The C++ code was very fragile in terms of ABI dependencies and broke
when the standard library was built with -enable-resilience.
The actual reason it broke is that case numbering changes when
resilience is enabled, but instead of messing with that, it seemed
more logical to rewrite this routine in Swift instead, to avoid
ABI dependencies altogether.
This requires using the "shadow protocol" trick to call NSNumber
methods, since we cannot import NSNumber from the stdlib.
Make _MagicMirrorData @_fixed_layout, but not the concrete mirror
implementations.
To make the calling convention work in resilient builds, make the
runtime entry points into top-level functions that take a
_MagicMirrorData, instead of adding @_silgen_name attributes on
methods.
This involves changing the convention on the 'owner' parameter
from +0 to +1.
Also, there was a memory leak in the old enum code that I noticed
by inspection. We would copy the enum value into a box, strip
the enum tag bits, take the box contents but never free the box
itself.
The fix isn't very satisfying either -- we have to modify the
source enum in-place to strip tag bits, then we copy it into a
box having the right payload type, and add the tag bits back.
At least this way, we can free the box after.
We have a special rule that Optional<T>.none successfully dynamically casts
to Optional<U>.none for any T and U. However the implementation was incorrect
if the source and destination types had a different size. We would initialize
the source to nil, and then copy to the result.
The correct implementation is to initialize the result using the result
payload type directly, and not call _succeed() at all.
Fixes <https://bugs.swift.org/browse/SR-1056>.
We have a special rule that Optional<T>.none successfully dynamically casts
to Optional<U>.none for any T and U. However the implementation was incorrect
if the source and destination types had a different size. We would initialize
the source to nil, and then copy to the result.
The correct implementation is to initialize the result using the result
payload type directly, and not call _succeed() at all.
Fixes <https://bugs.swift.org/browse/SR-1056>.
In PE-COFF formatted file, the offset of NT header can be greater than 0xFF. MS library defines it as LONG (32bit signed integer) in struct _IMAGE_DOS_HEADER.
This reverts commit 893d1dc523.
This looks like a likely culprit that broke tests on the iOS Simulator:
Failing Tests (6):
Swift :: IRGen/class_resilience.swift
Swift :: IRGen/concrete_inherits_generic_base.swift
Swift :: IRGen/enum_resilience.swift
Swift :: IRGen/foreign_types.sil
Swift :: IRGen/nested_types.sil
Swift :: IRGen/struct_resilience.swift
need to be modified by the runtime, and only actually store to them
when that would change anything.
Unfortunately, Linux is considerably better than Darwin at shaking
these bugs out because Darwin will leave global data mutable after
resolving relocations in it.
initialization in-place on demand. Initialize parent metadata
references correctly on struct and enum metadata.
Also includes several minor improvements related to relative
pointers that I was using before deciding to simply switch the
parent reference to an absolute reference to get better access
patterns.
Includes a fix since the earlier commit to make enum metadata
writable if they have an unfilled payload size. This didn't show
up on Darwin because "constant" is currently unenforced there in
global data containing relocations.
This patch requires an associated LLDB change which is being
submitted in parallel.
initialization in-place on demand. Initialize parent metadata
references correctly on struct and enum metadata.
Also includes several minor improvements related to relative
pointers that I was using before deciding to simply switch the
parent reference to an absolute reference to get better access
patterns.
The witness-table parameters got added to all witnesses as part of the
resilience work, but the hardcoded witness table in the runtime's
dynamic-casting infrastructure didn't get updated. Nothing seems to be
relying on these right now, so we cannot actually *test* it, but I've
verified that the types line up.
This lets us eliminate the _getObjectiveCType() value witness, which
was working around the lack of proper type witness metadata in witness
tables. Boilerplate -= 1.
clang and gcc provide a preprocessor macro called `__USER_LABEL_PREFIX__` which
provides the user label prefix for the specific target that the translation unit
is being built for. Rather than trying to reconstruct the logic in place via
various checks, fallback to the compiler to provide this information. Although
this limits the compilers (MSVC does not provide this preprocessor macro
definition), the only supported compiler ATM is clang, and it has provided this
definition for some time now.
This addresses the FIXME that was associated with the user label prefix being
applied under specific cases.
NFC.
It's to be used by code produced by the ReleaseDevirtualizer.
As the function is only used for non-escaping objects, the deallocating bit is set non-atomically.
Be more conservative in terms of masking ISAs. This reduces tight coupling with the objc runtime. This commit adds the required calls to IRGen and the runtime, and a test case to make sure IRGen is correct.
Teach swift_deallocPartialClassInstance how to deal with classes that
have pure Objective-C classes in their hierarchy. In such cases, we
need to make sure a few things happen:
1) We deallocate via objc_release rather than
swift_deallocClassInstance.
2) We only attempt to find an execute ivar destroyers for
Swift-defined classes in the hierarchy
3) When we hit the most-derived pure Objective-C class, make sure that we
only execute the dealloc of that class and not any of the subclasses
(which would end up trying to destroy ivars again).
Fixes rdar://problem/25023544.
This reverts commit 2262bd579a.
This information isn't necessary for field descriptor lookup,
after all. It's only the fields that need to have generic information,
which is already in the field descriptor.
Previously, the mangling didn't include generics, but these are
needed to key off of the new field descriptor metadata, as well
as to construct type references for the nominal type.
- Add RuntimeTarget template This will allow for converting between
metadata structures for native host and remote target architectures.
- Create InProcess and External templates for stored pointers
Add a few more types to abstract pointer access in the runtime
structures but keep native in-process pointer access the same as that
with a plain old pointer type.
There is now a notion of a "stored pointer", which is just the raw value
of the pointer, and the actual pointer type, which is used for loads.
Decoupling these allows us to fork the behavior when looking at metadata
in an external process, but keep things the same for the in-process
case.
There are two basic "runtime targets" that you can use to work with
metadata:
InProcess: Defines the pointer to be trivially a T* and stored as a
uintptr_t. A Metadata * is exactly as it was before, but defined via
AbstractMetadata<InProcess>.
External: A template that requires a target to specify its pointer size.
ExternalPointer: An opaque pointer in another address space that can't
(and shouldn't) be indirected with operator* or operator->. The memory
reader will fetch the data explicitly.
MetadataLookup.cpp and ProtocolConformance.cpp has same part for inspecting dynamic libraries.
The common code exist in one file and other uses it.
This uses the argument passing to callback in Linux/Cygwin and not applied to OS X.
"minimal" is defined as the set of requirements that would be
passed to a function with the type's generic signature that
takes the thick metadata of the parent type as its only argument.
