On Feb 2, 2013, at 5:01 PM, Felix Miata <mrmazda@xxxxxxxxxxxxx> wrote: > > My first response to thread, subsequent to time OP provided access to Xorg.0.log. http://lists.fedoraproject.org/pipermail/test/2013-February/113591.html It's unclear that nomodeset is coming from grub.cfg, rather than user edited entry in grub shell, or even if it does come from grub.cfg that it's also in /etc/default/grub. We need to see that file. And even then I'd sooner believe that file was user modified. > Almost perfect. So if Anaconda sees itself putting nomodeset into GRUB_CMDLINE_LINUX, nomodeset inclusion should be aborted, and a generic /etc/grub.d/49_nomodeset be enabled that actually contains it. Or some equivalent process that provides a conventional stanza plus at least one nomodeset stanza. Just don't put nomodeset in every stanza unless lspci | grep VGA finds a string that means KMS is incompatible with a required X driver included in the release. And I think anaconda producing an /etc/grub.d/49_nomodeset entry is not so straightforward for a distro because it's unclear (to me) that the GRUB syntax is now stable. To have to update this file as GRUB is updated obviates the whole point of grub2-mkconfig. Better to base a new "Nomodeset" creation feature in grub-mkconfig along the lines of Recovery entries, which merely add single to the end of the kernel parameter line, currently suppressed with: GRUB_DISABLE_RECOVERY="true" So the feature you're suggesting is perhaps, by default: GRUB_ENABLE_NOMODESET="false" And for anaconda to be able to flip that value to true if it detects cat /proc/cmdline contains nomodeset. Then when grub2-mkconfig is called, one nomodeset entry is created per kernel or something… but anyway, even this is not a trivial amount of code. It's a feature. > I thought I was being clear enough. Anaconda must have already known cmdline contained nomodeset or it it wouldn't have caused its inclusion in the installed grub.cfg, however that was implemented within Anaconda. I've read nothing that conclusively demonstrates anaconda presently behaves this way now. Chris Murphy -- test mailing list test@xxxxxxxxxxxxxxxxxxxxxxx To unsubscribe: https://admin.fedoraproject.org/mailman/listinfo/test
