This is part of our poor-man's internal compiler SPI hiding
in the standard library. We don't want these functions showing
up in code completion, etc.
Swift SVN r16916
Until we lock down the Swift ABI and ship with the OS, we need to be resilient
in the face of ObjC dynamic subclassing and OS changes. In practice, this means
that we need to have a swift runtime ABI to read the isa out of objects. I've
added it as of r. See: swift_getClassMetadata()
We can and will optimize swift_getClassMetadata into a single instruction once
we lockdown our ABI and ship with the OS.
See also: <rdar://problem/16735599>
Swift SVN r16889
We really don't need to support individual objects
this large, much less more than 4 billion fields in
a single type.
Also rearrange the fields to bring the instance
size/alignment fields closer to the class header,
just for a minor locality win.
Swift SVN r16879
Also fix a latent bug in the logic in the disabled code to scribble over memory durring dealloc.
<rdar://problem/15855042> QoI: abort on resurrection
Swift SVN r16827
Add value witnesses for destroyArray, initializeArrayWithCopy, and initializeArrayWithTake{FrontToBack,BackToFront}, and fill out the runtime value witness table implementations. Stub out the IRGen ones for now.
Swift SVN r16772
As it turns out, Swift classes can stumble into having a non-pointer isa.
This might be a problem for 1.0, but that is beyond the scope of this bug.
Swift SVN r16721
The true designated initializers take a variadic argument, which we can't directly support in Swift, so we'll map those definitions to refer to versions that drop the variadic parameters altogether, and reimplement the variadic interface in the overlay.
Swift SVN r16711
Currently this prevents building Swift on 10.9.
Right now CrashReportClient.a is used regardless of the SDK being used,
and the CMake logic does a check for the .a across different platforms
and SDKs without checking what the build product target OS actually is.
The current logic somewhat works because the SDKs will be in different
Platforms, but it is fragile.
Swift SVN r16678
The cost of hacks to swift_conformsToProtocol is starting to outweigh any benefit to being principled here. We'll get a linker error now if multiple modules declare a conformance for the same type to the same protocol, but that's arguably a good thing for 1.0 anyway, since we aren't set up to get that right in other ways.
Swift SVN r16554
When we build the standard libary with -parse-stdlib the frontend sets the
assert configuration to 'DisableReplacement'. Constant replacement does not take
place and the call to the builtin function 'assert_configuration' call stays in
the serialize SIL of the swiftmodule.
IRGen replaces the function call to the assert_configuration builtin function by
the value for Debug (0). The resuling standard library dylib hence contains the
debug version of the standard library assert function.
Frontend optimization flags can now determine whether asserts should be executed
or not.
This commit removes the SWIFT_ASSERTS cmake flag.
rdar://16458612
Swift SVN r16473
With this check-in, we get Mirrors for IntXX, UIntXX, Float32, and we don't lose the existing ones for Double, Int, String and Bool
This fixes rdar://16517273 and will make playgrounds able to display more data more often more consistently!
Swift SVN r16418
Double's like '0.9999999' were being printed as '1.0'.
'print' isn't a first-class I/O library anyway; it's main use right
now is accurate reporting in the REPL.
Fixes <rdar://problem/16603548>.
Swift SVN r16305
The strategy we are using is to obtain the class name from the ObjC runtime (as apparently nominal type descriptor for ObjC-wrapped types do not contain a useful name string) and then crafting the appropriate mangled name for the witness
Also, make AppKit another source of conformances. We currently have custom conformances for Reflectable that we are in Foundation (NSURL) and AppKit (NSView) - if more modules need to have custom mirrors, we might end up having to expand the list
Swift SVN r16280
the value buffer comes first.
The motivation for doing this is similar to the
motivation for moving it for class existentials:
it eliminates the need for an offset for the most
common accesses, which is particularly important
for the generic value witnesses.
Also try to hard-code that layout in fewer places,
or at least static_assert the places that have to
do so.
Swift SVN r16279
The actual logic to do this is simple; the vast majority of this
commit is just a pile of changes to test cases to reflect the fact
that Objective-C metadata now includes the module name for each class
and the mangling of Swift-defined @objc classes no longer goes into
the "So" namespace for Objective-C classes. Finishes
<rdar://problem/15506580>.
Swift SVN r16274
pointer first.
This most important effect of this is that accesses to that
field don't need to be dynamically offsetted past an arbitrary
number of value witnesses, which is pretty nice for the
generic value witnesses.
Swift SVN r16243
Its layout information in the ObjC runtime is a complete lie, so we crash if we try to use it to introspect the object. Fixes <rdar://problem/16592777>.
Swift SVN r16228
