Update to Xorg 1.9 , Painful for VBox User

lovat · 2010-09-30 10:18:27

When I installed Arch linux last month, I noticed that the Xorg version was a RC2 version, I didn't enable testing repo, why a rc package appears in stable repo?
There was a seg fault bug in this RC when using virtualbox video drivers, and this is not the worst...
When I update Arch today, It updates to Xorg 1.9 and Virtual Box guest additions doesn't even support 1.9.
What I only can do is by downgrading xorg, xorg has numerous dependencies, downgrading was painful, I wasted some time downgrading it and didn't finish my work(should be done if xorg works normally).
Can I simply install a stable version xorg ?
Thanks.

Last edited by lovat (2010-10-01 12:54:42)

.:B:. · 2010-09-30 10:21:24

This is Arch... It's bleeding edge. You can't get both (well you can - if you want rolling release but more stability, you can try Sidux).

Release candidates appear in the official repos all the time, if they're deemed stable. There have been a few KDE releases like that.

stryder · 2010-09-30 12:13:40

*whew* Thanks for the heads-up. I saw the xorg upgrade and decided to leave things till the weekend.

@.:B:. sidux is now aptosid.

karol · 2010-09-30 12:20:11

stryder wrote:

@.:B:. sidux is now aptosid.

Good thing they took the opportunity to change the name to a better one ;P

Arch routinely uses RC versions of xorg, nothing wrong with them. I don't have VirtualBox and I don't care if it supports the new xorg or not. If you care, you take care to do it right e.g. put some packages in IgnorePkg or don't update at all.

Last edited by karol (2010-09-30 12:21:31)

JGC · 2010-09-30 14:36:28

The 1.8 RC version we used was exactly the same as the version that was released after it. We often use RC versions from the stable branch, and add patches from the stable branch to it. Sometimes it turns out that they release exactly what we patched together.

As for "stable" X: xorg-server 1.9 is a stable release. This package has been in testing for months without major issues.

jt512 · 2010-09-30 18:50:33

JGC wrote:

As for "stable" X: xorg-server 1.9 is a stable release. This package has been in testing for months without major issues.

I don't think you guys took the problem with the nvidia-256 driver seriously enough. I spent the better part of the day yesterday diagnosing the problem and finding a solution. I had to switch to the nvidia-beta (260) driver, and in order to get it to load, eventually figured out that I had to blacklist the ahci driver.

Multiple people reported problems with nvidia driver in the testing forum. I don't understand why xorg 1.9 was released without a warning to users, or at all for that matter, while this problem was unresolved. I feel like I got hung out to dry by the devs on this one.

Jay

ngoonee · 2010-10-01 00:16:03

jt512 wrote:

JGC wrote:
As for "stable" X: xorg-server 1.9 is a stable release. This package has been in testing for months without major issues.
I don't think you guys took the problem with the nvidia-256 driver seriously enough. I spent the better part of the day yesterday diagnosing the problem and finding a solution. I had to switch to the nvidia-beta (260) driver, and in order to get it to load, eventually figured out that I had to blacklist the ahci driver.
Multiple people reported problems with nvidia driver in the testing forum. I don't understand why xorg 1.9 was released without a warning to users, or at all for that matter, while this problem was unresolved. I feel like I got hung out to dry by the devs on this one.
Jay

The problems were easily resolvable by the user, either by upgrading to nvidia-beta or by downgrading xorg. They do not cause data corruption or data loss.

If you were expecting to have everything Just Work (tm) without your intervention then this may not be the best distro for you. Yes, the devs could have made an announcement, but with the length of time xorg1.9 was in [testing] I can understand why they didn't.

Out of curiosity, how did you spend that 'better part of a day'?

jt512 · 2010-10-01 01:06:29

ngoonee wrote:

jt512 wrote:
JGC wrote:
As for "stable" X: xorg-server 1.9 is a stable release. This package has been in testing for months without major issues.
I don't think you guys took the problem with the nvidia-256 driver seriously enough. I spent the better part of the day yesterday diagnosing the problem and finding a solution. I had to switch to the nvidia-beta (260) driver, and in order to get it to load, eventually figured out that I had to blacklist the ahci driver.
Multiple people reported problems with nvidia driver in the testing forum. I don't understand why xorg 1.9 was released without a warning to users, or at all for that matter, while this problem was unresolved. I feel like I got hung out to dry by the devs on this one.
Jay
The problems were easily resolvable by the user, either by upgrading to nvidia-beta or by downgrading xorg. They do not cause data corruption or data loss.

First of all downgading isn't a solution. It's what you resort to when you can't actually find the solution.

Secondly, "upgrading to nvidia-beta" was not an "easy" solution either. Just determining that the problem was with the nvidia driver took time, then determining that the beta driver might be a solution took time, then figuring why the driver wouldn't load took time. And the only reason I was able to solve that problem on my own was that someone in testing had a similar (though not identical) problem, and luckily his solution (blacklisting the ahci driver) worked for me, apparently with no ill side effects.

If you were expecting to have everything Just Work (tm) without your intervention then this may not be the best distro for you.

Bullshit. You can use that line to excuse any possible error in judgment that any developer ever makes. "Oops, we broke Pacman. Well, if you were expecting to have everything work out of the box, this might not be the best distro for you."

Yes, the devs could have made an announcement, but with the length of time xorg1.9 was in [testing] I can understand why they didn't.

No matter how long it was in testing is irrelevant, if multiple people in testing are still having serious problems with it, which was the case. It's called "testing" to see if it passes. If it's failing, you either don't release it; or you release it with a warning, and an explanation of how to work around the problem.

Out of curiosity, how did you spend that 'better part of a day'?

I think I've already explained that.

Jay

karol · 2010-10-01 01:18:04

@ jt512
VirtualBox is not pacman. Everyone (minus one dev ;P) uses pacman.

Yes, I think that there should have been an announcement, but many people don't read them and come to the forums for help, because ... oh yes, they expect it to Just Work (tm). "I mean, I -Syu twice a day, it looks cool etc. and now everything is broken, please help." It might look as if no one cares about the ML, the News, Stickies, so why take time to post them?

Last edited by karol (2010-10-01 01:19:23)

ewaller · 2010-10-01 01:27:50

Guys,

You have all made good comments. Clearly there are passionate feelings on both sides of this. I suggest you take it down a notch before you start to damage your respective positions.

jt512 · 2010-10-01 01:30:45

karol wrote:

@ jt512
VirtualBox is not pacman. Everyone (minus one dev ;P) uses pacman.
Yes, I think that there should have been an announcement, but many people don't read them and come to the forums for help, because ... oh yes, they expect it to Just Work (tm). "I mean, I -Syu twice a day, it looks cool etc. and now everything is broken, please help." It might look as if no one cares about the ML, the News, Stickies, so why take time to post them?

I suspect that the silent majority of users actually do read the announcements, but, since they do, we don't hear from them on issues that have been reported in those announcements. So, why take the time to post them? (1) for the benefit of the people who do read them, and (2) because it seems unconscionable to release an upgrade with known issues, without disclosing those issues.

karol · 2010-10-01 01:38:48

jt512 wrote:

karol wrote:
@ jt512
VirtualBox is not pacman. Everyone (minus one dev ;P) uses pacman.
Yes, I think that there should have been an announcement, but many people don't read them and come to the forums for help, because ... oh yes, they expect it to Just Work (tm). "I mean, I -Syu twice a day, it looks cool etc. and now everything is broken, please help." It might look as if no one cares about the ML, the News, Stickies, so why take time to post them?
I suspect that the silent majority of users actually do read the announcements, but, since they do, we don't hear from them on issues that have been reported in those announcements. So, why take the time to post them? (1) for the benefit of the people who do read them, and (2) because it seems unconscionable to release an upgrade with known issues, without disclosing those issues.

Yes, I understand, but as long as you have a working kernel and pacman, you can (and often are expected to) downgrade.

Last edited by karol (2010-10-01 01:39:08)

jt512 · 2010-10-01 01:48:40

karol wrote:

jt512 wrote:
karol wrote:
@ jt512
VirtualBox is not pacman. Everyone (minus one dev ;P) uses pacman.
Yes, I think that there should have been an announcement, but many people don't read them and come to the forums for help, because ... oh yes, they expect it to Just Work (tm). "I mean, I -Syu twice a day, it looks cool etc. and now everything is broken, please help." It might look as if no one cares about the ML, the News, Stickies, so why take time to post them?
I suspect that the silent majority of users actually do read the announcements, but, since they do, we don't hear from them on issues that have been reported in those announcements. So, why take the time to post them? (1) for the benefit of the people who do read them, and (2) because it seems unconscionable to release an upgrade with known issues, without disclosing those issues.
Yes, I understand, but as long as you have a working kernel and pacman, you can (and often are expected to) downgrade.

Really? I don't think so. How many times are we told that the entire system is intended to be upgraded? I've rarely had to downgrade. Although there have been plenty of times I could have done so, I've always attempted to diagnose the problem properly and implement a solution. And, again, this is all beside the point. The devs released an upgrade with known problems, without notifying the users. "They can just downgrade?" Then why release the upgrade? "Let them eat cake?"

Jay

ngoonee · 2010-10-01 03:04:02

jt512 wrote:

First of all downgading isn't a solution. It's what you resort to when you can't actually find the solution.
Secondly, "upgrading to nvidia-beta" was not an "easy" solution either. Just determining that the problem was with the nvidia driver took time, then determining that the beta driver might be a solution took time, then figuring why the driver wouldn't load took time. And the only reason I was able to solve that problem on my own was that someone in testing had a similar (though not identical) problem, and luckily his solution (blacklisting the ahci driver) worked for me, apparently with no ill side effects.

In this situation downgrading is a (temporary) solution. Explanation later.

jt512 wrote:

If you were expecting to have everything Just Work (tm) without your intervention then this may not be the best distro for you.
Bullshit. You can use that line to excuse any possible error in judgment that any developer ever makes. "Oops, we broke Pacman. Well, if you were expecting to have everything work out of the box, this might not be the best distro for you.

Have you actually read how the decision was made in [arch-dev-public]? Gnome-unstable was coming into [testing], xorg had been sitting there for ages, and [testing] needed to be cleared for new Mesa (which is just about to be released). This was not a spur-of-the-moment decision.

jt512 wrote:

Yes, the devs could have made an announcement, but with the length of time xorg1.9 was in [testing] I can understand why they didn't.
No matter how long it was in testing is irrelevant, if multiple people in testing are still having serious problems with it, which was the case. It's called "testing" to see if it passes. If it's failing, you either don't release it; or you release it with a warning, and an explanation of how to work around the problem.

[testing] is meant to test for Arch errors. In this case there are no Arch-induced errors. If upstream has a problem (in this case nvidia), that can delay the move from [testing] (and already has for ages in this case) but not stop it. Ask the catalyst users about their experiences when xorg1.8 came out. Ask those with older nvidia cards whom nvidia has stopped supporting on later xorg versions. With these binary-only drivers there is no way of knowing when (or if) support will come.

Frustration is a natural reaction, but unhelpful in this case. For the future I'd suggest subscribing to [arch-dev-public], since you'll get heads-up on things like this which would be happening. Ideally you'd have a machine using [testing] as well, but not everyone has multiple machines.

karol · 2010-10-01 12:39:52

jt512, just because there is an update doesn't mean you have to, or are advised to, update.

Some people have setups that are unaffected by the bugs that hurt you and yet they would have to wait until it's fixed _for you_ even though it already works for them. Sure, they can use [testing] or compile on their own.

I '-Syu' regularly and if I see that there's e.g. an xorg update (or kernel, pacman etc.), I go and read the ML. If The nvidia-173xx drivers for the new release aren't there yet, I run a script to switch to nouveau or wait a bit with the update.
On my other box I have xorg-server and his dozen friends in IgnorePkg :-)

jt512 · 2010-10-01 21:22:16

ngoonee wrote:

Frustration is a natural reaction, but unhelpful in this case. For the future I'd suggest subscribing to [arch-dev-public], since you'll get heads-up on things like this which would be happening. Ideally you'd have a machine using [testing] as well, but not everyone has multiple machines.

I do subscribe to arch-dev-public, and there was not a word there about a problem with nvidia under xorg-1.9. I simply cannot fathom how the devs can release a major upgrade which they know will cause users problems without warning them about it. By writing one or two sentences they could have saved the user base 100s of frustrating man–hours.

Jay

jt512 · 2010-10-01 21:25:39

karol wrote:

jt512,
[ ... ]
Some people have setups that are unaffected by the bugs that hurt you and yet they would have to wait until it's fixed _for you_ even though it already works for them.

I see that reading comprehension is still alive and well on the Internet. How many times can you miss the part where I said that the devs should have at least posted a warning about the update causing problems with nvidia.

Jay

karol · 2010-10-01 21:36:12

jt512 wrote:

karol wrote:
jt512,
[ ... ]
Some people have setups that are unaffected by the bugs that hurt you and yet they would have to wait until it's fixed _for you_ even though it already works for them.
I see that reading comprehension is still alive and well on the Internet. How many times can you miss the part where I said that the devs should have at least posted a warning about the update causing problems with nvidia.
Jay

I admit that recently I've been misreading parts of poeple's posts way too often, but I've already agreed with you:
https://bbs.archlinux.org/viewtopic.php … 29#p833929

karol wrote:

Yes, I think that there should have been an announcement (...)

eldragon · 2010-10-01 22:31:34

jt512 wrote:

I do subscribe to arch-dev-public, and there was not a word there about a problem with nvidia under xorg-1.9. I simply cannot fathom how the devs can release a major upgrade which they know will cause users problems without warning them about it. By writing one or two sentences they could have saved the user base 100s of frustrating man–hours.
Jay

ehem...

http://mailman.archlinux.org/pipermail/ … 18022.html

jt512 · 2010-10-02 04:44:04

eldragon wrote:

jt512 wrote:
I do subscribe to arch-dev-public, and there was not a word there about a problem with nvidia under xorg-1.9. I simply cannot fathom how the devs can release a major upgrade which they know will cause users problems without warning them about it. By writing one or two sentences they could have saved the user base 100s of frustrating man–hours.
Jay
ehem...
http://mailman.archlinux.org/pipermail/ … 18022.html

Well, you're right. It's there. I didn't see it because the subject line, "Clearing out testing," had nothing to do with xorg, nvidia, or anything that I would seem to be of interest to me as an ordinary user.

Edited to add:

From the same thread. The smoking gun, as far as I'm concerned:

Date: Tue, 28 Sep 2010 21:54:58 +0300
From: Ionuț Bîru <ibiru@archlinux.org>    
Subject: Re: [arch-dev-public] Clearing out [testing]
To: Public mailing list for Arch Linux development <arch-dev-public@archlinux.org>      

                                                                                                                                                                                      On 09/28/2010 09:35 PM, Pierre Schmitz wrote:                                                                                                                                                                                                  
> On Tue, 28 Sep 2010 20:25:27 +0200, Andreas Radke<a.radke@arcor.de>                                                                                                                                                                          
> wrote:                                                                                                                                                                                                                                       
>> A closed source driver should never hold back an OSS package. So what's                                                                                                                                                                     
>> the state with nvidia and the new beta driver? If it won't hit the repos                                                                                                                                                                    
>> until the end of the week I don't want to wait any longer.                                                                                                                                                                                  
>                                                                                                                                                                                                                                              
> I agree. If it's really that driver holding xorg back it should just be                                                                                                                                                                      
> moved. Nvidia doesn't tellanybody when they plan new releases or fix a                                                                                                                                                                       
> bug. Last but not least I didn't notice any slowdown and there is also                                                                                                                                                                       
> no report in our tracker about this. So it cannot be that critical.                                                                                                                                                                          
>                                                                        
                                                                                                                                                                                                                                               
i can't agree with you. every time somebody reported an issue with                                                                                                                                                                             
nvidia, being even a performance regression, you personally closed it                                                                                                                                                                          
as "we can't do anything"                                                                                                                                                                                                                      
                                                                                                                                                                                                                                               
i'm hit with this issue and i know a dozen. this regression makes                                                                                                                                                                              
rendering antialised fonts in gtk 10 times slower and in kde when                                                                                                                                                                              
using non-antialiased fonts [1]                                                                                                                                                                                                                
                                                                                                                                                                                                                                               
we have to think about our users first. your statistics say that                                                                                                                                                                               
nvidia is dominant in our community[2]                                                                                                                                                                                                         
                                                                                                                                                                                                                                               
[1] [url]http://www.nvnews.net/vbulletin/showthread.php?t=154563[/url]                                                                                                                                                                                    
[2] [url]https://www.archlinux.de/?page=FunStatistics[/url]                                                                                                                                                                                               
                                                                                                                                                                                                                                               
--                                                                                                                                                                                                                                             
Ionuț

Last edited by jt512 (2010-10-02 04:56:32)

na12 · 2010-10-03 19:31:46

if anyone else has a problem with xf86-video-nv driver,I can't start x.

alterkacker · 2010-10-06 23:31:28

So back to the original VBox problem: there's a pre-release version of the Virtualbox Guest Additions that's compatible with Xorg 1.9. Get it at http://www.virtualbox.org/download/test … r65982.iso

lovat · 2010-10-07 10:43:37

alterkacker wrote:

So back to the original VBox problem: there's a pre-release version of the Virtualbox Guest Additions that's compatible with Xorg 1.9. Get it at http://www.virtualbox.org/download/test … r65982.iso

That's great, but the link doesn't work.

I Once heard on VBox forum, a developer said guest additions would support Xorg 1.9 before Ubuntu 10.10 release.

xduugu · 2010-10-07 11:51:59

From http://www.virtualbox.org/ticket/7306

http://www.virtualbox.org/download/test … -66155.iso

lovat · 2010-10-07 14:09:00

Well, I still got a few problems when using VBoxGuestAdditions_3.2.9-66155.iso in Arch Linux. The following error says failed to load kernel module "vboxvideo".

[    46.910] [drm] failed to load kernel module "vboxvideo"
[    46.910] (EE) [drm] drmOpen failed.
[    46.910] (EE) VBoxVideo(0): DRIScreenInit failed, disabling DRI.

However, in the previous output, I can see the vboxvideo module has already been loaded.

[    46.249] (II) LoadModule: "vboxvideo"
[    46.250] (II) Loading /usr/lib/xorg/modules/drivers/vboxvideo_drv.so
[    46.266] (II) Module vboxvideo: vendor="Oracle Corporation"
[    46.266]     compiled for 1.5.99.901, module version = 1.0.1
[    46.266]     Module class: X.Org Video Driver
[    46.266]     ABI class: X.Org Video Driver, version 8.0
[    46.266] (**) Load address of symbol "VBOXVIDEO" is 0xb74b7060

I can still startx, but when I drag the window around, I can see the window flickering.

when I run lsmod ：

 lsmod | grep vbox
vboxsf                 31332  0 
vboxguest             115335  7 vboxsf

Should I also see vboxvideo in this list?

Arch Linux

#1 2010-09-30 10:18:27

Update to Xorg 1.9 , Painful for VBox User

#2 2010-09-30 10:21:24

Re: Update to Xorg 1.9 , Painful for VBox User

#3 2010-09-30 12:13:40

Re: Update to Xorg 1.9 , Painful for VBox User

#4 2010-09-30 12:20:11

Re: Update to Xorg 1.9 , Painful for VBox User

#5 2010-09-30 14:36:28

Re: Update to Xorg 1.9 , Painful for VBox User

#6 2010-09-30 18:50:33

Re: Update to Xorg 1.9 , Painful for VBox User

#7 2010-10-01 00:16:03

Re: Update to Xorg 1.9 , Painful for VBox User

#8 2010-10-01 01:06:29

Re: Update to Xorg 1.9 , Painful for VBox User

#9 2010-10-01 01:18:04

Re: Update to Xorg 1.9 , Painful for VBox User

#10 2010-10-01 01:27:50

Re: Update to Xorg 1.9 , Painful for VBox User

#11 2010-10-01 01:30:45

Re: Update to Xorg 1.9 , Painful for VBox User

#12 2010-10-01 01:38:48

Re: Update to Xorg 1.9 , Painful for VBox User

#13 2010-10-01 01:48:40

Re: Update to Xorg 1.9 , Painful for VBox User

#14 2010-10-01 03:04:02

Re: Update to Xorg 1.9 , Painful for VBox User

#15 2010-10-01 12:39:52

Re: Update to Xorg 1.9 , Painful for VBox User

#16 2010-10-01 21:22:16

Re: Update to Xorg 1.9 , Painful for VBox User

#17 2010-10-01 21:25:39

Re: Update to Xorg 1.9 , Painful for VBox User

#18 2010-10-01 21:36:12

Re: Update to Xorg 1.9 , Painful for VBox User

#19 2010-10-01 22:31:34

Re: Update to Xorg 1.9 , Painful for VBox User

#20 2010-10-02 04:44:04

Re: Update to Xorg 1.9 , Painful for VBox User

#21 2010-10-03 19:31:46

Re: Update to Xorg 1.9 , Painful for VBox User

#22 2010-10-06 23:31:28

Re: Update to Xorg 1.9 , Painful for VBox User

#23 2010-10-07 10:43:37

Re: Update to Xorg 1.9 , Painful for VBox User

#24 2010-10-07 11:51:59

Re: Update to Xorg 1.9 , Painful for VBox User

#25 2010-10-07 14:09:00

Re: Update to Xorg 1.9 , Painful for VBox User

Board footer