I recently trie to switch from XFree86 to X.org's X server. I had 4.3.0-r4 installed, which whas working nicely on my Compaq Presario x1010us laptop. I 'emerge unmerge'd xfree to get rid of the blocks on xorg-x11, then emerged xorg-x11. I did not change my XF86Config, which xorg-x11 apparently uses if it can't find an xorg.conf. X.org's Xserver wouldn't start because it couldn't load the radeon driver (the laptop has a Radeon Mobility 9200) because of unresolved symbols (related to framebuffer stuff; the symbol it bails on is fbdevHWValidMode); it also can't find the synaptics driver. My attempt at a quick fix was to unmerge xorg-x11, but I'd sync'd and by that time, xfree-4.3.0-r4 had been removed from the Portage tree. 4.3.0-r5 won't start after building; it complains about a doubly-defined symbol in /usr/X11R6/lib/modules/fonts/libbitmap.a. I'm running kernel 2.6.6 vanilla (SELinux enabled). Attached are the XFree86 log and the X.org Xserver log. Unfortunately, I spaced emerge info and my kernel config; I'll attach them separately later. BTW, when I edit the xorg-x11 ebuild file to include the line: #define LinuxFBDevSupport YES in the generated host.def file and add LoadModule "fbdevhw" to the X config file, the compaints about missing FB-related symbols go away. However, it then complains that gRADEONEntityIndex is undefined, even though it appears to be defined in radeon_probe.c. I will try to attach more info about my system later (assuming I can access this form w/ lynx, since I have no windowing system now).
Created attachment 31719 [details] /var/log/XFree86.0.log This is what the re-installed XFree86 complains about when I run startx.
Created attachment 31720 [details] /var/log/Xorg.0.log This is the error message from X.org's Xserver, before I hacked the ebuild file to get rid of the FBDev-related errors.
Created attachment 31772 [details] emerge info results of "emerge info"
Created attachment 31773 [details] linux .config Linux configuration, kernel 2.6.6
I'm starting to think that the problem rebuild-ing the XF86 server is due to a binutils/gcc bug or issue: I got the 4.3.0-r4 ebuild from gentoo.org CVS, and that build but refuses to start w/ the same double symbol definition error as the -r5 build ("Duplicate symbol __i686.get_pc_thunk.bx in /usr/X11R6/lib/modules/fonts/libbitmap.a:bitmapmod.o"). The symbol looks like a toolchain-generated one. I recently updated gcc to 3.3.3-r5, the only other change I remember making which could have effected this. I'm updating binutils and trying to emerge xfree again. Will report results.
So, no luck with updating binutils. Turns out I'm using gcc-3.3.3-r3 (there is no -r5). Tonight or tommorrow I will look in /usr/tmp/portage and figure out what version of gcc I was using before I updated. I still get the duplicate symbol errors. It turns out that every individual object file that makes up libbitmap.a has __i686.get_pc_thunk.bx defined; 'nm libbitmap.a' shows a line like: 00000000 T __i686.get_pc_thunk.bx for each of them. Could this be as simple as a missing #ifndef in some header file that only broke w/ the new gcc? Anyway, I will keep digging, but as I have no UI hopefully someone there can provide some extra insight soon :-)
Also found: http://gcc.gnu.org/ml/gcc-patches/2003-09/msg00173.html while searching Google, and more like it; they all have to do w/ using a non-GNU linker or a broken version of GAS, though.
*** This bug has been marked as a duplicate of 43177 ***
