| Commit message (Collapse) | Author | Age | Files | Lines |
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This commit modularizes TemplateTransformer by moving a few members
from TemplateTransformer into the newly added OmittablesHelper.
The members in question include
- a few methods:
# isParamCandidateForElision(Symbol)
# isOuterCandidateForElision(Symbol)
# mustbeKept(Symbol)
- a few vals:
# paramCandidatesForElision
# outerCandidatesForElision
# bodyOfOuterAccessor
- class UsagesDetector
That way, all trace of rewriting to elide param-accessor fields
has vanished from TemplateTransformer and is encapsulated in
OmittablesHelper.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This commit modularizes TemplateTransformer by moving a few methods
from TemplateTransformer into the newly added DelayedInitHelper.
The methods in question
- delayedInitCall()
- delayedInitClosure()
- delayedEndpointDef()
build trees that rewriteDelayedInit() puts together.
That way, all trace of rewriting related to delayed-init
have vanished from TemplateTransformer and are encapsulated in
DelayedInitHelper.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Before this commit, each activation of transformClassTemplate()
performed the rewriting that the constructors phase is well-known for.
Lots of auxialiary objects were created in the process,
with lifetime confined to said activation.
The same auxiliary objects (having same lifetimes as before)
also are in effect starting with this commit,
but now it's an instance of TemplateTransformer
that holds them together. In other words,
there's a one-to-one correspondence between:
- (what used to be) transformClassTemplate() activation
- TemplateTransformer initialization
After initialization, the result of TemplateTransformer
can be found in its `transformed` member val.
In fact, the refactoring to get here from the previous commit
basically involves taking the body of method transformClassTemplate()
as-is to become the template of TemplateTransformer.
The TemplateTransformer in question will allow modularizing
sub-transformations (e.g., DelayedInit) into dedicated traits
(see upcoming commits).
|
|
|
|
|
| |
Removing the old implementation of elision in constructors
in favor of the new one which is both faster, more readable.
|
|
|
|
|
|
| |
For now both old and new implementations of elision coexist,
allowing cross-checking their results.
In the next commit only the new one will remain.
|
| |
|
|
|
|
|
|
| |
This way the contract of `transformClassTemplate()` can focus on
non-AnyVal cases, which are more regular from the perspective of
transforming their templates in the constructors phase.
|
|
|
|
|
|
|
| |
The check in question relies on helper maps and methods that don't
belong outside that check. This commit encapsulates those helpers into
the newly added `checkUninitializedReads()` , thus uncluttering
`transformClassTemplate()`
|
|\
| |
| | |
Scrubbing up the IntelliJ Config
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
- Add recently sprouted modules (xml and parser-combinators)
- Replace some of the documentation with a setup script
- Update Ant build to copy Maven sourced JARs to ./build/deps.
These are included in the IntelliJ classpath.
- Define the library for Ant at the project level based on
./lib, rather than asking the user to define global library.
- Disable Type Aware Highlighting by default.
IntelliJ now can build everything within the IDE with CTRL-F9.
|
|\ \
| | |
| | | |
Refactor the cake so SymbolTable does not depend on Global
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
SymbolTable refactoring introduced some TODOs that were supposed
to be addressed after M5 release. The reason I couldn't address
those problems right away was a conflict with our plans to modularize
Scaladoc and interactive. However, we decided to delay that work
until after M5 is released so addressing TODOs is not blocked
anymore.
This commit introduces the following changes:
* Eclipse project definitions for interactive and scaladoc
depend on scala-compiler project so they are builded against
latest version of the compiler (quick) instead of STARR.
This aligns our Eclipse project definitions with build.xml
structure.
* Introduce GlobalSymbolLoaders class which wires dependencies
of SymbolLoaders with assumption of dependency on Global.
* Switch to GlobalSymbolLoaders in BrowsingLoaders,
interactive Global and ScaladocGlobal; this eliminates all
TODO comments introduced before
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
At the moment we have just one Platform: JavaPlatform in the
compiler. The whole Platform abstraction feels dubious and
Java backends need to downcast to JavaPlatform to implement some
optimizations. It seems like for now it's just better to fix platform
declared in Global to be JavaPlatform and get rid of downcasting.
I checked that even JavaScript backend declares itself as a subtype
of JavaPlatform so it seems like our abstraction is not that useful.
If we have an alternative platform with specific requirements we
might want to refactor our Platform abstraction again but for now
it seems dubious to pay a price for abstraction nobody uses.
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
We want to write JUnit tests for Scala compiler so we should depend
on `scala-compiler` project in `test-junit`.
Also, make dependencies of `scala-compiler` project exported so it's
enough to depend on `scala-compiler` to get all the transitive
dependencies that are needed by the compiler. Otherwise, one would
need to copy all dependencies of Scala compiler to all Eclipse projects
that depend on the compiler.
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
This is rather large commit so I'll first explain the motivation
behind it and then go through all changes in detail explaining
the choices I made.
The motivation behind this refactoring was to make SymbolTable
unit testable. I wanted a lightweight way of initializing
SymbolTable and then writing unit tests for subtyping algorithm,
various functionality related to Symbols, etc.
All of that should be possible by precisely controlling what we
test, e.g., create types and symbols by hand and not have them
defined in source code as we normally do in partest (functional)
tests.
The other motivation was to reduce and clarify dependencies we
have in the compiler. Explicit dependencies lead to cleaner
design. Also, explicit and reduces dependencies help incremental
compilation which is a big problem for us in compiler's code
base at the moment.
One of the challenges I faced during that refactoring was
cyclic dependency between Platform and SymbolLoaders.
Platform depended on `SymbolLoaders.SymbolLoader` because it
would define a root loader. SymbolLoaders depended on Platform
for numerous reasons like deferring decision how to load a given
symbol based on some Platform-specific hooks.
I decided to break that cycle by removing methods related to
symbol loading from Platform interface. One could argue, that
better fix would be to make SymbolLoaders to not depend on Platform
(backend) concept but that would be much bigger refactoring. Also,
we have a new concept for dealing with symbol loading: Mirrors.
For those reasons both `newClassLoader` and `rootLoader`
were dropped from Platform interface.
Note that JavaPlatform still depends on Global so it can
access phases defined in Global to implement `platformPhases`
method.
Both GenICode and BCodeBodyBuilder have some Platform specific
logic that requires casting because pattern matcher doesn't narrow
types to give them a proper refinement. Check the changes for details.
Some logging utilities has been moved from Global to SymbolTable
because they are accessed by SymbolTable. Since Global inherits from
SymbolTable this should be a source compatible change.
The SymbolLoaders has dependency on `compileLate` method defined in Global.
The purpose behind `compileLate` is not clear to me but the dependency looks
a little bit dubious. At least we made that dependency explicit.
ScaladocGlobal and Global defined in interactive has been adapted in a way
that makes them compile both with quick.comp and 2.11.0-M4 so my refactorings
are not blocking the modularization effort.
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
The 27c8266e introduced `Platform.BinaryRepr` abstract type that
was meant to abstract over binary representation of a classfile.
The abstraction was needed at the time because we had both jvm
backend and msil backend. The msil backend has been removed but
the abstraction was kept in place. The type has been deprecated
and set to be an alias for AbstractFile. The alias will be
removed later in 2.11 milestone cycle.
All references to `Platform.BinaryRepr` has been removed in
the compiler in favour of using AbstractFile directly.
If we ever need to abstract over multiple backends we should
do that based on real requirements of given backends.
This is a first step towards breaking the cyclic dependency
between Platform and SymbolLoaders.
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
ClassfileParser depends on forcing infos during typer phase.
It's not entirely clear why this is needed (git blame doesn't
help much) but I decided to preserve the logic.
Therefore, I introduced an abstract method called
`lookupMemberAtTyperPhaseIfPossible` which preserves the original
semantics but is implemented outside of ClassfileParser so the
ClassfileParser itself doesn't need to depend on typer phase and
phase mutation utilities.
I removed `loaders` override in `test/files/presentation/doc`.
I would have to update it to implement the
`lookupMemberAtTyperPhaseIfPossible` method. However, the override
doesn't seem to be doing anything useful so I just removed it.
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
ClassfileParser contained some ICodeReader-specific logic
like `forceMangledName` and `getMemberSymbol` methods.
The `getMemberSymbol` method was defined in ConstantPool class
because it must access some internal state of ConstantPool. In order
to move that method to ICodeReader we had two options:
1. Make all internal state accessible from outside of ConstantPool
class so getMemberSymbol could be implemented outside of
ConstantPool hierarchy
2. Make it possible to subclass ConstantPool in ICodeReader so
getMemberSymbol can be implemented in a subclass and can access
internal state of ConstantPool
Given the fact that getMemberSymbol mutates ConstantPool's internal
state I decided that subclassing is a cleaner approach. It required
significant refactoring because we had to make sure that we create
an instance of proper class when initializing the `pool` variable.
I ended up introducing `ConstantPoolManager` class which is essentially
a mutable variable that can be reset multiple times.
This change makes ClassfileParser independent from the ICodeReader and
its implementation details.
|
|\ \ \ |
|
| |\ \ \
| | | | |
| | | | | |
Improve type printing (toString/debugging)
|
| | | | | |
|
| | | |/
| | |/|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
More consistency as to how to understand aliases,
singletons, specialized symbols, subclasses. Fewer
weird special casings, like normalizing tuples
and functions during type printing, but nothing else.
I avoid "normalize" entirely now, and do not make
special cases for dealiasing, which is already well
handled when printing error messages.
Look at the change to test/files/neg/t2641.check to get
a sense of why we should resist calling normalize during
the early days of a compilation run.
Anonymous and refinement classes can be printed far more
usefully by revealing their parents, and that too is here.
Hardened toString against undesirable side effects.
Make toString final to discourage any further rogue
toString overriders. Make safeToString a little safer.
|
|\ \ \ \
| |/ / /
|/| | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Conflicts:
bincompat-backward.whitelist.conf
bincompat-forward.whitelist.conf
src/compiler/scala/reflect/reify/phases/Reshape.scala
src/compiler/scala/tools/nsc/symtab/classfile/ClassfileParser.scala
src/compiler/scala/tools/nsc/transform/Mixin.scala
src/compiler/scala/tools/nsc/typechecker/RefChecks.scala
src/compiler/scala/tools/nsc/typechecker/Typers.scala
src/library/scala/concurrent/impl/Promise.scala
src/reflect/scala/reflect/internal/StdAttachments.scala
test/files/neg/macro-override-macro-overrides-abstract-method-b.check
test/files/run/t7569.check
|
| |\ \ \
| | | | |
| | | | | |
SI-7455 Drop dummy param for synthetic access constructor
|
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
Java synthesizes public constructors in private classes to
allow access from inner classes. The signature of
that synthetic constructor (known as a "access constructor")
has a dummy parameter appended to avoid overloading clashes.
javac chooses the type "Enclosing$1" for the dummy parameter
(called the "access constructor tag") which is either an
existing anonymous class or a synthesized class for this purpose.
In OpenJDK, this transformation is performed in:
langtools/src/share/classes/com/sun/tools/javac/comp/Lower.java
(Incidentally, scalac would just emits a byte-code public
constructor in this situation, rather than a private constructor /
access constructor pair.)
Scala parses the signature of the access contructor, and drops
the $outer parameter, but retains the dummy parameter. This causes
havoc when it tries to parse the bytecode for that anonymous class;
the class file parser doesn't have the enclosing type parameters
of Vector in scope and crash ensues.
In any case, we shouldn't allow user code to see that constructor;
it should only be called from within its own compilation unit.
This commit drops the dummy parameter from access constructor
signatures in class file parsing.
|
| |\ \ \ \
| | | | | |
| | | | | | |
SI-7636 Fix a NPE in typing class constructors
|
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
If we encountered an erroneous super call due to a
failure in parent type argument inference, we must
avoid inspecting the untyped children of erroneous
trees.
|
| |\ \ \ \ \
| | | | | | |
| | | | | | | |
SI-7687 Handle spaces in %COMSPEC% path in scala.bat.
|
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
Double quoted %COMSPEC% to allow for spaces in the path to the
default interpreter (cmd.exe or equivalent).
|
| |/ / / / /
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
introduced in 5b54681: the end position of Postfix operators should
take the operator length into account.
review by @som-snytt
|
| |/ / / /
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
Currently we allow macros to override non-abstract methods (in order
to provide performance enhancements such as foreach for collections),
and we also disallow macros to override abstract methods (otherwise
downcasting might lead to AbstractMethodErrors).
This patch fixes an oversight in the disallowing rule that prohibited
macros from overriding a concrete method if that concrete method itself
overrides an abstract method. RefCheck entertains all overriding pairs,
not only the immediate ones, so the disallowing rule was triggered.
Now macros can override abstract methods if and only if either the base
type or the self type contain a matching non-abstract method.
|
| |\ \ \ \
| | | | | |
| | | | | | |
SI-7336 Link flatMapped promises to avoid memory leaks
|
| | |/ / / |
|
| |\ \ \ \
| | | | | |
| | | | | | |
SI-7265 General test for spec version
|
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
The test for isJavaAtLeast uses the specification.version.
The method argument must have the form "major.minor".
The scaladoc is updated to reflect the new reality and a
test is added under junit.
Note that this implementation aims to be a simple
compromise between the functional and imperative camps,
that is, to be free of both closures and while loops.
And to elicit no cruft like regexes and wrappers for strings.
No doubt even more could be done in this department, but we
don't wish to spoil the fun on codegolf.stackexchange.com.
However, we might decide to sponsor a new site:
codereviewpingpong.com
For 2.10.x, javaSpecVersion is provided as a private member.
The active test is under `run` and the `junit` test must
bide its time in `pending`.
For 2.11, the private members can be public and the app test
replaced with the unit test.
|
| |\ \ \ \ \
| | | | | | |
| | | | | | | |
SI-7649 Fix positions for reshaped tag materializers
|
| | | |/ / /
| | |/| | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
Calls to `materializeClassTag[T]` are replaced during reification with
`implicitly[ClassTag[T]]` in the `reify` macro. This is done to avoid
referring to symbols in scala-compiler.jar. Class- and Type-Tag
materialization is treated in the same way.
This commit positions the replacement trees to avoid triggering
assertions under -Yrangepos.
|
| |\ \ \ \ \
| | | | | | |
| | | | | | | |
fix typo in BigInt/BigDecimal deprecation messages
|
| | |/ / / / |
|
| |\ \ \ \ \
| | | | | | |
| | | | | | | |
SI-7617 typedAssign no longer expands lhs
|
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
This makes sure that setter and updateDynamic macros work as intended
rather than in some cases expanding incorrectly or prematurely.
Setter invocations are desugared from assignments of values to getters.
If a typecheck of an assignment's lhs yields an invocation of a getter,
then the assignment is rewritten into an invocation of a setter. However
if a getter is a macro, then it just expands, destroying the prerequisite
for desugaring. Therefore we need to disable expansion for the typecheck
of an lhs.
Similar thing happens to updateDynamic that first desugars a getter
invocation into q"$target.updateDynamic($fieldName)" and then expects
typedAssign to rewrite the corresponding Assign node into an additional
application of a partially applied updateDynamic to a rhs. Here we also
need to disable the typecheck of an lhs, because macros cannot be
partially applied.
|
| | | | | | | |
|
| | |/ / / /
| |/| | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
The crash was caused by a symbol lookup to rewire the super calls,
done after pickler, but specialization added new traits and new
members, thus making the super rewiring impossible. To avoid such
problems, this patch moves symbol lookup after specialization,
so the changes done by specialization (and miniboxing) become
visible to mixin.
NOTE: This patch will be followed by a similar patch to master.
Review by @adriaanm or @retronym.
|
|\ \ \ \ \ \
| | | | | | |
| | | | | | | |
SI-7668 Better return type inheritance for dep. method types
|
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
Return type inheritance already handles substitution of
type parameters of the overriding method for those of the
overriding.
This commit extends this to do the same for parameter symbols.
|
|\ \ \ \ \ \ \
| | | | | | | |
| | | | | | | | |
SI-7669 Fix exhaustivity warnings for recursive ADTs.
|
| |/ / / / / /
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | |
| | | | | | | |
The pattern matcher's analysis was correctly finding
models under which the match in the enclosed test
could fail. But, when trying to render that model
as a counter example, it ran into an internal inconsistency
and gave up.
That inconsistency arose from VariableAssignment, which:
> turn the variable assignments into a tree
> the root is the scrutinee (x1), edges are labelled
> by the fields that are assigned a node is a variable
> example (which is later turned into a counter example)
In the process, it notes the unreachable case `V2 = NotHandled`,
which can only arise if `V1 = Op`, ie the scrutinee is `Op(NotHandled())`.
V2 is assosicated with the path `x1.arg`. The code then looked for
any variable assosicated with the prefix `x1` and registered that its
field `arg` was assosicated with this variable assignment.
However, the assignment for `V1 = NotHandled` (another missing case)
is also associated with the path `x1`. Registering this field makes
no sense here; we should only do that for `Op`.
This commit conditionally registers the fields based on the class
of `VariableAssignment#cls`. We no longer hit the inconsistency in
`VariableAssignment#allFieldAssignmentsLegal`.
This favourably changes the results of two existing tests.
I had to tweak the counter example pruning to avoid relying on
CounterExample.==, which is flaky in the light of Nil and List().
It is possible to have:
A, B where A != B && A.coveredBy(B) && B.coveredBy(A)
Luckily it is straightforward to implement pruning entirely with
coveredBy.
|
|\ \ \ \ \ \ \
| | | | | | | |
| | | | | | | | |
SI-7620 Remove floating-point-literals-without-digit-after-dot
|