After system update (xorg-server 1.9.4 to 1.9.5 and gentoo-sources 2.6.37 to 2.6.38) I'm unable to install any version of ati-drivers (10.x, 11.x). Emerge hangs on "++ qatom x11-base/xorg-server-1.9.5". Reproducible: Always Steps to Reproduce: 1. emerge -av ati-drivers Actual Results: emerge hangs on: >>> Emerging (1 of 1) x11-drivers/ati-drivers-11.2 * ati-driver-installer-11-2-x86.x86_64.run RMD160 SHA1 SHA256 size ;-) ... [ ok ] * Package: x11-drivers/ati-drivers-11.2 * Repository: gentoo * Maintainer: lu_zero@gentoo.org enrico.tagliavini@gmail.com,x11@gentoo.org * USE: amd64 elibc_glibc kernel_linux modules multilib qt4 userland_GNU * FEATURES: preserve-libs sandbox Emerge with debuging flag: + local retval + pkg_setup + MODULE_DIR=/var/tmp/portage/x11-drivers/ati-drivers-11.2/work/common//lib/modules/fglrx/build_mod ++ cut -d ' ' -f 3 +++ best_version x11-base/xorg-server +++ '[' setup == depend ']' +++ [[ -n 1 ]] +++ /usr/lib64/portage/bin/ebuild-ipc best_version / x11-base/xorg-server +++ local retval=0 +++ case "${retval}" in +++ return 0 ++ qatom x11-base/xorg-server-1.9.5 Expected Results: Successful emerge ati-drivers Emerge --info >> http://pastebin.com/MvGSKhtB Full emerge ati-drivers debug information >> http://pastebin.com/fcCYuNQb I have tried downgrade xorg-server and gentoo-sources but without success...
Same problem here, q / qatom hangs with 100% CPU. Noticed when trying to update to 2.6.38 (last kernel was 2.6.37). I don't think it is related to ati-drivers actually, since the same hang occurs on a emerge --sync (q-reinitialize).
Worked around by downgrading portage-utils.
Comfirmed that downgrading portage-utils fixes the problem...
(In reply to comment #0) > Emerge --info >> http://pastebin.com/MvGSKhtB > Full emerge ati-drivers debug information >> http://pastebin.com/fcCYuNQb Please attach that as files.
Created attachment 266915 [details] emerge --info
Created attachment 266917 [details] ati-drivers emerge debug
Looks like sys-apps/portage-2.2.2 which Gentoo doesn't have. And apparently that uses portage-utils somehow, which Gentoo's doesn't. I've never seen anything like this but it looks like you're a funtoo user.
Oh wait - ati-drivers /does/ DEPEND on portage-utils. How weird.
Still don't know which version it is that fails, but it's rather strange that it wants qatom just to figure out a version of another installed package.
Not an ati-drivers issue, I reassigned it to bug wranglers.
(In reply to comment #9) > Still don't know which version it is that fails, but it's rather strange that > it wants qatom just to figure out a version of another installed package. portage-utils-5.0, but not < 5.0
(In reply to comment #9) > Still don't know which version it is that fails, but it's rather strange that > it wants qatom just to figure out a version of another installed package. The answer is simple: AMD catalyst linux package used to provide different binaries for different Xorg versions. Lastest versions now have a unified binary which should work with all recent Xorg version (they just dropped 6.8 support). If this policy will last i will remove the qatom check in the future :)
Why should bug-wranglers solve this? ati-drivers uses qatom, and if there's a problem in qatom, then you reassign to portage-utils@gentoo.org or CC them.
you are right, i'm sorry i was really in hurry, then my laptop died and i was cutted off :(
sys-auth/pambase also hangs up on qatom with portage-utils-5.0, but not in earlier version
ati-drivers depends no more on portage-utils from version 11.11 (but qatom is no more used from many versions, i can't rember which one exactly and i'm lazy sorry ;) )
there is no portage-utils-5.0. there is a 0.5 version, and current unstable is 0.6. however, that works just fine for me: $ qatom --version portage-utils-0.6: compiled on Oct 2 2011 $Id: qatom.c,v 1.8 2011/02/21 01:33:47 vapier Exp $ atom written for Gentoo by <solar and vapier @ gentoo.org> $ qatom x11-base/xorg-server-1.9.5 x11-base xorg-server 1.9.5 as for the ebuild usage, it really doesn't seem necessary. that sort of dependency checking should be handled with DEPEND/RDEPEND. http://sources.gentoo.org/x11-drivers/ati-drivers/ati-drivers-11.2.ebuild?revision=1.1
(In reply to comment #17) > as for the ebuild usage, it really doesn't seem necessary. that sort of > dependency checking should be handled with DEPEND/RDEPEND. > http://sources.gentoo.org/x11-drivers/ati-drivers/ati-drivers-11.2.ebuild?revision=1.1 In the past releases of ati-drivers i had no other way to make a single ebuild able to install on different xorg-server versions. It is a binary blob, and it was providing different binaries for different xorg version. In short i had to check the xorg version at ebuild runtime and install the right binaries for the current installed xorg. Now AMD provides unified binaries which works on all xorg version :)
I just kept the qatom usage a bit couse i was worried AMD was going back to provide a different binary blob for the next xorg-server, but they didn't. The original work where qatom was needed for example is here http://sources.gentoo.org/cgi-bin/viewvc.cgi/gentoo-x86/x11-drivers/ati-drivers/ati-drivers-10.10.ebuild?hideattic=0&view=markup
(In reply to comment #18) you should have been able to use `has_version`: if has_version =x11-base/xorg-server-1.9* ; then ... 1.9 stuff ... elif has_version =x11-base/xorg-server-1.8* ; then ... 1.8 stuff ... ... etc ... also, latest ati-drivers ebuild still depends on portage-utils even though it no longer uses qatom ?
It doesn't
well, once someone can show me the failing example like in Comment #17, we can move forward in debugging ...
