On 01/21/2016 03:46 AM, Chris Sherlock wrote:
I think that something odd is happening in the compilation of the boost library, I’m getting an error on Jenkins that there are too many failures in the compilation process.
Yes, multiple -Werror,-Wunknown-pragmas caused by my <http://cgit.freedesktop.org/libreoffice/core/commit/?id=2c9d652dc7f7a1e2d0d8ea0cfcdcf47608a3d84e> "Does that help old GCC that don't know -Wplacement-new." (And it is normal behavior that the compiler quits with a "too many errors" message when it has encountered a large amount of errors in a compilation, assuming something is so broken with that input that it doesn't make much sense to present ever more error messages to the user.) This problem should be fixed with <http://cgit.freedesktop.org/libreoffice/core/commit/?id=dda3fc05a2900539836f9748d418dc56247a2ddc> "Fix for Jenkins Gerrit Mac builds."
What is worrying is that this problem only showed up in Jenkins Gerrit builds, not in Jenkins tinderbox builds. The two Jenkins Mac tinderboxes on master, Jenkins_MacOSX and Jenkins_MacOSX_Dbg, stayed green after 2c9d652dc7f7a1e2d0d8ea0cfcdcf47608a3d84e.
Quoting IRC: "<sberg> shm_get_, cloph, Jenkins_MacOSX tinderbox is a mystery to me; in <http://tinderbox.libreoffice.org/cgi-bin/gunzip.cgi?tree=MASTER&full-log=1453200807.490> build it broke when compiling sal/cppunittester/cppunittester.cxx; according to <http://tinderbox.libreoffice.org/MASTER/status.html>, the next build of that tb was <http://tinderbox.libreoffice.org/cgi-bin/gunzip.cgi?tree=MASTER&full-log=1453203566.10951>, and that doesn't mention building sal/cppunittester/cppunittester.cxx at all, and succeeds; how can that be?"
(I didn't notice any problems with my local Mac builds, as I only do them with a trunk Clang, which does know the offending pragmas, The only other Mac tinderbox on master, MacOSX_x86_64_49-TDF operated by cloph, appears to be down and hasn't reported any builds in quite some time now.)