| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
| |
If we can't type check the `Foo` in `case Foo(a, b) => (a, b)`,
we should enter error symbols for `a` and `b` to avoid further
errors being reported in the case body.
|
|\
| |
| | |
Describe type parameter interpolation in @implicitNotFound documentation
|
| |
| |
| |
| |
| |
| | |
Using this feature is necessary for helpful error messages, so it should be documented.
Thank you to @adriaanm for recommending the this description.
|
|\ \
| | |
| | | |
SI-7886 unsoundness in pattern matcher.
|
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Introduces -Xstrict-inference to deal with the significant
gap between soundness and what presently compiles. I'm hopeful
that it's TOO strict, because it finds e.g. 75 errors compiling
immutable/IntMap.scala, but it might be that bad.
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
I tracked down what was behind the issue described here:
// TODO: fix the illegal type bound in pos/t602 -- type inference
// messes up before we get here:
/*override def equals(x$1: Any): Boolean = ...
// Span[Any] --> Any is not a legal type argument for Span!
val o5: Option[com.mosol.sl.Span[Any]] =
*/
...which led straight to the unsoundness seen in neg/t7886.
It is dangerous to have an expected type of "Any" because
the type system will blithely ignore kind errors, since "Any"
can absorb anything. The consequence in this instance was
that inferring the case constructor for a type like
Foo[T <: Bound]
if done with expected type Any, this would come up with Foo[Any].
I altered it to use expected type Foo[T], which lets the dummy
type parameter survive to carry the bound forward and restores
sense to the inference. The before/after output for -Xprint:patmat
on pos/t602.scala is:
15c15
< if (x1.isInstanceOf[com.mosol.sl.Span[Any]])
---
> if (x1.isInstanceOf[com.mosol.sl.Span[K]])
17c17
< <synthetic> val x2: com.mosol.sl.Span[Any] = \
(x1.asInstanceOf[com.mosol.sl.Span[Any]]: com.mosol.sl.Span[Any]);
---
> <synthetic> val x2: com.mosol.sl.Span[K] = \
(x1.asInstanceOf[com.mosol.sl.Span[K]]: com.mosol.sl.Span[K]);
19,20c19,20
< val low$0: Option[Any] = x2.low;
< val high$0: Option[Any] = x2.high;
---
> val low$0: Option[K] = x2.low;
> val high$0: Option[K] = x2.high;
A file in the library depended (needlessly) on the unsoundness.
It was easy to fix but reminds us this may affect existing code.
|
| | | |
|
|\ \ \
| | | |
| | | | |
Rework cff8b569 to heal the windows build.
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
- change newTermName to fix negative length names
rather than reject them
- restore the old logic in unspecializedName for names that
result from AnyRef specialized type parameters.
Why does fix the windows build? I remain none the wiser.
|
|\ \ \ \
| |_|_|/
|/| | | |
Update of Eclipse project files
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
- Update of project dependencies
- Deletion of scala-xml project
- Deletion of scala-parser-combinators project
- Update of .gitignore to match nested .cache files
|
|\ \ \ \
| | | | |
| | | | | |
SI-7859 Value classes may wrap a non-public member
|
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
We allow value class constructors to be non-public, so to be regular,
we should also allow the same for the param accessor.
This commit uses the 'makeNotPrivate' machinery to ensure that
the backend can generate the requisite unboxing calls.
This commit:
- refactors the code that enforced the restrictions, improving
a few error messages and positions. The remaining restrictions
needed some rewording in light of this change.
- allows value classes to have non-public, val parameters.
private[this] / protected[this] are still disallowed as value
classes don't have a concept of `this`, and because trying to
accomdate then would complicate the implementation.
This means that `class C(x: Int) extends AnyVal` is not allowed,
the user still must write `private val x: Int` or `val x: Int`.
- Outlaw `class C()()(val x: Int) extends AnyVal` to curtail any
bugs that might lurk in such a formulation.
The tests:
- Show that the privacy is respected in the typer phase, under
joint and separate compilation. We don't want a repeat performance
of SI-6601.
- Show that code that needs compiler-generated unboxes works under
both compilation scenarios
- Checks that the remaining restrictions are enforced and well
communicated.
|
|\ \ \ \ \
| | | | | |
| | | | | | |
assorted fixes for bundles
|
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
One of the previous commits relaxed the top-level restriction for bundles,
turning it into a requirement of staticness (i.e. bundles nested in static
objects are also okay now).
This means that we can now define bundles in repl. Almost.
There's still a little problem remaining that arises from the fact that
when compiling a line of input, repl doesn't automatically import all
previously defined symbols, but rather uses an heuristic to scan the
input and guess what symbols need to be imported.
Unfortunately for bundles, this heuristic fails, because when scanning
a macro definition that looks like `def foo = macro Macros.foo`, it thinks
that it's only necessary to import a term symbol called Macros (a vanilla
way of defining macro impls), but not a type symbol called Macros (a new
way of writing macro impls in bundles).
This commit fixes the problem by making the repl look for both term and
type symbols corresponding to the identifiers used in macro definitions.
|
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
Bundles are emitted in compilation units that wrap virtual source files.
Previously we had those virtual files named randomly to ensure freshness,
but that led to infinite compilation loops in SBT (see the commit message
for more details and a link to a scala-user discussion).
Now the names are generated deterministically from full names of bundles,
which fixes the problems with SBT.
|
| | |_|/ /
| |/| | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
Previously it was enough to just extend scala.reflect.macros.Macro, which
created some loopholes, but now scalac enforces that bundles:
1) Are static (not necessarily top-level, but just static)
2) Are traits (objects shouldn't be bundles anyway, and classes bring
complications with their ctors which require special treatment in
generated classes, so why support them if they don't bring anything
new to the table?)
3) Are monomorphic (again, this brings unnecessary complications wrt
auxiliary code generation, so I don't see merit in supporting
polymorphic bundles, whatever that a polymorphic bundle could mean)
4) Don't provide concrete implementation for Macro.c (if they do then
what is the point?)
|
|\ \ \ \ \
| |/ / / /
|/| | | |
| | | | |
| | | | | |
sjrd/topic/remove-classpath-logic-dependent-on-inline
Don't avoid to load trait impl .class without inliner.
|
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
The parent commit, a3f71badf67bbaac1a4ba16f68211ea6e31aa473,
removed some logic preventing $class.class files to be
loaded. It did so only when the inliner was off.
Should this cause any issue, this option provides a means to
restore the old behavior by *never* loading $class.class files.
So, using -inline -Yno-load-impl-class will not load
$class.class files either (where previously -inline would
load them). The two old behaviors being available by *either*
using -inline *or* -Yno-load-impl-class (not both).
|
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
When the inliner is on, it is necessary to load class
files containing the implementation of traits, i.e.,
those ending in $class.class. This is needed to be
able to inline these methods. However it is useless
if the inliner is disabled.
The previous logic avoided to put these files on the
classpath when the inliner was off. However, this
complicates things, and it makes the classpath
mechanism dependent on something totally unrelated.
On this basis, this commit removes that logic, and
instead trait impl .class files are always kept on
the classpath.
|
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
Most of this was revealed via -Xlint with a flag which assumes
closed world. I can't see how to check the assumes-closed-world
code in without it being an ordeal. I'll leave it in a branch in
case anyone wants to finish the long slog to the merge.
|
| |_|_|/
|/| | |
| | | |
| | | |
| | | | |
This is the change which broke the compiler until I made the
changes found in 693ecffbaf.
|
|\ \ \ \
| | | | |
| | | | | |
Updating Position call sites.
|
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
Calling position factories rather than instantiating these
particular classes. Not calling deprecated methods. Added a few
position combinator methods.
|
|\ \ \ \ \
| | | | | |
| | | | | | |
transformers no longer ignore UnApply.fun
|
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
Centralizes wrappers in a single location and makes it impossible to use
the compiler without going through that location.
Also fixes the option string tokenization bug that was there for ages.
This time I figured out that there's already an implementation of the
tokenizer in our compiler, so I just used it :)
|
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
Second time's the charm. I remember trying to do exactly the same somewhen
around 2.10.0-M4, but then some continuations tests were failing.
Luckily, today everything went smoothly.
Please note that this fix changes the way that SI-5465 manifests itself.
Previously it produced type errors, now it simply crashes the compiler.
Therefore I had to attach the try/catch FatalError clause to invocations
of toolbox methods, so that compiler crashes get caught and translated to
ToolBoxErrors.
Also fixes SI-7871, and that clears the way for implementing quasiquotes
with conventional macros rather than relying on a special case in typer.
|
| |/ / / /
|/| | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
I had covered a few more cases working on this recently.
The warnings in several more cases involving polymorphism,
currying, and selects vs. idents receive more refined
handling.
|
|\ \ \ \ \
| | | | | |
| | | | | | |
SI-7629 Deprecate view bounds
|
| |/ / / /
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
This introduces a warning(/error with -Xfuture) with a general
migration advice. The IDE can use the warning to offer a quick fix
with the specific refactoring necessary.
|
|\ \ \ \ \
| | | | | |
| | | | | | |
Only look for unapplies in term trees
|
| |/ / / /
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
Since Scala 2.10.2, the enclosed test case has crashed
in the backend. Before, we correctly rejected this pattern match.
My bisection landed at a merge commit f16f4ab157, although both
parents were good. So I don't quite trust that.
I do think the regression stems from the changes to allow:
case rx"AB(.+)" =>
Examples of this are in run/t7715.scala.
This commit limits the search for extractors to cases where the
function within the Apply is a term tree.
|
|\ \ \ \ \
| | | | | |
| | | | | | |
SI-7848 Xlint no warn on $sym with params
|
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
This idea brought to you by retronym.
Also improve implicitNotFound detection at typer;
and avoid checking the standard interpolation
expression for cases like s"some $$x".
Some minor refactorings of implicitNotFound strings.
The intersobralator allows extra spaces, i.e., trims.
|
|\ \ \ \ \ \
| | | | | | |
| | | | | | | |
SI-3971 error message carat mispoints at curried methods.
|
| | |_|_|/ /
| |/| | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
Point at the beginning of the first argument list when
reporting an error, as this is most easily associated with
the application taking place (which may involve multiple
applies in succession.)
Thanks to retronym for figuring out why issuing a better
error message broke the compiler on non-erroneous compile runs.
The changes to "treesInResult" are the consequence.
|
|\ \ \ \ \ \
| | | | | | |
| | | | | | | |
SI-6120 multiple warnings at same position.
|
| |/ / / / /
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
An error suppresses all further warnings at the same position,
but multiple warnings can be heard.
|
|\ \ \ \ \ \
| | | | | | |
| | | | | | | |
SI-6762 rename emptyValDef to emptySelfType.
|
| |/ / / / /
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
Looks like emptyValDef.isEmpty was already changed to return
false, so now all that's left is a name which means something.
|
|\ \ \ \ \ \
| | | | | | |
| | | | | | | |
Fix up DEFAULTPARAM semantics.
|
| |/ / / / /
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
I foolishly believed the deprecation message on "hasDefaultFlag"
which suggested I use "hasDefault" instead. After lots of head
scratching, I hardened the semantics so it's like this:
- A method parameter with a default value is PARAM | DEFAULTPARAM
- A default getter for such a parameter is METHOD | DEFAULTPARAM
- And "hasDefault" is has(DEFAULTPARAM) && has(PARAM | METHOD)
Why all the bonus logic, why not just hasFlag(DEFAULTPARAM)? For
some reason we have a handful of overloaded flags spanning uses
which someone apparently thinks can never intersect but I have
not been so lucky overall. So since DEFAULTPARAM is overloaded with
TRAIT, unless we think it's fine that default getters and method
parameters with defaults will pose as traits all the time, there
has to be an anchor bit alongside it.
|
|\ \ \ \ \ \
| | | | | | |
| | | | | | | |
Small refactorings giving hooks for Scala.js
|
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
Other methods taking the same kind of decisions were already
in ClassPathContext, e.g., isValidName() or, in some sense,
even toBinaryName().
This makes ClassPath itself be completely agnostic of how
particular kinds of files or directories are named.
It also allows to override this logic at the context level.
Without it, overriding this logic required a fair amount of
code duplication from ClassPath.
|
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
This was previously an overridable method of Platform.
The loader creation was moved in
afbee09c8e0e7b1a4da1f8517c723dad9f1adb6f directly in
SymbolLoaders, but inside a method doing more logic,
namely initializeFromClassPath().
This commit simply moves the actual creation of the
class loader (`new ClassfileLoader(bin)`) into its
own method `newClassLoader`, but in SymbolLoaders.
This allows to override only that method in subclasses
of SymbolLoaders.
|
|\ \ \ \ \ \ \
| |_|/ / / / /
|/| | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
Conflicts:
bincompat-backward.whitelist.conf
bincompat-forward.whitelist.conf
build.xml
src/compiler/scala/tools/nsc/typechecker/RefChecks.scala
src/library/scala/concurrent/Future.scala
src/reflect/scala/reflect/internal/Types.scala
|
| |\ \ \ \ \ \ |
|
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | | |
Callbacks internal to the implementation of Futures should be
executed with the `InternalCallbackExecutor`, rather than the
user supplied `Executor`.
In a refactoring da54f34a6, `recoverWith` and `flatMap` no longer
played by these rules. This was noticed by a persnickety test in
Play.
Before this patch, the enclosed test outputs:
% scala-hash v2.10.3-RC2 test/files/run/future-flatmap-exec-count.scala
mapping
execute()
flatmapping
execute()
execute()
recovering
execute()
execute()
|
| |\ \ \ \ \ \ \
| | |/ / / / / /
| |/| | | | | | |
SI-7815 Dealias before deeming method type as dependent
|
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | |
| | | | | | | | |
To enable eta-expansion of method types seen from a prefix that
renders the result type as independent from the parameter symbols.
The enclosed test shows that we dealias types before checking
dependence, and that we do this deeply (e.g. type arguments are
also dealised.)
An existing test, neg/error_dependentMethodTpeConversionToFunction,
confirms that bona-fide dependent methods are still prohibited from
eta expansion.
|