To get a fully UTF-8 compatible system, getting proper support in ncurses is crucial. However, UTF-8 enabled (configure --enable-widec) ncurses libs are binary incompatible with normally compiled ones. Since many packages depend on ncurses, this is a bit of a problem. As far as I have checked, many core utils (bash, gcc etc.) do list ncurses in RDEPEND, but do not use the shared lib itself. (Are they statically compiled?) BTW, what about texinfo and readline? They use ncurses too, don't they? First, it would be nice to know if we actually need to have ncurses.so to run emerge. If not, with a custom ebuild for the unicode-enabled ncurses, the system could be rebuilt for UTF-8 support. However, if we need it, there is no other way to switch than to bootstrap the system again, with the unicode-enabled ncurses. I believe this would call for an alternative bootstrap script, which would also come in handy for new installations that are intended to be unicode-based. I know this breaks lots of things, but I need unicode, and I think I'm not the only one. :) First, I would like to get some info about how much the base system needs ncurses - I mean, can I just replace it with a binary incompatible, UTF-8 enabled version, and start reemerging everything that depends on it? If so, I will create an ebuild for the unicode ncurses install, and post it here if it works for me. :) Second, I propose a creation of a modified bootstrap script. Since I'm not too familiar with the inner workings of the Gentoo base system, so I really don't have the courage to hack away at it. :)
Shouldn't we mark this bug as duplicate of bug #27700 or bug #20006 ?
Komachi: Yes, I think so... I really should learn to search these bugs better, I didn`t find those at all... BTW, is there some sort of mailing list or repository, where I can get in touch with Unicode related stuff? Like, development info, patches for ebuilds and the like...
Created attachment 27525 [details, diff] Patch for ebuild ncurses-5.4-r1 I adopted Zhen Lin's code from his unicode-enabled ncurses-5.3-r1 ebuild in bug 27700, modifying it to IUSE "unicode" instead of "utf8". It makes symlinks to the *w.so libs, to maintain compile time compatibility. I also added Aethyr's gpm support patch from bug 22005. I have emerged it recently, and everything seems to work fine so far... ^^
azarah: could you add the ebuild to CVS tree? --enable-widec works on my machine these two months. AFAIC there are two issues with --enable-widec. First, centericq is known to be broken with --enable-widec ncurses (there was a post about it to gentoo-dev list a while ago). Second, --enable-widec installs libncursesw.so and some programs will depend on libncursesw.so instead of libncurses.so if libncursesw.so is present, so you cannot switch off unicode USE flag once you enabled it unless you compile all ebuilds depending on libncursesw.so. (perhaps this isn't a big issue as we need to rebuild packages upon shlib manor/major version change) I think unicode USE flag for ncurses package deserves at least hard-masked ebuild.
I'm not able to find the referenced posting to gentoo-dev regarding centericq. However, there was a bug-fix that I made in August 2003 that affected centericq. Perhaps that was the same issue. If so, it's been fixed for some time.
*** This bug has been marked as a duplicate of 20006 ***