On Mon, 2013-03-11 at 17:29 +0100, Dimitry Andric wrote: > On 2013-03-11 14:15, Niclas Zeising wrote: > > On 03/11/13 14:13, Steve Kargl wrote: > ... > >> No. Here's my make.conf. > >> > >> KERNCONF=SPEW > >> CPUTYPE?=opteron > >> FFLAGS+= -O2 -pipe -march=native -mtune=native -funroll-loops -ftree-vectorize > >> MALLOC_PRODUCTION="YES" > >> WITHOUT_LIB32="YES" > >> WITHOUT_MODULES="YES" > >> WITHOUT_NLS="YES" > >> WITH_BSD_GREP="YES" > >> WITH_PROFILE="YES" > >> WITH_PKGNG=yes > >> PRINTERDEVICE=ps > >> # > >> # Crap for ports. > >> # > >> DISABLE_MAKE_JOBS="YES" > >> WITH_GHOSTSCRIPT_VER=8 > >> # > >> # added by use.perl 2013-02-19 12:45:06 > >> PERL_VERSION=5.12.4 > >> > > > > This is most likely due to a incompatibility between bsd grep and gnu > > grep. Try to switch to gnu grep, and the problem will most likely go away. > > Yes, this is definitely due to a BSD grep bug. The depcomp tests > create a file sub/conftest.Po, containing: > > ======================================================================== > sub/conftest.o: sub/conftest.c sub/conftst1.h sub/conftst2.h \ > sub/conftst3.h sub/conftst4.h sub/conftst5.h sub/conftst6.h > > sub/conftst1.h: > > sub/conftst2.h: > > sub/conftst3.h: > > sub/conftst4.h: > > sub/conftst5.h: > > sub/conftst6.h: > ======================================================================== > > Then it runs "grep sub/conftest.o sub/conftest.Po", which fails with BSD > grep, and succeeds with GNU grep. > > BSD grep does something very strange here: > > $ echo 'foo.bar' | grep foo.bar > foo.bar > $ echo 'foo.barx' | grep foo.bar > foo.barx > $ echo 'sub/foo.bar' | grep sub/foo.bar > sub/foo.bar > $ echo 'sub/foo.barx' | grep sub/foo.bar > $ echo $? > 1 > > So why does it not match in the last case? GNU grep works: > > $ echo 'sub/foo.barx' | gnugrep sub/foo.bar > sub/foo.barx After disabling WITH_BSD_GREP and rebuild of the system, it seems that the machines in question now build lang/gcc.
This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:40:35 UTC