On Tue, Sep 15, 2015 at 4:54 PM, Jack Howarth <howarth.at.f...@gmail.com> wrote:
> Martin, > Fortunately, even with '-d' passed to fink make, I was able to > trigger at least one of these failures with the texlive-base build. > > Need a job token; we have children > Live child 0x7fd364b14bc0 (cweave.c) PID 32720 > Live child 0x7fd364b10f70 (synctex) PID 32702 > Live child 0x7fd362518520 > (luatexdir/luafontloader/src/libjitff_a-luafflib.o) PID 32689 > Live child 0x7fd364a0d2a0 > (luatexdir/luafontloader/fontforge/fontforge/libjitff_a-ttfspecial.o) PID > 32663 > Live child 0x7fd364a119d0 > (luatexdir/luafontloader/fontforge/fontforge/libjitff_a-tottf.o) PID 32644 > Live child 0x7fd364912260 > (luatexdir/luafontloader/fontforge/fontforge/libjitff_a-splineutil.o) PID > 32601 > Live child 0x7fd36490ee60 > (luatexdir/luafontloader/fontforge/fontforge/libjitff_a-psread.o) PID 32491 > Live child 0x7fd364b0f8f0 > (luatexdir/luafontloader/fontforge/fontforge/libjitff_a-parsettf.o) PID > 32473 > Reaping winning child 0x7fd364b14bc0 PID 32720 > Removing child 0x7fd364b14bc0 PID 32720 from chain. > Reaping losing child 0x7fe93ba55e90 PID 30489 > Makefile:13666: recipe for target 'all-recursive' failed > make[4]: *** [all-recursive] Error 1 > Removing child 0x7fe93ba55e90 PID 30489 from chain. > make[4]: Leaving directory > '/sw/src/fink.build/texlive-0.20140523-2/texlive/Work/texk/web2c' > Reaping losing child 0x7f820970c880 PID 30488 > Makefile:3580: recipe for target 'all' failed > make[3]: *** [all] Error 2 > Removing child 0x7f820970c880 PID 30488 from chain. > libtool: link: gcc -Wimplicit -Wreturn-type -Wdeclaration-after-statement > -Wno-unknown-pragmas -g -O2 -o synctex synctexdir/synctex-synctex_main.o > synctexdir/synctex-synctex_parser.o > synctexdir/synctex-synctex_parser_utils.o -L/sw/lib -lz > make[3]: Leaving directory > '/sw/src/fink.build/texlive-0.20140523-2/texlive/Work/texk/web2c' > Reaping losing child 0x7f91b3425400 PID 20021 > Makefile:878: recipe for target 'recurse' failed > make[2]: *** [recurse] Error 1 > Removing child 0x7f91b3425400 PID 20021 from chain. > make[2]: Leaving directory > '/sw/src/fink.build/texlive-0.20140523-2/texlive/Work/texk' > Reaping losing child 0x7fb98bc2d7a0 PID 20014 > Makefile:458: recipe for target 'all-recursive' failed > make[1]: *** [all-recursive] Error 1 > Removing child 0x7fb98bc2d7a0 PID 20014 from chain. > make[1]: Leaving directory > '/sw/src/fink.build/texlive-0.20140523-2/texlive/Work/texk' > Reaping losing child 0x7fa0cb5275f0 PID 66196 > Makefile:544: recipe for target 'all-recursive' failed > make: *** [all-recursive] Error 1 > Removing child 0x7fa0cb5275f0 PID 66196 from chain. > make: INTERNAL: Exiting with 1 jobserver tokens available; should be 8! > ### execution of /tmp/fink.EEyh_ failed, exit code 2 > ### execution of /tmp/fink.CtwYk failed, exit code 2 > I should also note that with '-d', the texlive-base fink build only randomly fails as opposed to consistently fails without it. Definitely this seems to be some threading race condition issue either exposed by or caused by 10.11 in GNU make 4.1. > Searching through all of the rest of the build logs I generated searching > for this failing test case using '-d'. this is the only one with instances > of 'Reaping losing child'. > > Jack > > On Tue, Sep 15, 2015 at 2:43 PM, Jack Howarth <howarth.at.f...@gmail.com> > wrote: > >> >> >> On Tue, Sep 15, 2015 at 11:45 AM, Jack Howarth <howarth.at.f...@gmail.com >> > wrote: >> >>> >>> >>> On Tue, Sep 15, 2015 at 11:04 AM, Martin Costabel <costa...@wanadoo.fr> >>> wrote: >>> >>>> On 15/09/15 14:50, Jack Howarth wrote: >>>> >>>>> Martin, >>>>> In libcurl4, the error always occurs in make install as... >>>>> >>>>> Makefile:481: recipe for target 'all-recursive' failed >>>>> >>>> [...] >>>> >>>> I don't have 10.11 yet, so I cannot look myself, but I still think >>>> there must have been an error just before this line. >>>> The question is where to look for a more detailed error message. >>> >>> >>> I can reproduce the build failure with 100% reproducibility on a MacPro >>> 3,1 with 8 cores. However if I change the cmake.info to have... >>> >>> make -d >>> >>> >> Jeremy agrees that since this seems to be a time-sensitive threading >> issue in GNU make it will be difficult to debug with the -d option. I am in >> the process of rebuilding chunks of the package set with Fink modified to >> pass -d with -j and with fink make installed. I really don't expect this to >> be able to replicate the bug since the act of trying to observe it with -d >> likely spoils the timings needed to reproduce it at all. >> >> >>> the build completes without issues. This smells like a race condition >>> in the pthread support or such. We might have to try a test fink where >>> MaxBuildJobs also defaults make to -d in order to capture a failing build >>> in verbose mode. Although -d may shift the timing in make around such that >>> the race condition never triggers at all. >>> Jack >>> >>> >>> >>>> -- >>>> Martin >>>> >>>> >>>> >>> >> >
------------------------------------------------------------------------------
_______________________________________________ Fink-devel mailing list Fink-devel@lists.sourceforge.net List archive: http://news.gmane.org/gmane.os.apple.fink.devel Subscription management: https://lists.sourceforge.net/lists/listinfo/fink-devel
