I fear my patch from bug 145277 comment 12 isn't enough. i686-pc-linux-gnu-gcc -march=prescott -O2 gdb -O2 -pedantic -Wall -fPIC -fno-strict-aliasing -I/usr/include -DGMT_SHARE_PATH=\"/usr/share/gmt-4.5.0\" -DTRIANGLE_D -c -o pslib.o pslib.c i686-pc-linux-gnu-gcc: gdb: No such file or directory make[1]: *** [pslib.o] Error 1 make[1]: Leaving directory `/var/tmp/portage/sci-geosciences/gmt-4.5.0-r1/work/GMT4.5.0/src' make: *** [suppl] Error 2 make: *** Waiting for unfinished jobs.... make[1]: *** [pslib.o] Error 1 make[1]: Leaving directory `/var/tmp/portage/sci-geosciences/gmt-4.5.0-r1/work/GMT4.5.0/src' make: *** [gmt] Error 2 * * ERROR: sci-geosciences/gmt-4.5.0-r1 failed. * Call stack: * ebuild.sh, line 49: Called src_compile * environment, line 2701: Called die * The specific snippet of code: * emake gmt ${mymake} || die "Make ${mymake} failed." * The die message: * Make suppl failed. The reason is that configure does CFLAGS=`echo $CFLAGS | sed 's/-g//'` but I have CFLAGS="-march=prescott -O2 -ggdb". There seems to be an awful lot of flags hackery in that configure script. Makes me think of bug #257489 where similar hacks were used upstream, and dropped at the distro level. I'll attach a more rigorous patch weeding more of these hacks.
Created attachment 199284 [details, diff] Drop manipulation of several flags This patch is a replacement for the current no-strip patch. It now also contains a verbose header for later reference. Maybe I'll even try to get upstream interested in this thing.
Created attachment 199288 [details, diff] Drop manipulation of several flags I had created that previous version against the configure script created by the autoreconf called by the ebuild. It failed to apply against the configure script shipped with the sources. This one here works for me. I know I probably should be testing my patches more thoroughly before submitting them here...
(In reply to comment #1) > Maybe I'll even try to get upstream interested in this thing. Upstream discussion: http://thread.gmane.org/gmane.comp.gis.gmt.user/11760
Patch available for well over four months now. Any reason to not include it in portage?
+*gmt-4.5.0-r2 (24 Jun 2010) + + 24 Jun 2010; Justin Lecher <jlec@gentoo.org> gmt-4.0.ebuild, + gmt-4.1.1.ebuild, -gmt-4.5.0-r1.ebuild, +gmt-4.5.0-r2.ebuild, + files/gmt-4.5.0-no-strip.patch: + Correct handling of *FLAGS, thanks Martin von Gagern for providing the + patch +
Just filed bug 670226 reporting a regression on this for gmt-4.5.18.