I'm getting blank screen with mouse cursor when logging in via xdm (gdm). Reproducible: Always Steps to Reproduce: 1. eclean-dist 2. eclean-pkg 3. emerge --sync 4. emerge -a --depclean 5. emerge -1 linux-headers glibc binutils gcc-config gcc 6. gcc-config 1 7. source /etc/profile 8. emerge -1b glibc binutils gcc portage 9. emerge -bke system 10. emerge -ke world 11. revdep-rebuild 12. dispatch-config 13. reboot Actual Results: GNOME displays blank screen with mouse cursor after login Expected Results: Working GNOME environment After I had performed full system upgrade I got blank screen with mouse cursor when tried to log in via xdm (gdm). The same thing for the root account. I even did 'emerge -e gnome-light'. It didn't help.
Created attachment 206452 [details] printout of 'emerge --info'
Created attachment 206453 [details] make.conf
Created attachment 206454 [details] printout of 'gnome-session'
- Firstly please read this link http://www.gentoo.org/doc/en/gcc-optimization.xml. - Then please remerge these packages by using normal CFLAGS. -> like following : CFLAGS="-march=core2 -O2 -pipe"
Please get back us
(In reply to comment #5) > Please get back us > Thanks for your attention and information! I forgot to reassess the CFLAGS after I had switched to the new compiler. Previous one didn't set -mmmx -msse -msse2 -msse3 -mssse3 flags when you stated march=core2 even though these are supported by the hardware! Neither do this version: printout of "gcc <flags go here> -Q --help=target" also shows them disabled. Anyway the system is rebuilding right now with the flags you advised. I'll inform you on the results when it's done.
reopen per comment 6
(In reply to comment #5) > Please get back us > Rebuilding with CFLAGS="-O2 -march=core2 -pipe" solved the problem! Thank you very much for the idea! So I suggest adding -mmmx -msse -msse2 -msse3 -mssse3 to the CFLAGS and rebuilding again. I guess it was because of -O3. Am I right?
re-opening for the toolchain guys.
@toolchain, another -03 gcc-4.4 bug it seems.
some of the bugs I've seen w/ -O3 and 4.4 end up being strict-aliasing violations in the code, not an issue with the compiler. sometimes these are benign until -finline-functions is enabled by -O3, but they are still violations. compiling gdm with -Wstrict-aliasing=2 reveals: * QA Notice: Package has poor programming practices which may compile * fine but exhibit random runtime failures. * gdm-config.c:1042: warning: dereferencing type-punned pointer will break strict-aliasing rules * gdm-config.c:1354: warning: dereferencing type-punned pointer will break strict-aliasing rules * gdm.c:1500: warning: dereferencing type-punned pointer will break strict-aliasing rules * gdm.c:1500: warning: dereferencing type-punned pointer will break strict-aliasing rules * QA Notice: Package has poor programming practices which may compile * fine but exhibit random runtime failures. * gdm-xdmcp-manager.c:808: warning: dereferencing pointer 'clnt_sa.60' does break strict-aliasing rules * gdm-xdmcp-manager.c:2110: warning: dereferencing pointer 'clnt_sa.126' does break strict-aliasing rules * gdm-xdmcp-manager.c:2134: warning: dereferencing pointer 'clnt_sa.129' does break strict-aliasing rules * gdm-xdmcp-manager.c:2389: warning: dereferencing pointer 'clnt_sa.149' does break strict-aliasing rules * gdm-xdmcp-manager.c:1510: warning: dereferencing pointer 'clnt_sa.102' does break strict-aliasing rules
i.e. try building with -fno-strict-aliasing in your CFLAGS and see if the issue is resolved
