[LLVMdev] non-x86 sanitizer buildbots: no rule to make target check-lsan etc.

[Evgeniy Stepanov]

I think check-all-compiler-rt target would be useful on its own, but on a
buildbot it would result in a single "step" instead of multiple smaller
"steps", and even less informative messages. I don't really mind, "run
sanitizer tests in gcc build" is a single step anyway. I would not want
that to happen to the following "check-asan", "check-msan", etc steps
though.

With ninja, we could filter against "ninja -t targets".



On Mon Dec 01 2014 at 3:15:27 PM Jay Foad <jay.foad at gmail.com> wrote:

> Hi,
>
> Currently the first stage ("run sanitizer tests in gcc build") of the
> sanitizer-ppc64-linux1 buildbot is only failing because of:
>
> + cd clang_build
> + make -j16 check-lsan
> make: *** No rule to make target `check-lsan'.  Stop.
> + echo @@@STEP_FAILURE@@@
> @@@STEP_FAILURE@@@
> + cd clang_build
> + make -j16 check-msan
> make: *** No rule to make target `check-msan'.  Stop.
> + echo @@@STEP_FAILURE@@@
> @@@STEP_FAILURE@@@
> + cd clang_build
> + make -j16 check-tsan
> make: *** No rule to make target `check-tsan'.  Stop.
> + echo @@@STEP_FAILURE@@@
> @@@STEP_FAILURE@@@
> + cd clang_build
> + make -j16 check-ubsan
> make: *** No rule to make target `check-ubsan'.  Stop.
> + echo @@@STEP_WARNINGS@@@
> @@@STEP_WARNINGS@@@
> + cd clang_build
> + make -j16 check-dfsan
> make: *** No rule to make target `check-dfsan'.  Stop.
> + echo @@@STEP_WARNINGS@@@
> @@@STEP_WARNINGS@@@
>
> The problem is that the buildbot script
> (zorg/buildbot/builders/sanitizers/buildbot_cmake.sh) explicitly lists
> all the sanitizer check-* targets, but only some of them are supported
> on non-x86 configurations. It seems silly for the buildbot script to
> duplicate all the logic from compiler-rt's cmake/config-ix.cmake about
> which sanitizers are supported on which targets.
>
> Maybe compiler-rt could have a new check-all-compiler-rt target, which
> runs the checks for all supported sanitizers? Is it possible to
> implement this (it would be a lit target that subsumes a bunch of
> other lit targets)?
>
> Any other ideas?
>
> Thanks,
> Jay.
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20141201/a4017e08/attachment.html>