This option puts a special symbol into the generated object files that other
object files can reference to force the library to be loaded.
The next commit will modify the way we serialize autolinking information so
that importers of this module will always emit a reference to this symbol.
This means the library will be linked into the final binary even if no other
symbols are used (which happens for some of our overlays that just add
category methods to Objective-C classes).
Part of <rdar://problem/16829587>
Swift SVN r17750
This ensures that if the witness is a foreign-to-native, curry, or other lazily-generated thunk, that the thunk gets generated. This fixes natively-ObjC witnesses to Swift protocols.
Swift SVN r17739
This is fairly ugly, because we're halfway between a-function-type-takes-a-tuple and a-function-type-takes-a-set-of-parameters. However, it's another step toward -strict-keyword-arguments.
Swift SVN r17727
Previously 'NSError**' was imported directly as
ObjCMutablePointer<NSError?>, which is verbose and quite ugly
in the synthesized interfaces and overrides. This remapping
is semantically equivalent by syntactically prettier.
This name needs to go through API review.
The test uses the actual SDK; the clang-importer-sdk didn't work
because the dump of the Foundation module only shows the pieces
from the Swift files, and the Objective-C pieces are just shown
as re-exported modules. Testing the actual SDK provides good
fidelity anyway.
Swift SVN r17707
- Mine conjunction constraints for constraint failure data. (rdar://problem/16833763)
- Rather than crash, add a diagnostic to signify a missing user constraint. (rdar://problem/16747055) I don't have a deterministic repro for this to include as a test, but users hit it from time to time, I'd like to address this issue holistically, and we're hoping that the new diagnostic will help us collect isolated repros.
- As promised, remove the temporary "compiler_submit_version" build configuration predicate in time for WWDC. (rdar://problem/16380797)
Swift SVN r17705
We want to support three configurations:
* Debug (-Onone, -O0): user assertions, library precondition checks, runtime
checks enabled and verbose.
* Release (-O): library precondition checks, runtime checks enabled but succinct
(trap).
* Fast (-Ofast): all checks off.
The _isFast predicate will allow to write standard library functions to support
this plan. This commit changes fatal() to differentiate between the three modes.
Support for rdar://16477198
Swift SVN r17697
This prevents us going in an infinite optimization loop. According to Michael,
we need at least 15 for performance.
The fix at r17635 does not handle the case where the recursive function calls
itself only once.
rdar://16761933
Swift SVN r17686
of T[]() instead of Array<T>().
I didn't do this in the guts of Array.swift because the abstraction
is harmful in the implementation of Array itself.
NFC.
Swift SVN r17683
We used to just use FileContext at deserialization time. That causes
multiple NormalProtocolConformances with the same protocol and type,
but from different modules. And we will have issues at IRGen.
With this commit, we serialize the parent module and deserialize the parent
module, and we are not going to have multiple NormalProtocolConformances with
the same protocol and type.
rdar://16809311
Swift SVN r17675
Even though declarations end up in the top-level module, this can still
/expose/ them if the submodule is explicit. I'll have to think more about
this.
This reverts r17656 / <rdar://problem/16818519>
Swift SVN r17668
corresponding getter method is audited.
There are a host of other potential consistency problems
here, but this should fix a particular egregious one
that was preventing use of NSColor's CGColor property.
rdar://16846555
Swift SVN r17667
go back to disallowing ; in switch statements, people should use break
for empty statements. It is much more explicit and obvious what you mean.
Swift SVN r17662
The REPL was being rather cavalier about keyword arguments and call
expressions.
This fixes the last of the crashes under
-strict-keyword-arguments. There are still ~54 failing tests that need
to be audited or fixed.
Swift SVN r17655
While they do "write" to memory in terms of invalidating enums, that is
not a required operation, so it is ok to remove it.
rdar://16752890
Swift SVN r17654
Introduce a new locator kind for argument/parameter comparisons that
tracks both the argument and the parameter, which we will eventually
use in diagnostics more regularly. For now, this helps us smooth over
scalar-to-tuple/tuple-to-tuple/tuple-to-scalar nonsense when dealing
with calls.
Fix a pile of fallout from this change.
Swift SVN r17648
