This isn't a "complete" port of the standard library for embedded Swift, but
something that should serve as a starting point for further iterations on the
stdlib.
- General CMake logic for building a library as ".swiftmodule only" (ONLY_SWIFTMODULE).
- CMake logic in stdlib/public/core/CMakeLists.txt to start building the embedded stdlib for a handful of hardcoded target triples.
- Lots of annotations throughout the standard library to make types, functions, protocols unavailable in embedded Swift (@_unavailableInEmbedded).
- Mainly this is about stdlib functionality that relies on existentials, type erasure, metatypes, reflection, string interpolations.
- We rely on function body removal of unavailable functions to eliminate the actual problematic SIL code (existentials).
- Many .swift files are not included in the compilation of embedded stdlib at all, to simplify the scope of the annotations.
- EmbeddedStubs.swift is used to stub out (as unavailable and fatalError'd) the missing functionality.
Use 'copy_if_different' instead of 'file(GENERATE'. Somehow
'file(GENERATE' failed with
CMake Error: Files to be generated by multiple different commands:
Write source file list using 'file(GENERATE' so that they are updated
only if thier content is changed. Otherwise stdlib modules are rebuild
everytime cmake configuration happens.
'.private.swiftinterface' should be a part of the module. And it's
declared as `OUTPUT` file list, but they were never created. Because of
that the target was always rebuilt.
Ubuntu 18.04 seems to be defaulting to 'ascii', causing an error when
using line-directive combined with the utf-8 diagnostic output used when
early swift syntax is enabled.
Some non-stdlib thing e.g. swift-backtrace still might be built before
the runtime is built. For building Swift code in Linux "hosttools",
always set 'LD_LIBRARY_PATH' to the runtime in the builder.
Header search path for Swift shims is "usr/lib/swift". Don't rely on
clang looking for module maps in immediate subdirectories like "shims"
but add a module map to "usr/lib/swift" sourcing "shims" modules.
rdar://106677321
These headers need to be ahead of all the LLVM headers, which are added
before the flags are. Add a new parameter to pass through headers to add
as a prefix.
Resolves rdar://113647684.
The CxxStdlib module is built statically only. We would previously
build the static library but indicate dynamic linking. This would
incorrectly code generate in the client on Windows making it impossible
to use the module.
Windows names static libraries with a `lib` prefix and a `lib` suffix.
This differentiates them from the import libraries which have no prefix
and a `lib` suffix. This adjustment enables the parallel installation
of import libraries and static library variants for a given module.
This is required to support static and dynamic library co-existence in
Swift.
This makes the `stdint` module implicit which repairs the ability to
build some components. In order to accomplish this, we need to
potentially break the fragile Swift build system. Due to the incorrect
handling of compilers we need some workarounds to support
cross-compilation. This removes the injected system header paths when
building on Windows to ensure that the clang resource headers are not
following the system headers which breaks the modules as the clang
resources are dependent on the system headers when running in hosted
mode.
Because the rpath isn't set, the dynamic linker can't find the Swift
libraries when we try to run `swift-backtrace` on Linux (when we
actually install everything).
rdar://113337854
As of CMake 3.25, there are now global variables `LINUX=1`, `ANDROID=1`,
etc. These conflict with expressions that used these names as unquoted
strings in positions where CMake accepts 'variable|string', for example:
- `if(sdk STREQUAL LINUX)` would fail, because `LINUX` is now defined and
expands to 1, where it would previously coerce to a string.
- `if(${sdk} STREQUAL "LINUX")` would fail if `sdk=LINUX`, because the
left-hand side expands twice.
In this patch, I looked for a number of patterns to fix up, sometimes a
little defensively:
- Quoted right-hand side of `STREQUAL` where I was confident it was
intended to be a string literal.
- Removed manual variable expansion on left-hand side of `STREQUAL`,
`MATCHES` and `IN_LIST` where I was confident it was unintended.
Fixes#65028.
Section scans (for metadata, protocols, etc.) can be costly. This change adds tracing calls to those scans so we can more easily see how much time is spent in these scans and where they're initiated.
This adds an os_signpost implementation controlled by SWIFT_STDLIB_TRACING, and a default empty implementation for when that's disabled.
rdar://110266743
Cxx & CxxStdlib modules are Swift-only, they do not require invoking clang directly.
When building with `SWIFT_INCLUDE_TOOLS=NO`, Clang is not available as a CMake target (see `swift_common_standalone_build_config`).
rdar://107780733
We have various pieces of SPI that are getting used in tests that are
failing occasionally. The swiftmodule file has the SPI, so if the
compiler loads from the swiftmodule file, everything works fine. If the
compiler loads the interface from the swiftinterface file, it fails to
find the `startOnMainActor` declaration.
We need to disable opaque pointers when we compiler the runtime the
linker complains:
```
ld: Opaque pointers are only supported in -opaque-pointers mode
```
Alternatively, we could pass the opaque-pointers flag to the linker but
then it would complain about the swiftc generated files which still use
typed pointers.
Until swiftc is fixed, disable opaque pointers for .cpp runtime files.
rdar://106515243
Implicit imports were off for library builds already, but we need them off
for executable builds too, otherwise we have problems with _StringProcessing.
Added some extra code to AddSwiftStdlib.cmake so executable targets can
specify target SDKs the same way libraries currently can.
Updated the Backtracing targets to specify just OS X for now.
While I was doing this, it turns out Saleem was fixing things to avoid
having to patch the Windows include directories, which is awesome but
necessitates an extra change to the backtracing stuff to make the build
not fail on Windows.
rdar://105409147
