Previously, when choosing the local cache directory, if there was no
root source file, an explicitly chosen path, or other clues, zig would
choose cwd + zig-cache/ as the local cache directory.
This can be problematic if Zig is invoked with the CWD set to a
read-only directory, or a directory unrelated to the actual source files
being compiled. In the real world, we see this when using `zig cc` with
CGo, which for some reason changes the current working directory to the
read-only go standard library path before running the C compiler.
This commit conservatively chooses to use the global cache directory
as the local cache directory when there is no other reasonable choice,
and no longer will rely on the cwd path to choose a local cache directory.
As a reminder, the --cache-dir CLI flag and ZIG_LOCAL_CACHE_DIR
environment variable are available for overriding the decision. For the
zig build system, it will always choose the directory that build.zig is
+ zig-cache/.
Closes#7342
* Field global_cache_root was added to Builder struct along with
mandatory argument for build_runner.zig. Logic for using the custom
global cache was also added.
* The arguments --cache-dir and --global-cache-dir are no longer passed
directly through to build_runner.zig and are instead only passed through the
mandatory cache_root and global_cache_root arguments.
Previously, --name would only be inferred if there was exactly 1 C
source file or exactly 1 object. Now it will be inferred if there is at
least one of either.
This allows to have multiple instances of `zig build` at the same
time. For example when you have a long running `zig build run` and
then want to run `zig build somethingelse`.
This commit adds:
ZIG_LOCAL_CACHE_DIR corresponding to --cache-dir
ZIG_GLOBAL_CACHE_DIR corresponding to --global-cache-dir
ZIG_LIB_DIR corresponding to --override-lib-dir
The main use case is for `zig cc` where we are bound by clang's CLI
options and need alternate channels to pass these configuration options.
* Remove function parseAnyBaseInt.
* Replace calls to parseAnyBaseInt to calls to std.fmt.parseUnsigned
with radix 0.
* Replace calls to parseInt where the type is unsigned with calls to
parseUnsigned. Note that the functionality of these arguments haven't
changed, they still retain the original radix specified.
* 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
Before this commit, the branch regressed this case:
zig build-lib foo.zig -dynamic
readelf -d libfoo.so
x000000000000000e (SONAME) Library soname: [libfoo.so]
Now it works again, but it preserves the property that using `zig cc`
has the SONAME off by default to match C compilers.
* stage2: Make zig cc more verbose
Make `zig cc` print more info from Clang itself and from our own linker
invocation, this is needed for CMake to properly discover all the
include directories and library search paths.
Closes#7110
* Update `update_clang_options`
* Typo fixes
Co-authored-by: Jakub Konka <kubkon@jakubkonka.com>
On Darwin, according to the man pages for setrlimit(), when adjusting
max number of open fds, the reported hard max by getrlimit() is only
theoretical, while the actual maximum, set in the kernel, is hardcoded
in the header file. Therefore, the reported max has to be adjusted
as `min(OPEN_MAX, lim.max)`.
Signed-off-by: Jakub Konka <kubkon@jakubkonka.com>
* extract logic into a `os_can_execve` and use it in the other place
that we execve
* outdent some code by introducing `run_or_test` variable
* delete unnecessary and wasteful memory management logic
* better error message for when execve fails
* add comment to explain why we do not execve for `zig test`
