| Commit message (Collapse) | Author | Age | Files | Lines |
| |
|
|
|
|
|
|
| |
Disables NotNull checking unless -Ynotnull is given. Closes #3568,
review by odersky.
|
|
|
|
|
| |
No review.
|
| |
|
| |
|
| |
|
|
|
|
|
|
| |
Made getters treated more like private members when debating whether to
inline. Closes #3420, review by dragos.
|
| |
|
|
|
|
|
|
| |
Further improved combine for hash tries, cutting of another 30ms (160
downto 130). Review by rompf.
|
|
|
|
|
|
| |
is now guaranteed to be initialized at most once, even when accessed
from different threads. Closes #3007, review by odersky.
|
|
|
|
|
|
| |
compile because the specialized variants of apply were not implemented
as well. Fixed by extending runtime.AbstractFunction0 instead.
|
|
|
|
|
|
| |
Under test/benchmarks there is a `bench` script to run benchmarks - it can be invoked after running building the library.
Review by rompf.
|
|
|
|
|
|
|
|
| |
Taking another shot at negative constants as annotation arguments since
r22175 didn't quite get there. I call into the constant folder with the
unfolded tree at the last point before it's going to fail the compile
anyway. Closes #3521, review by odersky.
|
|
|
|
|
|
|
|
| |
Most of the iterate implementations were calling the given function one
too many times, leading to tragic failure if the function could not
handle this (such as repeatedly applying tail.) Closes #3540, review by
prokopec.
|
|
|
|
|
| |
sabbus also edited to add parallel collections to the library jar - review by phaller
|
|
|
|
|
|
|
| |
Patch from Jason Zaugg so singleton types aren't accidentally deprived
of a fourth try at satisfaction. Already reviewed by odersky, so no
review.
|
|
|
|
|
|
|
|
| |
but in the interest of not breaking backwards compatibility, the
JSON.parse method has been marked deprecated for now.
Unit tests have been fixed so that this won't break the build this time.
|
|
|
|
|
|
|
| |
Tracked down why the jvm/natives.scala fails for me and apparently not
anyone else. Rebuilt libnatives.jnlib to accomodate x86-64, and it seems
to pass. No review.
|
|
|
|
|
|
|
| |
ran across some tortured logic trying to accomodate the long abandoned
idea of having 5 != 5L, so simplified the contains method. Closes #3518,
no review.
|
|
|
|
|
|
|
| |
Burned by a last minute adjustment, I lost the downward counting
direction. It is a seriously fiddly process to adjust Range and I don't
recommend it to anyone. Closes #3529 over again. Review by prokopec.
|
|
|
|
|
|
|
|
|
|
|
|
| |
scala> (1 to 1 drop 1) == (1 to 1)
res0: Boolean = true
It was introduced in r21349 which was to fix #2535, but led to #3529.
I humbly suggest we can't afford to introduce bugs of this severity in
the pursuit of corner cases such as Ranges which use Int.MaxValue as a
boundary. And looking at the patch I find the "after" code a lot harder
to follow. Closes #3529. Review by prokopec.
|
|
|
|
|
|
|
| |
Reverts r21973, the patch I characterized as "hacky but no-risk" in my
commit message, for causing #3480. Closes #3480. I'd say no review but
who can trust a guy who throws around "no risk" with such abandon.
|
|
|
|
|
|
|
|
|
| |
Restored Source factories to a form source compatible with 2.7.7. No
default implicit arguments, now low priority saves the day with a low
priority codec which io.Codec offers as last resort. Dropped the line
separator argument to getLines and made it act in a separator agnostic
way (any of \r\n, \r, or \n is a separator.) Review by community.
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
|
| |
This commit reimplements filter for Streams, but does not reimplement
map in StreamWithFilter.
The problem is that GC can't collect instances of Streams residing on the stack if there are multiple references to the Stream (more than a single one on the stack on which a Stream method is invoked). In the case of a StreamWithFilter, being an inner class, there is always an `$outer` reference to the outer object, so there is little GC can do. Possible solution - change the return type of WithFilter to something else (in TraversableLike) to allow it to return objects that don't have to subclass TraversableLike.WithFilter, and reimplement the withFilter method in Stream to simply call `filter` method - in the case of Streams, `withFilter` has little sense in either case...
|
| |
|
| |
|
| |
|
|
|
|
|
| |
according to my tests, they all currently work.
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
also consider non-implicit locals when checking shadowing of implicits:
closes #3453
nonImplicitSynonymInScope implements the predicate that is used in
tryImplicit's checks for shadowing of locally defined implicits
benchmarking shows the predicate does not significantly affect quick.comp+quick.lib (goes from 11min to 11min2s on my machine -- no optimisations)
review by odersky
|
|
|
|
|
|
| |
Adds comments for #3471, and a test case which should have been added
earlier. Review by extempore.
|
| |
|
| |
|
|
|
|
|
|
| |
Added a migration warning for matches and instance tests when it might
be an Array/Seq test whose answer has changed. Review by odersky.
|
|
|
|
|
| |
attention in case the ticket was being left open intentionally.)
|
| |
|
|
|
|
|
| |
Addresses see #3470 by adding a method Reactor.restart. Review by rompf.
|
|
|
|
|
|
| |
Fixed an xml issue arising from arrays no longer being recognized as
sequences. Review by dpp.
|
| |
|
|
|
|
|
|
| |
I figured it'd be a miracle if I got the script test past our fickle
robot overseers. Disabled, no review.
|
|
|
|
|
|
| |
Added tests for bugs which have been fixed since beta1. Closes #2691,
#3020. No review.
|
|
|
|
|
|
|
|
| |
Changed the script runner mechanism to alchemize from AST atoms rather
than generating wrapper source, and fixed script position reporting.
This patch does not include a discussed change to mark some positions as
synthetic. Closes #3119, #3121. Review by milessabin.
|
| |
|
| |
|
|
|
|
|
|
|
|
| |
Removed some flotsam which should have been removed when I rolled back
partest. (We ended up with two copies of the methvsfield test, one in a
subdir and one not.) I would suppose it is responsible for #3464 but I
don't know for certain. No review.
|
|
|
|
|
|
|
| |
Moved the test for #3420 to pending and reopened the ticket. The bug
only materializes if scalac was itself built with -optimise. (We may
need a richer language for describing optimise bugs.) No review.
|
|
|
|
|
|
| |
Make the inliner be more careful about where it pokes around. Closes
#3252, #3430. Review by dragos.
|
| |
|