Hi Dimitry, On Tue, 22 Jul 2014, Dimitry Andric wrote: > On 22 Jul 2014, at 00:34, Benjamin Kaduk <kaduk_at_MIT.EDU> wrote: >> Building some out-of-tree software with a rather long set of compiler flags, I can reliably get our clang to crash. >> The system is current as of r267362 (June 11), with clang reporting itself as FreeBSD clang version 3.4.1 (tags/RELEASE_34/dot1-final 208032) 20140512 >> Target: x86_64-unknown-freebsd11.0 >> (freefall's clang crashes as well.) >> >> Unfortunately, I don't have debug symbols around for that clang binary. >> If someone does have a clang with debug symbols handy, I'd be interested in seeing the backtrace. >> >> The processed source file and invocation shell script may be found at: >> http://web.mit.edu/kaduk/Public/clang/dumptool-09e584.c (1.1M) >> http://web.mit.edu/kaduk/Public/clang/dumptool-09e584.sh > > Hi Ben, > > I was able to reproduce the assertion, even with clang trunk. It is > apparently caused by the implementation of the -Wconsumed warning, which > is enabled in your case because you had used -Weverything. > > I have reported an upstream bug with a reduced test case here: > > http://llvm.org/PR20402 Thank you very much! I was planning to try to reproduce on clang trunk and reduce the test case, but I think you are much more practiced at doing so than I am ;) > Meanwhile, you can work around the assertion by adding -Wno-consumed to > your compilation flags. I will do so, thanks. As you may have guessed from the compiler command line, I'm enabled -Weverything for development and am disabling the pieces that cause annoyance and/or too much spew. The openafs build system interacts in an exciting way with bsd.kmod.mk, passing through my -Weverything and also taking the kernel's -Werror... -BenReceived on Tue Jul 22 2014 - 17:06:03 UTC
This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:40:51 UTC