Restore and verify GCC compatibility #177
Conversation
Makes it easier to skip targets, such as those using FuzzedDataProvider when using gcc as CC.
Using the weak attribute together with an explicitly undefined symbol should still work on macOS, see: https://stackoverflow.com/a/34983229/297261
fmeum
force-pushed
the
gcc
branch
2 times, most recently
from
5eba34e
to
8eead03
Compare
fmeum
requested review from
asraa,
inferno-chromium,
markus-kusano,
oliverchang,
stefanbucur and
xinhaoyuan
as code owners
fmeum
force-pushed
the
gcc
branch
4 times, most recently
from
8c46709
to
0eba32d
Compare
|
@stefanbucur The CI is broken due to a regression in Bazel 4.2.0 which came out a few days ago and is now used by both OSS-Fuzz and the Bazel tests. Don't worry, projects using rules_fuzzing do not seem to be affected. |
|
There will be a 4.2.1 release that fixes the regression, see bazelbuild/bazel#13890. @stefanbucur Could you as the maintainer follow the instructions at bazelbuild/bazel#13890 (comment) to add |
|
Now that Bazel 4.2.1, the pipeline passes. I had to exclude the Java regression tests from the macOS regression test for now, but will enable them once I have understood where to find the required dylibs. All regression tests are now buildable and runnable with gcc and C++ tests are also compatible with macOS. |
Excludes the tests in //examples/java, which require further fixes to find libjvm.dylib.
Implements the missing readlink -f command in Python.
The fix we are pulling in is: bazel-contrib/rules_fuzzing#177
Fixes #176.