The latest version of the librep 0.16.1 ebuild removes the dependency on (and binding to) gmp, which causes librep's handling of integers and floats to break, which causes a great deal of breakage, most notably in sawfish. The most obvious bug (to me) was not being able to resize text-mode windows (like xterm), but I noticed a number of other bugs as well (such as the ,help command in sawfish-client not working). Revision 1.8 in CVS <http://cvs.gentoo.org/cgi-bin/viewcvs.cgi/gentoo-x86/dev-libs/librep/librep-0.16.1.ebuild?r1=1.1> introduced these changes, and also removed the optional readline binding. Also strange is that the comment for this patch is 'mark x86 stable', but x86 was stable in revision 1.1. I suggest backing out this revision (with the possible exception of the emake -> make -j1 change).
well hmm.. dunno, it seems this is the only way some ppl can get sawfish to work.
I'm confused. Disabling gmp is the only way some people can get sawfish to work? What kind of bugs are they seeing when it's enabled? And what about the thing with readline? I use sawfish-client; having readline enabled is nice. If people don't want it, isn't that what USE flags are for?
bugs long ago.. forgot what it was exactly about.
You checked in a change nine days ago to fix bugs from long ago that you can't even remember? Does that sound weird to you?
is this nine days ago ? then its a mistake.. ill have a look.
I just started from scratch with 1.4_rc2 a couple of days ago and ran into librep and sawfish weirdness. I tried to emerge sawfish, which wanted librep-0.16.1 and rep-gtk-0.17 (but not gmp). When rep-gtk tries to build, it compiles a couple of files and then gets into a deadlock waiting for some files to be removed. I get the following lines repeated infinitely: Waiting for rep-types.o.lock to be removed Waiting for rep-gtk.o.lock to be removed
Man, this is frustrating. Now, I can't reproduce the deadlock. Anyway, even if I do get sawfish to compile, which I have a couple of times, it doesn't work. It just complains like this: Bad argument: #<closure make-lut>, , 3 Bad argument: #<closure make-lut>, , 3 Bad argument: #<closure make-lut>, , 3
The original link I meant to include was <http://cvs.gentoo.org/cgi-bin/viewcvs.cgi/gentoo-x86/dev-libs/librep/librep-0.16.1.ebuild.diff?r1=1.7&r2=1.8>, which shows the changes I'm talking about. Jonathan, have you tried reversing these changes in the ebuild and rebuilding librep to see if it solves your problems?
i think the current stable version in portage 0.16.1-r1 should fix this screwup, can you guys confirm so i can close ? Sorry for the mess.
Works for me. Thanks.
ok closing, again sorry for the mess up