The addition of `addDeclErr` introduced a memory leak at every call
site, and I also would like to push back on having more than 1
compilation error per `Decl`.
This reverts commit 1634d45f1d53c8d7bfefa56ab4d2fa4cc8218b6d.
Previously Zig would need to recompile runtime libs if you changed the
values of --strip or -O. Now, unless the `debug_compiler_runtime_libs`
flag is set (which is currently not exposed to the CLI), Zig will always
choose ReleaseFast or ReleaseSmall for compiler runtime libraries.
When the main application chooses ReleaseFast or ReleaseSmall, that
value is propagated to compiler runtime libraries. Otherwise a decision
is made based on the target, which is currently ReleaseSmall for
freestanding WebAssembly and ReleaseFast for everything else.
Ultimately the purpose of this commit is to have Debug and ReleaseSafe
builds of applications still get optimized builds of, e.g. libcxx and
libunwind, as well as to spend less time unnecessarily rebuilding compiler
runtime libraries.
* rename is_compiler_rt_or_libc to skip_linker_dependencies
and set it to `true` for all sub-Compilations. I believe
this resolves the deadlock we were experiencing on Drone
CI and on some users' computers. I will remove the CI workaround in
a follow-up commit.
* enabling TSAN automatically causes the Compilation to link against
libc++ even if not requested, because TSAN depends on libc++.
* add -fno-rtti flags where appropriate when building TSAN objects.
Thanks Firefox317 for pointing this out.
* TSAN support: resolve all the undefined symbols. We are still seeing
a dependency on __gcc_personality_v0 but will resolve this one in a
follow-up commit.
* static libs do not try to build libc++ or libc++abi.
* split std.ResetEvent into:
- ResetEvent - requires init() at runtime and it can fail. Also
requires deinit().
- StaticResetEvent - can be statically initialized and requires no
deinitialization. Initialization cannot fail.
* the POSIX sem_t implementation can in fact fail on initialization
because it is allowed to be implemented as a file descriptor.
* Completely define, clarify, and explain in detail the semantics of
these APIs. Remove the `isSet` function.
* `ResetEvent.timedWait` returns an enum instead of a possible error.
* `ResetEvent.init` takes a pointer to the ResetEvent instead of
returning a copy.
* On Darwin, `ResetEvent` is implemented using Grand Central Dispatch,
which is exposed by libSystem.
stage2 changes:
* ThreadPool: use a single, pre-initialized `ResetEvent` per worker.
* WaitGroup: now requires init() and deinit() and init() can fail.
- Add a `reset` function.
- Compilation initializes one for the work queue in creation and
re-uses it for every update.
- Rename `stop` to `finish`.
- Simplify the implementation based on the usage pattern.
We generally get away with atomic primitives, however a lock is required
around the refresh function since it traverses the Node graph, and we
need to be sure no references to Nodes remain after end() is called.
std.crypto.random
* cross platform, even freestanding
* can't fail. on initialization for some systems requires calling
os.getrandom(), in which case there are rare but theoretically
possible errors. The code panics in these cases, however the
application may choose to override the default seed function and then
handle the failure another way.
* thread-safe
* supports the full Random interface
* cryptographically secure
* no syscall required to initialize on Linux (AT_RANDOM)
* calls arc4random on systems that support it
`std.crypto.randomBytes` is removed in favor of `std.crypto.random.bytes`.
I moved some of the Random implementations into their own files in the
interest of organization.
stage2 no longer requires passing a RNG; instead it uses this API.
Closes#6704
The steps to repro this issue are:
zig build-obj hello.zig -target x86_64-windows-msvc
zig build-exe hello.obj -target x86_64-windows-msvc --subsystem console
-lkernel32 -lntdll
What was happening is that the main Compilation added a work item to
produce kernel32.lib. Then it added a sub-Compilation to build zig's
libc, which ended up calling a function with extern "kernel32", which
caused the sub-Compilation to also try to produce kernel32.lib. The main
Compilation and sub-Compilation do not coordinate about the set of
import libraries that they will be trying to build, so this caused a
deadlock.
This commit solves the problem by disabling the extern "foo" feature
from working when building compiler_rt or libc. Zig's linker code is now
responsible for putting the appropriate import libs on the linker line,
if any for compiler_rt and libc.
Related: #5825
Cache exposes BinDigest.
Compilation gains a set of a BinDigest for every C/C++ source file. We
detect when the same source/flags have already been added and emit a
compile error. This prevents a deadlock in the caching system.
Closes#7308
passthrough mode does not mean always exit - it just means to pass
through stdio and exit if the child process exits, without doing any
special error reporting.
Previously, when mixing Zig and C/C++ code for windows-gnu targets, zig
would get codeview format but the C/C++ code would not get any debug
info. Now, C/C++ code properly emits debug info in codeview format and
everything just works.
Use case:
zig build-exe non_existent_file.zig
Previous behavior:
error.FileNotFound, followed by an error return trace
Behavior after this commit:
error: unable to read non_existent_file.zig: FileNotFound
(end of stderr, exit code 1)
This turns AllErrors.Message into a tagged union which now has the
capability to represent both "plain" errors as well as source-based
errors (with file, line, column, byte offset). The "no entry point found"
error has moved to be a plain error message.
This commit version-gates appending `-syslibroot` flag to lld. This
is predicated upon the fact that for versions of macOS lower than
11, lld would fail to find and link against frameworks with this
flag specified.
Co-authored-by: Andrew Kelley <andrew@ziglang.org>
If we enter the `if` because `comp.disable_c_depfile` is false the `man`
object has no manifest and calling .final on it will trip an assertion.
Closes#7096
See #5854
Some tiny tweaks too:
* Use `wasm-freestanding-musl` instead of `wasm32-freestanding-musl`,
making it pointer-size-agnostic.
* Fix trying to build non-existent wasm musl start files.
Positional shared library arguments were not being detected as causing
dynamic linking, resulting in invalid linker lines. LLD did not have an
error message for this when targeting x86_64-linux but it did emit an
error message when targeting aarch64-linux, which is how I noticed the
problem.
This surfaced an error having to do with fifo.pipe() in the cat example
which I did not diagnose but solved the issue by doing the revamp that
was already overdue for that example.
It appears that the zig-window project was exploiting the previous
behavior for it to function properly, so this prompts the question, is
there some kind of static/dynamic executable hybrid that the compiler
should recognize? Unclear - but we can discuss that in #7240.
* it is now -fcompiler-rt and -fno-compiler-rt to override the (quite
reasonable) default of bundling compiler-rt only for executables and
dynamic libraries.
- the build.zig API is still called bundle_compiler_rt however it is
now an optional bool instead of a bool. leaving it as `null` means
to use the compiler default.
* renamed some internal identifiers to make the source more readable
* additionally support -fcompiler-rt when doing build-obj for ELF files
since that target already supports linking multiple objects into one.
- includes an error message when attempting this for non-ELF. in the
future this could additionally be supported with a more advanced
implementation that does not rely on the linker.
* properly populate the linker cache hash
It is now possible to force linking with system linker `ld` instead
of the LLVM `lld` linker when building natively on the target. This
can be done at each stage by specifying `--system-linker-hack` flag,
and can be useful on platforms where `lld` fails to operate properly
such as macOS 11 Big Sur on ARM64 where every binary/dylib is expected
to be codesigned.
Some example invocations for each stage of compilation of Zig
toolchain:
```
cmake .. -DCMAKE_PREFIX_PATH=/path/to/llvm -DSYSTEM_LINKER_HACK=1
```
```
build/zig build test --system-linker-hack
```
```
build/zig build --prefix $(pwd)/stage2 -Denable-llvm
--system-linker-hack
```
```
build/zig build-exe hello.zig --system-linker-hack
```
This commit fixes linking issue on macOS 11 BigSur by appending
a prefix path to all lib and framework search paths known as
`-syslibroot`.
The reason this is needed is that in macOS 11, the system libraries
and frameworks are no longer readily available in the filesystem.
Instead, the new macOS ships with a built-in dynamic linker cache
of all system-provided libraries, and hence, when linking with either
`lld.ld64` or `ld64`, it is required to pass in `-syslibroot [dir]`.
The latter can usually be obtained by invoking `xcrun --show-sdk-path`.
With this commit, Zig will do this automatically when compiling natively
on macOS. However, it also provides a flag `-syslibroot` which can be
used to overwrite the automtically populated value.
To summarise, with this change, the user of Zig is not required to
generate and append their own syslibroot path. Standard invocations
such as `zig build-exe hello.zig` or `zig build` for projects will
work out of the box. The only missing bit is `zig cc` and `zig c++`
since the addition of the `-syslibroot` option would be a mismatch
between the values provided by `clang` itself and Zig's wrapper.
