Not only it does not achieve what the comment says (_storage has the
type _ContiguousArrayStorageBase), but optimizations that would break
semantics of this code would be invalid anyway.
Swift SVN r29056
The <Redacted>13A261 iOS SDK adds an availability attribute marking the imported
WatchKitErrorCode enum as available only on iOS 8.2 and newer. The WatchKit overlay
has a protocol extension for WatchKitErrorCode, so mark the extension as only available
on 8.2 and newer. On <Redacted>13A261 this silences a diagnostic complaining that
WatchKitErrorCode is only available on 8.2 and newer.
However, on <Redacted>13A254 this new annotation itself causes an error because on that
SDK WatchKitErrorCode claims to be available on all iOS versions and yet the conformance
would only hold on 8.2 and newer. To make this second diagnostic go away, this
commit temporarily disables availability checking for the WatchKit overlay.
rdar://problem/21099550 tracks re-enabling availability checking when we update the bots,
etc. to <Redacted>13A261.
rdar://problem/21098396
Swift SVN r29026
CNErrorDomain is lacking proper availability annotations, and
explicitly linking with -weak_framework is not working reliably. For
now, just remove the overlay to solve rdar://problem/21030937. We'll
bring it back when rdar://problem/21032649 makes it into all of the
builds we care about.
Swift SVN r29003
Fixes <rdar://21091371>. When SequenceType acquired a O(N) 'contains'
method, the code for pattern matching a Range started using that method
instead of implicitly deducing a HalfOpenInterval.
We still have a problem though:
(1_000_000..<1_000_000_000).contains(1)
will compile and appear to hang at runtime. I'll bring this up on the
mailing list.
Swift SVN r28998
This changes things like Swift.Dictionary<Swift.Int, Swift.AnyObject> to Dictionary<Int, AnyObject>
It has been suggested that playgrounds would benefit from not showing the fully qualified name.
Playgrounds use the runtime demangler to obtain type names, and honestly, I do not see enough value in those qualifiers being printed out to justify hackery in PlaygroundLogger or separate demangling logic
Swift SVN r28997
The protocol is still needed in order to make sure mixed signed/unsigned
arithmetic doesn't compile, but after pulling all its "real"
requirements into UnsignedIntegerType, that's now its only function.
Thus, rename it to _DisallowMixedSignArithmetic.
Swift SVN r28980
Darwin libc and glibc use different typedefs for the signal handler type, and SIG_HOLD has different values. Conditionalize the definitions in their respective overlays.
Swift SVN r28921
This reverts commit r28892, r28894, and r28895.
They broke validation tests; JoeG is going to look at what's needed to
make them work again.
Swift SVN r28897
This has passed review, or at least satisfied Tony Parker, provided we
do something to hide SetAlgebraDispatchType. I think I can eliminate it
in an imminent commit.
Swift SVN r28892
Previously a mirror for an empty case would have one entry, consisting
of the case name and a value of ().
Now, the mirror's summary shows both the type and the case. If there is
no payload, the mirror will not have any children. The presence of a
child indicates there is an actual payload.
Also put a non-ASCII character in the reflection test to ensure the
runtime's primitives can round-trip UTF8.
Fixes <rdar://problem/20994093>.
Swift SVN r28874
