Hello, webkitgtk/webkitgtk3 currently fail to build with parallel make. An investigation upstream [1] lead to a conclusion that it's actually an issue within make, and that make-intermediate-parallel-bug.patch from http://savannah.gnu.org/bugs/?30653 fixes it. Would it be possible to apply this patch to rawhide, please? webkitgtk is a huge library and using parallel make would cut down the build time from ~6h to less than 2h, making its packaging much easier to handle. [1] https://bugs.webkit.org/show_bug.cgi?id=79498
This same bug is affecting postgresql. Here is a self-contained test case that fails for me on both F16 and F17: Grab postgresql-9.1.5.tar.bz2 (eg, from current Fedora F17 git) Be sure you have gcc, bison, flex installed tar xfj postgresql-9.1.5.tar.bz2 cd postgresql-9.1.5 ./configure --enable-depend make -j4 -s Building the full tree may or may not fail, but if you then do cd src/interfaces/ecpg you can repeat the following sequence of steps: make -s maintainer-clean rm -rf ~/.ccache/ make -j4 -s and it will fail upwards of 75% of the time with config.status: creating src/interfaces/ecpg/include/ecpg_config.h parser.c:25:21: fatal error: preproc.h: No such file or directory compilation terminated. make[1]: *** [parser.o] Error 1 make[1]: *** Waiting for unfinished jobs.... keywords.c:20:21: fatal error: preproc.h: No such file or directory compilation terminated. make[1]: *** [keywords.o] Error 1 make: *** [all-preproc-recurse] Error 2 make: *** Waiting for unfinished jobs.... Depending on what hardware you try it on, you might need to adjust the -j value a bit (a value near your number of CPUs seems most prone to trigger the problem). Sometimes there are other errors, which I suspect are due to bogosities in the makefiles, but if you look into ecpg/preproc/Makefile I think you will agree there is no way that it should be trying to compile parser.c or keywords.c before it's run bison to create preproc.h. But it does. I have reproduced this with make-3.82-8.fc16 on F16, and with make-3.82-9.fc17 on F17 on different hardware. I rebuilt make with make-intermediate-parallel-bug.patch from the upstream bug report, and that seems to fix it. Dunno if the fix is any good stylistically, but please press upstream to do something about this.
BTW, I forgot to mention that --enable-depend seems to be necessary to trigger the bug in postgresql builds. I am not sure if the modified build rules (see the "autodepend" stanza in src/Makefile.global.in for details) actually have some influence on the bug, or if it's just a matter of getting the timing right.
FYI, make maintainers just applied the patch: https://savannah.gnu.org/bugs/?30653 http://cvs.savannah.gnu.org/viewvc/make/remake.c?root=make&r1=1.153&r2=1.154
Oh, this is good new. Petr, what would you think about backporting the patch to the Fedora package, or would you mind if I do this?
OK, the test case provided in 30653 triggers, and the patch indeed fixes it. I'll add this to impacted Fedora releases.
make-3.82-13.fc18 has been submitted as an update for Fedora 18. https://admin.fedoraproject.org/updates/make-3.82-13.fc18
make-3.82-13.fc17 has been submitted as an update for Fedora 17. https://admin.fedoraproject.org/updates/make-3.82-13.fc17
Package make-3.82-13.fc18: * should fix your issue, * was pushed to the Fedora 18 testing repository, * should be available at your local mirror within two days. Update it with: # su -c 'yum update --enablerepo=updates-testing make-3.82-13.fc18' as soon as you are able to. Please go to the following url: https://admin.fedoraproject.org/updates/FEDORA-2012-13808/make-3.82-13.fc18 then log in and leave karma (feedback).
make-3.82-13.fc18 has been pushed to the Fedora 18 stable repository. If problems still persist, please make note of it in this bug report.
make-3.82-13.fc17 has been pushed to the Fedora 17 stable repository. If problems still persist, please make note of it in this bug report.