| Summary: | [PATCH] xfree using kernel 2.6.3 un Ultra 5 (mach64) does not work | ||
|---|---|---|---|
| Product: | Gentoo Linux | Reporter: | Daniel Seyffer <gentoo> |
| Component: | [OLD] Unspecified | Assignee: | Gentoo X packagers <x11> |
| Status: | RESOLVED FIXED | ||
| Severity: | major | CC: | sparc |
| Priority: | High | ||
| Version: | unspecified | ||
| Hardware: | Sparc | ||
| OS: | All | ||
| Whiteboard: | |||
| Package list: | Runtime testing required: | --- | |
| Attachments: |
gentoo-sparc-mach64.patch
*UNTESTED* updated patch that shouldn't nuke 2.4/non-sparc *UNTESTED* updated patch that shouldn't nuke 2.4/non-sparc attempt 2 |
||
|
Description
Daniel Seyffer
2004-03-03 10:51:34 UTC
Created attachment 26787 [details, diff]
gentoo-sparc-mach64.patch
Is this patch compatible with 2.4 kernels? Nope, the patch as it stands will break everything except 2.6.x on sparc. It should also be using sysfs rather than the procfs interface. Adding a runtime check for kernel version would be nice, as well as making this patch generic enough to apply on all archs (might need #if/#ifdef magic). I think the best thing to do would be to try in this order at runtime: * sysfs, domains enabled * sysfs, no domains * procfs, no domains Checking for procfs + domains is pretty much irrelevant, because domains are only in 2.6.x kernels, and we can assume that /sys is there. Shouldn't need any #if magic, just someone who can remember how to check whether a directory exists in C :) btw, patch works for me, revelling in full X glory again .. Created attachment 27127 [details, diff]
*UNTESTED* updated patch that shouldn't nuke 2.4/non-sparc
Ok. I can't test this right now because my only mach64able box can't be
rebooted for another few days... However, assuming I didn't screw up, the patch
I'm attaching should work on both 2.4.x and 2.6.x, and won't break non-sparc.
Then again, I've not even tried to compile it, and I can't code C, so could
someone brave please test? :)
Created attachment 27166 [details, diff]
*UNTESTED* updated patch that shouldn't nuke 2.4/non-sparc attempt 2
Uhm, I think I got the conditional the wrong way round... Oops... How about
this one?
Worked fine for me, back at 16 bit color using ati driver.
All I did was:
1) created patch file
2) edited up the ebuild using epatch to include the patch
(epatch /usr/portage/distfiles/patchname)
3) let it compile, worked fine.
ciaranm's latest patch above works both on ultra 5 and blade 100 systems for me. how soon can we get it rolled into the X ebuild? Anyone try it on other archs yet? I could try building it on x86, but I have no mach64 cards to test against. It's not just mach64, it's *any* PCI card. Works on x86 for me... No, this should not be converted to use sysfs, why do you think it should? Xfree86 mmaps the pci device through the /proc/bus/pci/... file and then directly talks to it, right? sysfs has no such functionality. Patch looks sane to me. Works for me too, please include soon. Nice, finally it works. /me does a happy dance Based on some comments on IRC, my understanding is that this patch is only for 4.3.99. Since I'm not particularly interested in continuing to maintain xfree 4.3.99, especially once X.org makes a release, I'll plan to add it to the X.org release instead. It should apply pretty clean since that's quite close to xfree-4.4. What timeframe are we looking at for the next X.org release? I only ask because this patch applies to the more popular machines for users to have, and we get questions about how they will work on 2.6 and people trying it out every day. Currently without this patch, these machines cannot use XFree at all on 2.6, and if we're talking more than a week or so until the next X.org release, I'd really like to get this patch put into X.4.3.99. Tuesday or so. here's just another 'me too' ... i needed the aforementioned patch in order to get X up and running In xorg-x11 patchset 0.3 |
