Attempting to emerge graphviz on OS X (gentoo-alt prefix) leads to the following error. Some part of the script attempts to move and remove dynamic library files using the wrong naming convention (.so instead of .dylib). The .dylib files exist, the .so files do not. (cd /gentoo/var/tmp/portage/media-gfx/graphviz-2.12/image//gentoo/usr/lib/ruby/site_ruby/1.8/i686-darwin8 && { ln -s -f libgv_ruby.0.0.0.dylib libgv_ruby.dylib || { rm -f libgv_ruby.dylib && ln -s libgv_ruby.0.0.0.dylib libgv_ruby.dylib; }; }) /gentoo/usr/bin/install -c .libs/libgv_ruby.lai /gentoo/var/tmp/portage/media-gfx/graphviz-2.12/image//gentoo/usr/lib/ruby/site_ruby/1.8/i686-darwin8/libgv_ruby.la libtool: install: warning: remember to run `libtool --finish /gentoo/usr/lib/ruby/site_ruby/1.8/i686-darwin8' make install-data-hook make[4]: Entering directory `/gentoo/var/tmp/portage/media-gfx/graphviz-2.12/work/graphviz-2.12/tclpkg/gv' (cd /gentoo/var/tmp/portage/media-gfx/graphviz-2.12/image//gentoo/usr/lib/perl5/vendor_perl/5.8.8/darwin-2level; rm -f gv.so libgv_perl.so libgv_perl.so.0 libgv_perl.la; mv libgv_perl.so.0.0.0 gv.so;) mv: cannot stat `libgv_perl.so.0.0.0': No such file or directory make[4]: *** [install-data-hook] Error 1 make[4]: Leaving directory `/gentoo/var/tmp/portage/media-gfx/graphviz-2.12/work/graphviz-2.12/tclpkg/gv' make[3]: *** [install-data-am] Error 2 make[3]: Leaving directory `/gentoo/var/tmp/portage/media-gfx/graphviz-2.12/work/graphviz-2.12/tclpkg/gv' make[2]: *** [install-am] Error 2 make[2]: Leaving directory `/gentoo/var/tmp/portage/media-gfx/graphviz-2.12/work/graphviz-2.12/tclpkg/gv' make[1]: *** [install-recursive] Error 1 make[1]: Leaving directory `/gentoo/var/tmp/portage/media-gfx/graphviz-2.12/work/graphviz-2.12/tclpkg' make: *** [install-recursive] Error 1 * * ERROR: media-gfx/graphviz-2.12 failed. * Call stack: * ebuild.sh, line 1595: Called dyn_install * ebuild.sh, line 1030: Called qa_call 'src_install' * ebuild.sh, line 44: Called src_install * graphviz-2.12.ebuild, line 178: Called die * * emake install failed * If you need support, post the topmost build error, and the call stack if relevant. * A complete build log is located at '/gentoo/var/tmp/portage/media-gfx/graphviz-2.12/temp/build.log'. * Reproducible: Always Steps to Reproduce: 1. emerge graphviz
Thanks, this has to do with building one of the extentions (truggered by USE flags). We're looking into how to fix this the best way.
More info: The bug is indeed related to the use of the "perl" USE-flag. An analogous error is produced for the "ruby" flag. Emerging with neither of these two flags works. Cheers, Burnce
This issue has been solved upstream in v2.14.1. Note that USE=python will not work even with v2.14.1, because there's no shared library for -lpython2.5 to link against. @grobian: please make this bug depend on bug #194749, #192606
graphviz 2.14.1 will still need a minor patch, because one ends up with broken symlinks: perl/gv.so -> libgv_perl.so python/_gv.so -> libgv_python.so ruby/gv.so -> libgv_ruby.so tcl/gv.so -> libgv_tcl.so the rest is fine and the above might not even matter, though: perl/gv.pm perl/gv.so -> libgv_perl.so perl/libgv_perl.0.0.0.dylib perl/libgv_perl.0.dylib -> libgv_perl.0.0.0.dylib perl/libgv_perl.dylib -> libgv_perl.0.0.0.dylib perl/libgv_perl.la
I've committed a fix in r11046. [1] * python bindings still don't work{1} * i cannot test the tcl bindings because tcl does not compile{2} with the patch applied, perl's dir (and it's basically the same for ruby) looks like this: perl5/vendor_perl/5.8.8/darwin-2level/ gv.dylib -> libgv_perl.dylib gv.pm libgv_perl.0.0.0.dylib libgv_perl.0.dylib -> libgv_perl.0.0.0.dylib libgv_perl.dylib -> libgv_perl.0.0.0.dylib libgv_perl.la all of the dynamic libraries are linked against this file: ${EPREFIX}usr/lib/perl5/vendor_perl/5.8.8/darwin-2level/libgv_perl.0.dylib That's why doing exactly what the original patch does, as you can see here[2] is not a good idea: > rm -f gv.so libgv_perl.so libgv_perl.so.0 libgv_perl.la; mv libgv_perl.so.0.0.0 gv.so we could make that > rm -f gv.so libgv_perl.dylib libgv_perl.0.dylib libgv_perl.la; mv libgv_perl.0.0.0.dylib gv.dylib but: * we'd still link against libgv_perl.0.dylib - removing it is therefore not a good idea * why don't we keep the other dylibs around * why don't we keep the .la file around? (looks perfectly fine although i might be wrong) * should the gv.dylib be there at all? do we need it? does it need to go by the name gv.so for some reason? I've not remove the package.use.mask for perl and ruby because i'm not sure if they really *work* now - so both testing and answers are highly appreciated. @grobian: this bug does not depend on bug #194749, sorry {1} ld naturally doesn't find anything for -lpython2.5 because of bug #192606 {2} I should file a bug about that [1] http://overlays.gentoo.org/proj/alt/changeset/11046 [2] http://overlays.gentoo.org/proj/alt/browser/trunk/prefix-overlay/media-gfx/graphviz/files/graphviz-2.12-bindings-prefix.patch?rev=11046
per pipping's comment
* we'd still link against libgv_perl.0.dylib - removing it is therefore not a good idea * why don't we keep the other dylibs around * why don't we keep the .la file around? (looks perfectly fine although i might be wrong) * should the gv.dylib be there at all? do we need it? does it need to go by I don't see the ebuild doing any removing of dylibs, does it?
not the ebuild but the makefile/this patch[1] removes dylibs. [1] http://overlays.gentoo.org/svn/proj/alt/trunk/prefix-overlay/media-gfx/graphviz/files/graphviz-2.12-bindings-prefix.patch
Can someone running OSX please let us know the status of this package? If no comment is, say..2 weeks, I'll assume this package is working "good enough" (based on the comments) and resolve. Thanks.
I think the bug should stay open even if nobody says anything (which is what I expect to happen) -- just my opinion of course.
(In reply to comment #3) > This issue has been solved upstream in v2.14.1. v2.20.2 is in the tree now. (In reply to comment #10) > I think the bug should stay open even if nobody says anything (which is what I > expect to happen) -- just my opinion of course. No one has had any comments about borkdness for 8 months. Assumed this bug is resolved. Please file a new bug if there is any additional problems.
