Hello, On Tue, Feb 09, 2016 at 03:09:09PM -0500, Aaron M. Ucko wrote: > > Automated builds of simgrid generally failed:
Yes, that hurts. It's all red! pitty. > ERRORCould not figure the stack direction. > -- LTO seems usable. > -- Add CFLAGS: "-g -O2 -fstack-protector-strong -Wformat > -Werror=format-security -O3" to CMAKE_C_FLAGS > [...] > /«PKGBUILDDIR»/src/simix/smx_context.c:142:16: error: 'PTH_STACKGROWTH' > undeclared (first use in this function) > > It's not entirely clear from the public logs what went wrong, but I > suspect CMake may have tried to build prog_stackgrowth.c with -O3, > which turns on -finline-functions, breaking the test. Parallel building does not help here, for sure. And the diagnosis further is complicated by the fact that my cmake output was broken when the stack direction detection fails (I'm in the upstream team). I just commited a fix for that message, so that we get more info. That is why the build starts even after the line "ERRORCould not figure the stack direction", sorry. Your analysis would be correct: these flags clearly break the stack test. That is why they get added at the end of the configuration, once all tests are ran. So, this is not it. > The fallback > logic didn't help even on the limited set of processor types it tried > to cover because CMAKE_SYSTEM_PROCESSOR corresponds to the output of > uname -p (typically "unknown" on Debian) rather than uname -m. I would say that the fallback works, as it managed to compile on i386 ;) But this is not an option, I'll investiguate further. Right now, I forgot my laptop adapter at lab, and I will soon be offline... If you want to give it a spin, the modified version is in the package git. > Could you please take a look? Many thanks to you for investigating so far. Bye, Mt. -- It is not in improving the candle technology that electricty was discovered, understood and mastered.
signature.asc
Description: PGP signature
