Add a `SWIFT_STDLIB_OVERRIDABLE_RETAIN_RELEASE` CMake option. When set to true, swift_retain/release and the other functions in InstrumentsSupport.h can be overridden by setting the appropriate global function pointer, as is already the case. When set to false, those function pointers are removed and the functions always go into the default implementation.
Set `SWIFT_STDLIB_OVERRIDABLE_RETAIN_RELEASE` to false when building the minimal stdlib, and set it to true otherwise by default.
rdar://115987924
swift-evolve hasn’t been used in the recent past. @beccadax and I agreed that it doesn’t provide any value anymore to keep swift-evolve around. In Becca’s words:
> The resilience implementation is pretty stable at this point, and it never seems to have had many of the bugs that swift-evolve would have been able to shake out.
This patch adds a new build products to build and test the Swift stdlib
for WebAssembly.
This adds WebAssembly specific stdlib product instead of adding the new
target in `stdlib-deployment-targets` because unlike darwin platforms
Wasm target has quite different stdlib configuration from host target
and there is no way to specify different stdlib configuration for each
target in `stdlib-deployment-targets` for now.
This patch adds 1 new dependency `wasi-libc` and 3 new build products:
1. `WASILibc` - wasi-libc is a libc implementation for WebAssembly
System Interface (WASI). This product is required to build the
compiler runtimes and Swift stdlib.
2. `WasmLLVMRuntimeLibs` - This product builds `./runtimes` directory
in llvm-project, which contains compiler-rt, libcxx and libcxxabi.
This product is required to build Swift stdlib.
3. `WasmStdlib` - This product builds standalone Swift stdlib for
WebAssembly. (Similar to `MinimalStdlib` product, but configured for
WebAssembly target)
* swift_build_support: fix `shell.run` bytes printing
`output.splitlines()` returns a value of `bytes` type, not `str`, which leads to `b` prefixed to it when printing.
This led to broken output when running `update-checkout`. Before:
```
[swiftpm] b"Already on 'main'"
```
After:
```
[swiftpm] Already on 'main'
```
* swift_build_support/shell.py: use `errors='replace'` in `decode`
Use FetchContent to include swift-syntax directly in swift. This can be
thought of as an `add_subdirectory` for a directory outside the root.
The default build directory will be `_deps/swiftsyntax-subbuild/`, though
the modules and shared libraries will be built in `lib/swift/host` by
passing down `SWIFT_HOST_LIBRARIES_DEST_DIR` to avoid copying them as we
were doing previously.
It doesn’t really matter but the SourceKit stress tester and SwiftEvolve share the same repo and have a very similar build architecture and they should be build next to each other by build-script.
Some build scenario (e.g. lldb) want swift-syntax libraries installed
but don't want to install "swift" compiler products.
Install 'earlyswiftsyntax' products independently when
'--install-swiftsyntax' is specified but '--install-swift' is NOT
specified.
In the fullness of time, we want to split the full build-script-impl pipeline so
that we can begin moving library like products (libdispatch, foundation) from
build-script-impl into build-script. We are not there yet since some of swift's
concurrency tests have a dependency on swift dispatch being built. This breaks
the build and we need to extract those tests into a separate product. But for
now, this makes sense to repair the build.
rdar://89046735
For compiling codes required for macro support, we now need swiftc
compiler in the build machine.
Unlike Darwin OSes, where swiftCore runtime is guaranteed to be present
in /usr/lib, Linux doesn't have ABI stability and the stdlib of the
build machine is not at the specific location. So the built compiler
cannot relies on the shared object in the toolchain.
These builds do not work outside of darwin currently. Rather than adding
the skip to all the necessary build presets, just disable in the product
instead.
This library served its purpose and has overstayed its welcome.
The library shipped in Xcode can't change again from the state it was
in, so changes to the sources here will not affect that library, which
is confusing. The library does not currently build due to changes in the
runtime headers, so it does not give meaningful signal to anything
anymore. If you need to test things in a backdeploy concurrency
environment, use the copy from the toolchain in Xcode as that will give
you a far clearer picture of what the code will actually be running with
than the state of the sources here did.
This solves deprecation warnings in build-script:
```
DeprecationWarning: 'pipes' is deprecated and slated for removal in Python 3.13
```
This change assumes that the minimum version
of Python3 is 3.3, which has `shlex.quote`.
Since our build bots currently use 3.6 as their
Python version, we can't yet use `shlex.join`
to further simplify some of the code using
quote.
Currently, when building LLVM/clang on Alpine Linux, CMake toolchain file specifies incorrect `<cpu_arch>-unknown-linux-musl` target, which makes the build immediately fail. Correct target that allows building on Alpine should be specified as `<cpu_arch>-alpine-linux-musl`, with `<cpu_arch>` replaced with respective CPU platform, e.g. `aarch64`.
Currently, when building LLVM/clang on Alpine Linux, CMake toolchain file specifies incorrect `<cpu_arch>-unknown-linux-musl` target, which makes the build immediately fail. Correct target that allows building on Alpine should be specified as `<cpu_arch>-alpine-linux-musl`, with `<cpu_arch>` replaced with respective CPU platform, e.g. `aarch64`.
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
CMake's bootstrapping script can't infer a number of parallel build jobs from the number of CPU cores on its own, which means that it's not parallelized by default. Let's get the value of CPU cores when `build_jobs` argument is not explicitly set.
`make -j` only has an effect on CMake's post-bootstrap build jobs, while a substantial amount of time was spent building code sequentially in CMake's bootstrap script without that `--parallelize` option, which is called on a line preceding the `make -j` invocation.
