| Commit message (Collapse) | Author | Age | Files | Lines |
| |
|
|
|
|
|
|
|
|
|
|
| |
Stole two bits from Tree#id to record whether a tree contains an
error tree and whether its subtrees have been checked. This leaves us
with only a single gigatree of headroom for replaying trees in the
presentation compiler. Alert the media.
Review by odersky.
|
|
|
|
|
|
| |
"Pure expression in statement position" is hopefully a lot quieter now.
Review by dragos.
|
| |
|
|
|
|
|
|
| |
Started work on compiler toolbox that can compile reflect trees at
runtime.
|
| |
|
|
|
|
|
| |
Added module for importing trees and symbol tables from another universe
|
|
|
|
|
| |
fixed svn props and scaladoc tags, did some cleanup
|
|
|
|
|
| |
No review.
|
|
|
|
|
|
| |
Changes to Liftcode to use new reflection semantics, where a compiler
uses type checking.
|
|
|
|
|
| |
Review by Dragos.
|
|
|
|
|
|
|
|
|
|
| |
I added local logging to zip and zipped and listened for who was
dropping things on the floor. Everything in this commit stems from that.
Sometimes the fix was uncertain and I sprinkled some logging. If you've
been hanging back with lots of internals knowledge waiting for the right
commit to review, this would be a good one. But since knowledgeable
people are hard to find, I'll go with review by moors.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
If you build and then run
tools/proxy-report <outdir>
you will be given files revealing interesting things like:
/** With respect to trait SetLike, trait SetProxyLike does NOT wrap:
*/
trait Unwrapped {
def +(elem1: A,elem2: A,elems: A*): This
def ++(elems: scala.collection.GenTraversableOnce[A]): This
def subsets(len: Int): Iterator[This]
def subsets: Iterator[This]
protected override def newBuilder: scala.collection.mutable.Builder[A,This]
protected override def parCombiner: scala.collection.parallel.Combiner[A,scala.collection.parallel.ParSet[A]]
}
Lots more possible here, for now I just want to get a backstop
against our worst transgressions (not just in the library -- look
at SimpleTypeProxy, if you dare!) I will inquire about the results
separately, so no review.
|
|
|
|
|
|
|
| |
Iterators should have def next(), not def next. Clearing
a mutable structure should be done with clear(), not clear.
And etc. No review.
|
| |
|
|
|
|
|
|
| |
Not O(n) like it was. Here's another good candidate for some mythical
performance regression tests. Closes SI-4933, no review.
|
|
|
|
|
|
| |
This seems less likely to surprise people attempting to use them as
implicits. No review.
|
|
|
|
|
|
| |
Scaladoc enhancements to continuations package for docspree. Review by
rompf
|
|
|
|
|
|
| |
Small changes to the API documentation of the AnyVal types. Contributed
by Michael Allman during the September doc spree. No review.
|
|
|
|
|
|
| |
Improved API documentation for scala.compat.Platform. Contributed by
Matthew Farwell during the September doc spree.
|
|
|
|
|
|
|
| |
Hugely improves the API documentation for JavaTokenParsers, Parsers,
and RegexParsers. Contributed by Daniel Sobral during the September doc
spree.
|
|
|
|
|
|
| |
Fixes build issue involving missing rootdoc.txt in strap build. No
review.
|
|
|
|
|
|
| |
Scaladoc now hides members with @bridge annotation, contributed by Simon
Ochsenreither. Closes SI-4924
|
|
|
|
|
|
| |
Fixed a Scaladoc layout issue affecting how type parameters (@tparams)
are displayed. No review.
|
| |
|
|
|
|
|
|
|
| |
Adds ability to document root package, as well the first cut of an
introduction to the API. Contributed by Iain McGinniss during the
September doc spree.
|
|
|
|
|
|
|
|
|
|
|
| |
extempore can't live with that level of duplication. I eliminated
400 lines. I could eliminate more: there are distinctions which seem
unimportant, but I'm not sure so I left them in place. Example, is a
"ContextError" something meaningful other than "an error which one emits
by calling context.error"?
Review by plocinic.
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
| |
Eliminated the var, the Option[Boolean]s, and the Unit-returning init
method. If they were accomplishing something I could not discern, I
trust hubert will now apprise me. Review by plocinic.
|
|
|
|
|
| |
Hey, that new warning works. No review.
|
|
|
|
|
|
| |
Amend previous commit by suppressing type errors for syntactically
incorrect units only when -Ypresentation-strict is set. no review.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Better behavior with problematic sources in the presentation compiler:
- if the presentation compiler crashes while type checking a source file, it puts it into an
ignore buffer, and it will not type check it during background compilation rounds.
Any subsequent calls to waitLoadedTyped will fail with the last exception.
To remove it from the ignore list call either 'enableIgnoredFile' or 'askToDoFirst'
on that file.
- sources with syntax errors are not type checked during background compilation. This leads
to less spurious errors. For instance, type errors may occur much earlier in the file than
the actual syntax error that caused them.
review by odersky.
|
|
|
|
|
|
|
| |
Added new disabled test for eclipse scala-ide ticket1000609. Corrected
test framework to fail nicely when the tree retrieved after an has no
associated symbol.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This commit doesn't include the line which would change anything
interesting (that being making toString in Name a val), it is all
the changes elsewhere which one would have to make to do that. I
didn't want to waste them. A quirk of scala is that if you call an
empty-parameter-list method:
foo.bar() // e.g. x.toString()
foo.bar // where you could have called x.toString
then if you decide in the future to make it a val, every call site must
potentially be modified. This is a bit of a bummer. One can try to
defend against that future decision by eschewing parentheses, but it's
not very satisfying or effective. No review.
|
|
|
|
|
| |
Fixed stackoverflow problem when initializing large scopes.
|
|
|
|
|
|
|
| |
This saves an unnecessary field (the lazy bitmap) and means hashing is
faster. The hashCode is called on every created type anyway because of
the lookup in unique, so this is pure gain. No review.
|
|
|
|
|
| |
No review.
|
|
|
|
|
|
|
| |
Removed redundant UnCurry transform (now in LiftCode.)
No review.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The pieces of TypeMap which returned Option[T] were contributing many
unnecessary allocations as well as inconsistency and complexity. By
creating distinguished objects UnmappableTree and UnmappableAnnotation,
I was able to gain simplicity, eliminate all the Some allocations, use
the normal mapConserve, and so on.
I also went after the monster that was dropNonConStraintAnnotations by
generalizing it into "AnnotationFilter", a trait which extends TypeMap.
SpecializeTypes was able to take advantage of this as well. (The means
by which (some) annotations are dropped, and the related fact that
NullaryMethodType returns constant false for isTrivial only so the
continuations plugin will keep working, leaves the situation still in
dire need of improvement.)
Review by moors.
|
| |
|
|
|
|
|
|
|
| |
When a reasonable lub cannot be arrived at by direct means, use the
bounds of the lub inputs to derive bounds for the lub rather than giving
up. Closes SI-4938, review by moors.
|
| |
|
|
|
|
|
| |
Improved getFiles/getDirs in PartestTask; did some cleanup.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Partially reverted r25636 change (forget notice about the starr
rebuild).
Change originated in my misunderstanding of Ant filters:
The pattern string if <exclude name="<pattern>"/> may not contain
Ant variables such as eg. ${compiler.excludes}.
Use <excludesfile name="${compiler.excludes}" if="<property>"/>
instead !
|
| |
|
|
|
|
|
| |
No review.
|
|
|
|
|
|
|
| |
Report erroneous tree instead of Error tree in case we are dealing with
presentation compiler. This way scala plugin doesn't have to be aware of
Error trees in most common cases. Some minor tweaks
|
|
|
|
|
|
|
|
|
| |
I had all the variations of zip/zipped start logging when they being
were given differently lengthed arguments, to find out what all we might
be throwing away. This uncovered at least one bug with free variable
identification, fixed herein, and also solidifed my belief that we
should have a lot less ad hoc zipping of things. No review.
|