
| |
| | |
SI-6863 Fix verify error in captured var inited from block w try/catch
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
If a captured var was inited from a try/catch we did something
reasonable. But if the var was inited from a more complicated expression
(if/else, a block, match/case, etc) that ended with
a try/catch then we didn't and we were generating faulty byte code.
This fix patches LambdaLift to add the missing cases.
For known simple expressions, the translation is just new *Ref(expr).
For try/catch, if/else, match/case, and blocks this recursively
walks down the internal result expressions to translate them. E.g.
if(cond) trueExpr else falseExpr becomes if(cone) translate(trueExpr)
else translate(falseExpr)
For unknown expression types, the translation is {val temp = expr; new
*Ref(expr) }
|
|\ \
| | |
| | | |
SI-6932 StackOverflowError in chained Future.flatMap calls
|
| | |
| | |
| | |
| | | |
internal execution context of Future
|
|\ \ \
| | | |
| | | | |
SI-6443 Widen dependent param types in uncurry
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
These were already working, due to a serendipitous
ordering of transformations. The tests will keep
it this way.
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Bridge building operates on unusual method signatures:
after uncurry, so parameter lists are collapsed; but before
erasure, so dependently typed parameters are still around.
Original:
def foo(a: T)(b: a.type, c: a.U): Unit
During computeBridges:
(a: T, b: a.type, c: a.U)Unit
This signature no longer appears to override the corresponding
one in a superclass, because the types of `b` and `c` are dependent
on method parameters.
The root of the problem is uncurry, which leaves the trees in
a poor state. This commit changes uncurry to remedy this.
An example illustrates it best:
// source
def foo(a: A)(b: a.type): b.type = b
// post uncurry before this patch.
// not well typed code!
def foo(a: A, b: a.type): a.type = {
// post uncurry after this patch
def foo(a: A, b: A): A = {
val b$1 = b.asInstanceOf[a.type]
b$1
}
|
| |_|/
|/| |
| | |
| | |
| | |
| | | |
The build is currently broken:
https://scala-webapps.epfl.ch/jenkins/view/2.10.x/job/scala-nightly-main-2.10.x/
|
|\ \ \
| | | |
| | | | |
SI-7011 Fix finding constructor type in captured var definitions
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
If a captured var was initialized with an empty tree then finding
the type of the empty tree was being handled improperly. The fix is
to look for primary constructors on the tree's type symbol rather than
the tree's symbol.
A test is included. In order to make the problem more testable the debug
logging of the issue is changed to a debug warn.
|
|\ \ \ \
| | | | |
| | | | | |
SI-6231 Report unsupported free var capture by a trait.
|
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
If a class nested in a trait captures a free variable from
the enclosing scope of the trait, the transformation to
add that variable to the `init` method of the trait
implementation class happens *after* the abstract trait
interface has been extracted. This would lead to a crash
when trying to find the corresponding interface method.
This commit detects this situation and reports an
implementation restriction. The enclosed test case
shows a workaround.
To lift this restriction, LambdaLifter should add the getters
and make sure they end up in the trait interface. Looks like
Martin tried this once:
// LambdaLift.scala
//
// Disabled attempt to to add getters to freeParams
// this does not work yet. Problem is that local symbols need local names
// and references to local symbols need to be transformed into
// method calls to setters.
// def paramGetter(param: Symbol): Tree = {
// val getter = param.newGetter setFlag TRANS_FLAG resetFlag PARAMACCESSOR // mark because we have to add them to interface
// sym.info.decls.enter(getter)
// val rhs = Select(gen.mkAttributedThis(sym), param) setType param.tpe
// DefDef(getter, rhs) setPos tree.pos setType NoType
// }
// val newDefs = if (sym.isTrait) freeParams ::: (ps map paramGetter) else freeParams
|
|\ \ \ \ \
| | | | | |
| | | | | | |
SI-6987 Fixes fsc compile server verbose output
|
| | |/ / /
| |/| | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
This commit includes a test of fsc's verbose output. In order for it
to work, CompileServer's main method had to be modified to remove a
sys exit 0 at the end. It was redundant and made testing a bit harder.
In order to prevent a race condition between server and client start
up, this commit also adds a server callback that decrements a
CountDownLatch that the main testing thread waits for.
Finally, the server had to be modified to use Console.withErr
and Console.withOut instead of mutating the global System.err
and System.out variables. Otherwise the test would be unreliable.
|
|\ \ \ \ \
| | | | | |
| | | | | | |
SI-6666 Restrict hidden `this` access in self/super calls.
|
| | |/ / /
| |/| | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
Detect when classes (user authored or compiler generated)
local to a self or super constructor argument would require
premature access to the in-construction instance.
The same restriction applies for classes and objects; for objects,
the premature access would result in a null via MODULE$ field.
A residual error has been lodged as SI-6997.
I'd like to remove calls to `Symbol#outerClass` (which relies on
the flaky flag INCONSTRUCTOR, see my comments in the JIRA issue
for more discussion) from `LambdaLift` and `ExplicitOuter`, and
instead use the stack of active self/super calls to know when to
skip an enclosing class. That will obviate that flag.
|
|\ \ \ \ \
| | | | | |
| | | | | | |
SI-6902 Check unreachability under @unchecked
|
| | |_|/ /
| |/| | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
Only exhaustiveness checking should be disabled if the
scrutinee of a match as annotated as `: @unchecked`.
This was the pre-2.10.x behaviour.
This also fixes a variation of the closed ticket,
SI-6011. The exhaustiveness check is needed to
safely fallback from emitting a table switch if
duplicate cases are detected.
|
|\ \ \ \ \
| | | | | |
| | | | | | |
Closes SI-6952: add correct error positions for Dynamic feature check.
|
| | | | | | |
|
|\ \ \ \ \ \
| | | | | | |
| | | | | | | |
SI-6969, mishandling of SoftReferences in method cache.
|
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
More interesting to test than it was to fix. The soft
reference is now dereferenced once, the locally stored
underlying value ascertained to be non-null, and the
remainder of the references to the value use the local var.
The enclosed test reliably NPEs without this patch.
|
|\ \ \ \ \ \ \
| |/ / / / / /
|/| | | | | | |
SI-6976 Fix value class separate compilation crasher.
|
| | |/ / / /
| |/| | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
We can't guarantee that the owner of the value class
is initialized, and if it isn't, the search for the
companion module will turn up bubkis.
This is a localized fix, but I'd be suprised if there
weren't other places that suffered from the same problem.
Wouldn't it be nicer to have something like:
// doesn't force info
sym.raw.info
sym.raw.companionModule
// forces info
sym.info
sym.companionModule
|
|\ \ \ \ \ \
| | | | | | |
| | | | | | | |
[backport] Fix SI-6637 (misoptimization in erasure)
|
| | |_|_|/ /
| |/| | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
commit f9ef5300ab561628e53c654df9000c75f488d74a
Author: Jan Niehusmann <jan@gondor.com>
Date: Fri Nov 9 15:05:58 2012 +0100
Fix SI-6637 (misoptimization in erasure)
Move the optimization one level deeper so the expression
being tested with isInstanceOf is always evaluated.
(cherry picked from commit b540aaee4ba30e2dd980456a44e8c6d732222df1)
|
|\ \ \ \ \ \
| | | | | | |
| | | | | | | |
[backport] Fix unsafe array opt. / opt. primitive Array(...)
|
| |/ / / / /
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
SI-6611, SI-6247 (partial fix)
The original commits on master were a bit circuitous, this
is squashed to a neat little package.
I had to add type arguments to the Array.apply calls in the
test case, they are inferred on master.
commit 41ff05dfdbcf032157b3509ace633f2e7a12295c
Author: Jason Zaugg <jzaugg@gmail.com>
Date: Sun Nov 4 14:44:59 2012 +0100
Refactor guards checking for a particular overload of Array.apply.
(cherry picked from commit 092345a24c22a821204fb358d33272ae8f7353be)
commit 1e5c942deccaf64f8d57bd8891b912381d7f220a
Author: Jason Zaugg <jzaugg@gmail.com>
Date: Sun Nov 4 14:17:25 2012 +0100
Expand optimization of Array(e1, ..., en) to primitive arrays.
(cherry picked from commit 8265175ecc42293997d59049f430396c77a2b891)
commit ab1bf77e39f2dfeacf3fc107ccb2907a1867f04c
Author: Jason Zaugg <jzaugg@gmail.com>
Date: Sat Nov 3 13:34:20 2012 +0100
SI-6611 Tighten up an unsafe array optimization
The net was cast too wide and was unsafely optimizing away array
copies.
(cherry picked from commit dad886659faca4fba2d4937c9bc6780591b02c27)
And also:
Optimize primitive Array(e1, ..., en)
Expands an existing optimization for reference arrays to
apply to primitives, as well.
Fixes one aspect of SI-6247.
(cherry picked from commit cac5a08611f9511ba4d94b99db630404efae190a)
Conflicts:
src/compiler/scala/tools/nsc/transform/CleanUp.scala
More principled tree copying.
Canonical > home-spun.
Conflicts:
src/compiler/scala/tools/nsc/transform/CleanUp.scala
|
|\ \ \ \ \ \
| | | | | | |
| | | | | | | |
[backport] SI-6567 Warning for Option(implicitView(foo))
|
| |/ / / / /
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
commit 284bd754fa5dfc8bc626b0c5ebe85d872dd044cb
Author: Jason Zaugg <jzaugg@gmail.com>
Date: Sat Nov 3 16:19:46 2012 +0100
SI-6567 Warning for Option(implicitView(foo))
I've seen the reported problem before in the wild. It seems
worthy of a special warning, so long as we advocate Option.apply
as an alternative to `if (x == null) Some(x) else None`.
It is behind -Xlint at the moment, an option that could do with
some promotion.
(cherry picked from commit 0bcb9e9169146e3f589c6c9f65cc4a5523b78120)
|
|\ \ \ \ \ \
| | | | | | |
| | | | | | | |
SI-6439 Avoid spurious REPL warnings about companionship
|
| | |_|_|/ /
| |/| | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
`val m` isn't a companion of `trait m`, check the pair of
eponymous symbols are a ((class|trait), object) pair before
emitting the warning.
In order to correctly check this one a type alias is involved,
`definedSymbols` must avoid normalizing through type aliases.
AFAICT this is an improvement to the other clients of that Map,
one such power mode progression is demonstrated at the end
of the test case.
|
|\ \ \ \ \ \
| | | | | | |
| | | | | | | |
SI-6994 Avoid spurious promiscuous catch warning
|
| |/ / / / /
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
It was being issued upon re-typechecking of a transformed
tree. Now we disable the warning post-typer.
|
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
It's all system admin, all the time, here at scala ranch.
|
| |/ / / /
|/| | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
We were pretty printing a function type with one by name arg as
=> A => B, but because => is right associative that's formally
equivalent to => (A => B) and that's entirely a different thing. This
commit changes the pretty printer in Typers.scala to check for a
byname argument on a function type and wrap it in parens. A REPL test
is included.
|
| |_|_|/
|/| | |
| | | |
| | | | |
Someone checked in a pair of .class files.
|
|\ \ \ \
| |/ / /
|/| | | |
SI-6942 more efficient CNF conversion in patmat analysis
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Avoid blowing the stack/the analysis budget by more eagerly translating
the propositions that model matches to CNF.
First building a large proposition that represents the match,
and then converting to CNF tends to blow the stack.
Luckily, it's easy to convert to CNF as we go.
The optimization relies on `CNF(P1 /\ ... /\ PN) == CNF(P1) ++ CNF(...) ++ CNF(PN)`:
Normalizing a conjunction of propositions
yields the same formula as
concatenating the normalized conjuncts.
CNF conversion is expensive for large propositions,
so we push it down into the conjunction and
then concatenate the resulting arrays of clauses (which is cheap).
(CNF converts a free-form proposition into an `Array[Set[Lit]]`, where:
- the Array's elements are /\'ed together;
- and the Set's elements are \/'ed;
- a Lit is a possibly negated variable.)
NOTE:
- removeVarEq may throw an AnalysisBudget.Exception
- also reworked the interface used to build formula,
so that we can more easily plug in SAT4J when the time comes
|
|\ \ \ \
| |/ / /
|/| | | |
SI-5568 Fixes verify error from getClass on refinement of value type
|
| | | |
| | | |
| | | |
| | | |
| | | | |
Based on code review here are a few comment cleanups and the removal of
some dead test code.
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
().asInstanceOf[AnyRef with Unit].getClass and
5.asInstanceOf[AnyRef with Int].getClass would cause a verify
error. Going the other way, i.e. [Unit with AnyRef] or [Int with AnyRef]
worked fine. This commit fixes it that both directions work out to
BoxedUnit or java.lang.Integer.
|
|\ \ \ \
| | | | |
| | | | | |
SI-6601 Publicise derived value contstructor after pickler
|
| | |/ /
| |/| |
| | | |
| | | |
| | | |
| | | |
| | | | |
Otherwise the access restrictions are not enforced under
separate compilation.
See also SI-6608.
|
|\ \ \ \
| | | | |
| | | | | |
SI-6923 Context now buffers warnings as well as errors
|
| |/ / /
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Code that was silently typed would not report warnings, even if it
returned a successful result.
This appeared in the following code which didn't show warnings even
with -Ywarn-adapted-args:
def foo(a: Any) = a; foo(1, 2)
While the following would show the expected warning:
def foo[A](a: Any) = a; foo(1, 2)
|
|\ \ \ \
| |_|_|/
|/| | | |
SI-6956 determine switchability by type, not tree
|
| | |/
| |/|
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Constant folding will set the type of a constant tree
to `ConstantType(Constant(folded))`, while the tree
itself can be many different things (in casu, an Ident).
We used to look at the tree directly when deciding whether
to emit a switch. Now we look at the tree's type. Voilà.
|
|\ \ \
| |/ /
|/| | |
SI-6479 Don't lift try exprs in label arguments.
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
The new pattern matcher uses label jumps to GOTO
the next case. Uncurry treated these like regular
method arguments, and performed the liftedTree()
transformation, which ensures that try expressions
are only used in a statement position. Even try
in statement position of a block used as such an argument
are subject to the same transform.
This transform stems from the JVM limitation,
that try/catch does not leave a value on the stack.
See b194446.
This commit changes Uncurry to avoid this transform
for arguments to label jumps. This avoids needlessly
indirect code, and enables tail call elimination in
more cases.
As an example, Scala 2.10.0 transforms the last
method of the enclosed test case to:
try {
case <synthetic> val x1: Int = 1;
case5(){
if (2.==(x1))
{
val x2: Int = x1;
matchEnd4({
{
def liftedTree2(): Unit = try {
throw new scala.runtime.NonLocalReturnControl[Unit](nonLocalReturnKey1, ())
} catch {
case (e @ (_: ClassNotFoundException)) => ()
};
liftedTree2()
};
TailrecAfterTryCatch.this.bad()
})
}
else
case6()
};
case6(){
matchEnd4(throw new MatchError(x1))
};
matchEnd4(x: Unit){
x
}
} catch {
case (ex @ (_: scala.runtime.NonLocalReturnControl[Unit @unchecked])) => if (ex.key().eq(nonLocalReturnKey1))
ex.value()
else
throw ex
}
After this patch:
@scala.annotation.tailrec final def bad(): Unit = {
case <synthetic> val x1: Int = 1;
case5(){
if (2.==(x1))
{
<synthetic> val x2: Int = x1;
matchEnd4({
try {
return ()
} catch {
case (e @ (_: ClassNotFoundException)) => ()
};
TailrecAfterTryCatch.this.bad()
})
}
else
case6()
};
case6(){
matchEnd4(throw new MatchError(x1))
};
matchEnd4(x: Unit){
x
}
}
|