[Bug testsuite/86446] 'gmake check-fortran' broken in libgomp
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=86446 Eric Botcazou changed: What|Removed |Added Status|UNCONFIRMED |RESOLVED CC||ebotcazou at gcc dot gnu.org Resolution|--- |INVALID --- Comment #1 from Eric Botcazou --- The correct invocation of a GCC testsuite is "make -k check-blah", otherwise the recursive Make processes will stop on errors.
[Bug testsuite/86446] 'gmake check-fortran' broken in libgomp
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=86446 --- Comment #2 from Steve Kargl --- On Mon, Jul 09, 2018 at 07:20:02PM +, ebotcazou at gcc dot gnu.org wrote: > > The correct invocation of a GCC testsuite is "make -k check-blah", otherwise > the recursive Make processes will stop on errors. > Since when? I've been doing 'gmake check-fortran' and 'gmake -j6 check-fortran' for more than 15 years. There was never an error. This broke recently. I just did a bootstrap of the 8-branch. 'gmake -j6 check-fortran' completes without an error. The last few line are === gfortran Summary === # of expected passes6413 # of expected failures 18 # of unsupported tests 2 /safe/sgk/gcc/obj8/gcc/gfortran version 8.1.1 20180709 (GCC) gmake[2]: Leaving directory '/safe/sgk/gcc/obj8/gcc' gmake[1]: Leaving directory '/safe/sgk/gcc/obj8/gcc' Something is broken with trunk. These do not occur on the 8-branch. gmake[4]: *** [Makefile:306: check-DEJAGNU] Error 1 gmake[4]: Leaving directory '/safe/sgk/gcc/obj/x86_64-unknown-freebsd12.0/libgomp/testsuite' gmake[3]: *** [Makefile:350: check-am] Error 2 gmake[3]: Leaving directory '/safe/sgk/gcc/obj/x86_64-unknown-freebsd12.0/libgomp
[Bug testsuite/86446] 'gmake check-fortran' broken in libgomp
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=86446 --- Comment #3 from Steve Kargl --- On Mon, Jul 09, 2018 at 07:33:48PM +, sgk at troutmask dot apl.washington.edu wrote: > > gmake[4]: *** [Makefile:306: check-DEJAGNU] Error 1 > gmake[4]: Leaving directory > '/safe/sgk/gcc/obj/x86_64-unknown-freebsd12.0/libgomp/testsuite' > gmake[3]: *** [Makefile:350: check-am] Error 2 > gmake[3]: Leaving directory > '/safe/sgk/gcc/obj/x86_64-unknown-freebsd12.0/libgomp > If I comment out lines 306-316 in ./x86_64-unknown-freebsd12.0/libgomp/testsuite/Makefile check-DEJAGNU: site.exp # srcdir='$(srcdir)'; export srcdir; \ # EXPECT=$(EXPECT); export EXPECT; \ # runtest=$(RUNTEST); \ # if $(SHELL) -c "$$runtest --version" > /dev/null 2>&1; then \ # exit_status=0; l='$(DEJATOOL)'; for tool in $$l; do \ # if $$runtest $(AM_RUNTESTFLAGS) $(RUNTESTDEFAULTFLAGS) $(RUNTESTFLAGS); \ # then :; else exit_status=1; fi; \ # done; \ # else echo "WARNING: could not find \`runtest'" 1>&2; :;\ # fi; \ # exit $$exit_status gmake -j6 check-fortran completes with === gfortran Summary === # of expected passes8017 # of expected failures 3 # of unsupported tests 24 /safe/sgk/gcc/obj/gcc/gfortran version 9.0.0 20180709 (experimental) (GCC) gmake[2]: Leaving directory '/safe/sgk/gcc/obj/gcc' gmake[1]: Leaving directory '/safe/sgk/gcc/obj/gcc' as I expect. This check is bogus.
[Bug testsuite/86446] 'gmake check-fortran' broken in libgomp
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=86446 --- Comment #4 from Eric Botcazou --- > Since when? The dawn of time, see https://gcc.gnu.org/contribute.html#testing
[Bug testsuite/86446] 'gmake check-fortran' broken in libgomp
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=86446 --- Comment #5 from Steve Kargl --- On Mon, Jul 09, 2018 at 10:05:23PM +, ebotcazou at gcc dot gnu.org wrote: > https://gcc.gnu.org/bugzilla/show_bug.cgi?id=86446 > > --- Comment #4 from Eric Botcazou --- > > Since when? > > The dawn of time, see https://gcc.gnu.org/contribute.html#testing > 'gmake -j6 check-fortran' has never died on an error like the one I've shown in the 15+ years that I've been contributing to GCC. Needing -k now, means someone has broken the build infrastructure.
[Bug testsuite/86446] 'gmake check-fortran' broken in libgomp
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=86446 --- Comment #6 from Eric Botcazou --- > 'gmake -j6 check-fortran' has never died on an error > like the one I've shown in the 15+ years that I've been > contributing to GCC. Needing -k now, means someone has > broken the build infrastructure. No, it's just https://gcc.gnu.org/ml/gcc-patches/2018-06/msg01023.html
[Bug testsuite/86446] 'gmake check-fortran' broken in libgomp
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=86446 --- Comment #7 from Steve Kargl --- On Mon, Jul 09, 2018 at 10:21:20PM +, ebotcazou at gcc dot gnu.org wrote: > https://gcc.gnu.org/bugzilla/show_bug.cgi?id=86446 > > --- Comment #6 from Eric Botcazou --- > > 'gmake -j6 check-fortran' has never died on an error > > like the one I've shown in the 15+ years that I've been > > contributing to GCC. Needing -k now, means someone has > > broken the build infrastructure. > > No, it's just https://gcc.gnu.org/ml/gcc-patches/2018-06/msg01023.html > So, instead of using -k to pape over an error in the Makefile, what is wrong with the lines check-DEJAGNU: site.exp srcdir='$(srcdir)'; export srcdir; \ EXPECT=$(EXPECT); export EXPECT; \ runtest=$(RUNTEST); \ if $(SHELL) -c "$$runtest --version" > /dev/null 2>&1; then \ exit_status=0; l='$(DEJATOOL)'; for tool in $$l; do \ if $$runtest $(AM_RUNTESTFLAGS) $(RUNTESTDEFAULTFLAGS) $(RUNTESTFLAGS); \ then :; else exit_status=1; fi; \ done; \ else echo "WARNING: could not find \`runtest'" 1>&2; :;\ fi; \ exit $$exit_status If the commands are command out, there are no errors. So, what is the above suppose to do other that cause 2 ERRORs in the build.
[Bug testsuite/86446] 'gmake check-fortran' broken in libgomp
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=86446 --- Comment #8 from Dominique d'Humieres --- > > The correct invocation of a GCC testsuite is "make -k check-blah", otherwise > > the recursive Make processes will stop on errors. > > > > Since when? I've been doing 'gmake check-fortran' and > 'gmake -j6 check-fortran' for more than 15 years. There > was never an error. This broke recently. Due to pr86417, there is a failure in the libgomp tests and make has to be used with -k.
[Bug testsuite/86446] 'gmake check-fortran' broken in libgomp
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=86446 --- Comment #9 from Eric Botcazou --- > So, instead of using -k to pape over an error in the Makefile, > what is wrong with the lines > > check-DEJAGNU: site.exp > srcdir='$(srcdir)'; export srcdir; \ > EXPECT=$(EXPECT); export EXPECT; \ > runtest=$(RUNTEST); \ > if $(SHELL) -c "$$runtest --version" > /dev/null 2>&1; then \ > exit_status=0; l='$(DEJATOOL)'; for tool in $$l; do \ > if $$runtest $(AM_RUNTESTFLAGS) $(RUNTESTDEFAULTFLAGS) > $(RUNTESTFLAGS); > \ > then :; else exit_status=1; fi; \ > done; \ > else echo "WARNING: could not find \`runtest'" 1>&2; :;\ > fi; \ > exit $$exit_status > > If the commands are command out, there are no errors. So, > what is the above suppose to do other that cause 2 ERRORs > in the build. But we are talking about errors in the testsuite, not in the Makefile! Every testsuite of the GNU toolchain (GCC, GDB, Binutils, etc...) must be run the same way with "make -k check" because, otherwise, the recursive Make processes will stop on errors due the way the various testsuites are invoked.
