5:51 p.m.
It's quite well-known that NVIDIA proprietary driver doesn't work on
Rawhide due to a GPL symbol compatibility issue (which only shows up
with kernel debugging enabled, hence only shows up in Rawhide).
There are workarounds for this, however:
http://forums.fedoraforum.org/showthread.php?t=231340
that look like they could be applied to the NVIDIA build in Fusion.
Should we apply one so that we get a working NVIDIA package for Rawhide
in Fusion?
--
adamw
8:02 p.m.
2009/10/6 Adam Williamson <adamwill(a)shaw.ca>:
It's quite well-known that NVIDIA proprietary driver doesn't
work on
Rawhide due to a GPL symbol compatibility issue (which only shows up
with kernel debugging enabled, hence only shows up in Rawhide).
There are workarounds for this, however:
http://forums.fedoraforum.org/showthread.php?t=231340
that look like they could be applied to the NVIDIA build in Fusion.
Should we apply one so that we get a working NVIDIA package for Rawhide
in Fusion?
No we shound'n't. There is no purpose of workarounding this
feature of
the kernel this way IMHO.
I meant that if we want debugging things from the kernel, we cannot
use nvidia at the same time because debuging will be pointless.
If we want to use nvidia, then we want to turn off debugging on a
kernel that's the other side of the fix (a way better) .
People using this hack is fine, they are on their own. But
redistributing this hack for people not aware of what was done is
another thing.
So what I have in mind is more about to add a warning on the beta serie such as:
"Please test this package only if your hardware is still under
warranty so you can return to the Manufacturer in case of problem."
I'm dirrecly referencing this problem
http://www.nvnews.net/vbulletin/showthread.php?t=131849 which was
raised to this mailing list earlier. (and isn't only proprietary
driver kind of problem, some intel nic had similar issue IIRC).
Nicolas.
11:25 p.m.
On Tue, 2009-10-06 at 22:02 +0200, Nicolas Chauvet wrote:
2009/10/6 Adam Williamson <adamwill(a)shaw.ca>:
> It's quite well-known that NVIDIA proprietary driver doesn't work on
> Rawhide due to a GPL symbol compatibility issue (which only shows up
> with kernel debugging enabled, hence only shows up in Rawhide).
>
> There are workarounds for this, however:
>
> http://forums.fedoraforum.org/showthread.php?t=231340
>
> that look like they could be applied to the NVIDIA build in Fusion.
> Should we apply one so that we get a working NVIDIA package for Rawhide
> in Fusion?
No we shound'n't. There is no purpose of workarounding this feature of
the kernel this way IMHO.
I meant that if we want debugging things from the kernel, we cannot
use nvidia at the same time because debuging will be pointless.
If we want to use nvidia, then we want to turn off debugging on a
kernel that's the other side of the fix (a way better) .
There's no way that 'we' - RPM Fusion - can do that except by issuing
our own kernel. It's a compile-time option, and you're not going to
convince the Fedora kernel maintainers to disable debugging in the
Rawhide kernel compile so that people can use the NVIDIA proprietary
driver.
"I meant that if we want debugging things from the kernel, we cannot use
nvidia at the same time because debuging will be pointless."
Well, yes. I don't see how that bears on the debate, though. If you
report a kernel bug with NVIDIA loaded, you get a comment asking you to
try without it loaded. That's the case whatever we do to the NVIDIA
package to make it work for people who want to use NVIDIA on Rawhide, so
I don't see the relevance of bringing it up. If this were a reason not
to make NVIDIA work, we shouldn't release packages of it at all, ever.
There are legitimate reasons to want to have NVIDIA running on Rawhide.
Fr'instance, it would make it much easier for me to test my video
acceleration stuff, since there is no video acceleration support in
nouveau.
--
adamw
8:28 a.m.
-------- Original Message --------
Subject: Re: NVIDIA Rawhide fix?
From: Nicolas Chauvet <kwizart(a)gmail.com>
To: RPM Fusion developers discussion list
<rpmfusion-developers(a)lists.rpmfusion.org>
Date: 06.10.2009 22:02
I meant that if we want debugging things from the kernel, we cannot
use nvidia at the same time because debuging will be pointless.
If we want to use nvidia, then we want to turn off debugging on a
kernel that's the other side of the fix (a way better) .
Still, on the other side, there are people who really do use their
rawhide machines for everyday tasks (I do). Maybe they are just
interested in developing/debugging their stuff on the latest versions of
software and don't really care about kernel debugging.
People using this hack is fine, they are on their own. But
redistributing this hack for people not aware of what was done is
another thing.
I believe we shouldn't make peoples lives harder here. I'd opt for a
build time option in the spec to enable this hack.
I don't see why we should leave people alone with the ugly NVIDIA
installer that is quite destructive in the way it installs it's files.
That way we don't need to warrant for the drivers packaged by us but
also offer the option to say: You can recompile the driver with this
special option at your own risk.
That way all people will love us and we eventually are able to take over
the world.
dd
Felix
11:08 a.m.
2009/10/8 Felix Kaechele <felix(a)fetzig.org>:
-------- Original Message --------
Subject: Re: NVIDIA Rawhide fix?
From: Nicolas Chauvet <kwizart(a)gmail.com>
To: RPM Fusion developers discussion list
<rpmfusion-developers(a)lists.rpmfusion.org>
Date: 06.10.2009 22:02
> I meant that if we want debugging things from the kernel, we cannot
> use nvidia at the same time because debuging will be pointless.
> If we want to use nvidia, then we want to turn off debugging on a
> kernel that's the other side of the fix (a way better) .
Still, on the other side, there are people who really do use their rawhide
machines for everyday tasks (I do). Maybe they are just interested in
developing/debugging their stuff on the latest versions of software and
don't really care about kernel debugging.
> People using this hack is fine, they are on their own. But
> redistributing this hack for people not aware of what was done is
> another thing.
I believe we shouldn't make peoples lives harder here. I'd opt for a build
time option in the spec to enable this hack.
I don't see why we should leave people alone with the ugly NVIDIA installer
that is quite destructive in the way it installs it's files.
Good! so you have
been elected to maintain a kernel build without
debuggin stuff enabled.
Nicolas (kwizart)
4:52 p.m.
Wouldn't someone just need to make the kernel.spec reflect the following edit:
--- kernel.spec.orig 2009-10-08 11:50:13.585861723 -0500
+++ kernel.spec 2009-10-08 11:49:31.384923220 -0500
@@ -129,7 +129,7 @@
# Set debugbuildsenabled to 1 for production (build separate debug kernels)
# and 0 for rawhide (all kernels are debug kernels).
# See also 'make debug' and 'make release'.
-%define debugbuildsenabled 0
+%define debugbuildsenabled 1
and then rebuild?
-Adam
--
http://maxamillion.googlepages.com
---------------------------------------------------------
() ascii ribbon campaign - against html e-mail
/\ www.asciiribbon.org - against proprietary attachments
6:12 p.m.
On Thu, 2009-10-08 at 13:08 +0200, Nicolas Chauvet wrote:
Good! so you have been elected to maintain a kernel build without
debuggin stuff enabled.
I still don't understand why you think doing this, rather than doing a
rather small patch in the NVIDIA package, is better. It _may_ be, I'm
not discounting the possibility, I just haven't got the rationale yet.
--
adamw
5558
days inactive
5561
days old
rpmfusion-developers@lists.rpmfusion.org
6 comments
4 participants
participants (4)
-
Adam Miller -
Adam Williamson -
Felix Kaechele -
Nicolas Chauvet