Slashdot Mirror
Ubuntu LTS Experiences X.org Memory Leak
MonsterTrimble writes "Ubuntu 10.04 LTS Beta 2 is experiencing a major memory leak due to patches for X.org. 'An X.Org Server update that was pushed into the Lucid repository last week has resulted in the system being slower and slower as it is left on, until it reaches a point where the system is no longer usable. ... In order to make the Ubuntu 10.04 LTS deadline, the developers are looking at just reverting three of the patches, which brings the GLX version back to 1.2. Ubuntu developers are now desperate for people willing to test out this updated X.Org Server package so they can determine by this Friday whether to ship it with Ubuntu 10.04 LTS or doing an early SRU (Stable Release Update). Right now this X.Org Server that's being tested is living in the ubuntu-x-swat PPA.'"
How come this wasn't caught when they were profiling? Notice I said "when" - the X.org people aren't seriously deploying patches to such a crucial app without profiling first, are they?
It's getting popular, better tear it down, and kill it!
It's funny how when a FOSS project gets to a certain level of popularity (Firefox, Ubuntu) there seem to be a vocal group of people that try to tear them down. Oh my god, a version of Linux that is nearly user friendly, it's not hardcore enough for me!
Um, they are calling you because they want your help to figure it out.
It is a concept known as "Open Source".
This is the reason why hard release schedules kill Ubuntu. The devs slipped 6.04 to 6.06 for similar reasons, and the release was great. Contrast that with the scramble to get 8.04 released on time and then look at the mess it was in when it was delivered. It wasn't stable until 8.04.1. Ubuntu needs to be more flexible. Slip a month, fix this problem, then release. No biggie.
Why not hold the release until the bug is fixed ?
The Ubuntu Wiki has details on this issue at the GEMLeak entry. It provides instructions on how to upgrade to (and remove) the candidate packages in the PPA. This comment is worthy of note for those already on Lucid:
This does not affect cards using proprietary drivers or not using DRI2. Intel will always be affected since DRI2 is used with and without KMS, ATI uses DRI1 without KMS.
I understand that fixed release dates are useful for planning, but I think Ubuntu has put too much emphasis on them. Software should not be released until it is ready.
The idea of releasing it on schedule, with this big bug in it, and then issuing a quick fix when it is ready (one of the options discussed) is silly and rather deceptive. If what they have on April 30th is only beta quality then don't call it a release just so you can say that you stuck to your schedule.
When I first heard about Ubuntu, I thought to myself, "Great, a user friendly Linux distro!"
When I first heard it I thought "that's the stupidest fucking name I've ever heard".
Then when I first tried it I thought "Man that is WAY too much brown and orange.".
Overall though, if you ignore the name, and change your theme around to something a bit more pleasant, it's really pretty slick. If anything has a chance to get people adopt Linux for general usage, Ubuntu is it.
Either that or LinuxMint, which is effectively "Ubuntu with the ugly removed".
"People who think they know everything are very annoying to those of us who do."-Mark Twain
You've all got to help them FAST!
Because the world would, you know, end in a fiery ball of flaming death if the LTS ended up being 10.05!
(This policy is why I replaced Ubuntu on my desktop)
If you read the wiki page referenced carefully, it would seem that the general consensus is that the bug is fixed in the testing packages. https://wiki.ubuntu.com/X/Testing/GEMLeak Seems a bit blown out of proportion to me.
10.04 is supposed to be a LTS release, and they are nearing their deadline. Roll back to the "stable" version of X, and push these patches forward to 10.10. Anyone who cares about having the latest and greatest will roll along with the 6 month release cycle.
chown -R us.
Which would be why they sent that to their -dev mailing list.
You do know the difference between someone sending a request, and someone else reporting on that request, and someone else reporting on someone else reporting on that request?
Right???
> system being slower and slower as it is left on, until it reaches
> a point where the system is no longer usable
At last Linux is feature-complete with MS Windows and ready for the desktop!
This isn't the only video problem in the Lucid Lynx betas. Since upgrading, I've been having a problem where x.org sometimes fails to start up when I boot. Presumably this is a separate problem from the one described in TFA, since you wouldn't expect to see a memory leak's effects showing up at boot time.
Jaunty and Karmic were really terrible releases, IMO. The good news for me is that sound, which broke when I upgraded to Jaunty, is now working for me again with Lucid. I'm hoping that Lucid gets nice and stable over the long lifetime it will have as an LTS release. In the past, I'd been upgrading ubuntu steadily rather than waiting for the next LTS, mainly because I wanted my apps upgraded. That was such a miserable experience that I'm planning not to do it anymore; I'll just stay with Lucid until the next LTS.
I like debian and ubuntu better than the other OSS systems I've used (Mandrake, Red Hat, FreeBSD), but this close tie-in between updating apps and updating the OS can really be a pain. The OS-level tweaking has never made my life any better. As a user, I couldn't care less about stuff like OSS versus ALSA. I would really love it if ubuntu would focus more on fixing bugs in the OS while keeping applications up to date, but not gratuitously breaking stuff in the OS just because they want to be on the cutting edge.
Another thing can be a drag about ubuntu is that they aren't very careful at all about keeping Gnome separate from the underlying OS. Anyone who uses a WM other than Gnome with ubuntu is going to run into lots of things that don't work properly, because the developers always seem to feel free to make changes without testing them on any other WM. For example, here is a bug in xsplash. It causes problems for people who aren't using Gnome. You know you're in trouble when you have functions whose names begin with "temporary_hack..." This one was not a bug in a beta, BTW, but a bug in a real release.
Find free books.
Seriously, they need to hide their source code better, so random incompetent people off the street don't mess with it. What, do they just let ANYONE see it?
Installed by the OEM not MS.
but... Linux doesn't have bugs!
Sure it does.
The point is that they get found and then get fixed fast.
Ubuntu's problem occurred because they have a shipping deadline and a really bad bug got inserted late and detected about a week from the scheduled release. So there's not much time left for testing a fix, another if the first fails, rinse-and-repeat...
Deadlines: "The light at the end of the tunnel is an oncoming locomotive."
Bantam Dominique roosters crow a four-note song. Once you've heard it as "Happy BIRTHday" you can't NOT hear it that way
The length of security patch support on the LTS releases is quite attractive for servers that don't need to be bleeding edge.
Not compared to Debian.
I operate a number of servers running Ubuntu, due to decisions made in the past. Inertia is enough to keep us on the platform, in the sense that I don't object strongly enough to go through the pain of migrating them to another distro. The servers run well enough, I suppose, but there's nothing particularly attractive about running Ubuntu on them.
Where servers are concerned, conservatism is a virtue, and Debian Stable is my favourite brand of conservatism. I find it philosophically unappealing to be running on Testing and/or Unstable (which, effectively, is what Ubuntu is) because the benefits don't outweigh the liabilities. Happily, my servers have behaved well so far, in part because I use minimally simple configurations, I check everything that happens on them all the time and I read the changelogs before I patch.
On the desktop, however, I quite like Ubuntu. Pushing out closer to the edge in order to get better hardware support and cool features really appeals to me, because the promise of an improved user experience makes it worth enduring a few nagging issues.
That said, Lucid and Karmic have a few bugs that are really silly. One recent one is the Edit Network Connections applet which (rightly) disables the 'Apply' button when there's only partial address information, but never re-enables it. This is a really basic programming mistake, and frankly I'm amazed it was never caught. Issues with removable devices have become increasingly bothersome as well. Karmic saw intermittent problems mounting CDs as well as USB disks and flash drives.
Most -if not all- of these issues can be laid squarely at the feet of the GNOME devs, who seem to be making more and more amateur mistakes at every release. I'm starting to wonder if they have any QA & testing environment at all. But Ubuntu has made its bed by tightly aligning itself with GNOME's release schedule, so they get to share the blame.
As a poster just below observed, becoming popular makes you a target for criticism. I don't really see a problem (or a contradiction) there. While I support Ubuntu and suggest it to anyone who asks, I still think that prominence means that they should be prepared to meet a higher standard and to address such criticism effectively.
Full marks to them, by the way, for getting out ahead of this issue. If this were a proprietary OS, we'd likely have to wait for the first Service Pack before this issue was addressed. (And of course, it wouldn't be documented except for numerous blog and forum posts peppered across the Web.)
Crumb's Corollary: Never bring a knife to a bun fight.
Ubuntu developers are now desperate for people willing to test out this updated X.Org Server package so they can determine by this Friday whether to ship it with Ubuntu 10.04 LTS or doing an early SRU (Stable Release Update).
They should have thought that before antagonizing over 80% of the tester community with the windows button issue.
Yes, it IS a petty issue, the problem is that everybody said "We don want it, please revert pretty please" and Mark was like "Thank you, your opinions are very valuable, however, just bite it".
So I'm not surprised at all if the tester community feels withdrawn. There is a growing feeling that the opinions of the community are being soundly ignored, for instance these (public) statements from the bug tracker I'm going to reproduce without permission:
Jef Spaleta:
First of all I think you put too much weight behind Brainstorm as a tool
to drive change inside Ubuntu. You actually shouldn't be at all
surprised that Brainstorm popularity has very little influence over
design decisions. It's never had influence in any technical decision
making and no one in a position of authority inside Canonical or Ubuntu
governance has ever claimed that it has. Canonical nor the external
Ubuntu governance structures make it a policy to rely heavily or to even
officially review highly popular ideas in Brainstorm on a regular basis
or part of technical decision making or public governance discussion.
Were highly popular Brainstorm ideas even discussed in an organized
session during the UDS in the run up to 10.04?
The track record of implemented ideas backs up my point. You look
really closely at the ideas marked implemented in Brainstorm and they
are at best mediocre in terms of Brainstorm popularity. None of the
highly popular ideas in Brainstorm get implemented..or even discussed
publicly as a matter of technical decision making or governance. Take
for example the music store idea. It has a negative voting total and is
marked implemented.
It's wishful thinking to suggest that Brainstorm popularity plays an
important role in decision making. It doesn't. At best brainstorm is a
dumping ground for random ideas. There's no evidence that the voting
process correlates with feature development or decision making at all.
The thing is, Ubuntu has dropped the ball massively with this release, there is simply nothing good about the new release, worse still is that it lost contact with its user base, most of the decisions are now either politically or corporately motivated, or driven by the team of Cupertino rejects that Mark appointed to drive Ubuntu development.
But really, this is interesting, I'll get some marsh mellows and enjoy the fireworks. The question no longer is if Lucid is going to be an embarrassment but whether Mark will learn anything from it. If Mark learns a lesson it's well worth it.
I really loved ubuntu, I want to love it again, but right now, I'm just deciding whether to switch to mint or debian.
But... the future refused to change.
It's funny how when a FOSS project gets to a certain level of popularity (Firefox, Ubuntu) there seem to be a vocal group of people that try to tear them down.
That's kind of a pointless statement. There's also a vocal group of people who talk it up every chance they get and actively shut down anyone who points out that there might be room for improvement.
This author takes full ownership and responsibility for the unpopular opinions outlined above.
Full marks to them, by the way, for getting out ahead of this issue. If this were a proprietary OS, we'd likely have to wait for the first Service Pack before this issue was addressed. (And of course, it wouldn't be documented except for numerous blog and forum posts peppered across the Web.)
Full agreement there. Even certain other players in the Linux market might be less "good" in this respect. While Debian's mailing lists can be a brutal place to exist if you are neither omnipotent nor immortal (or at least flameproof), both their core contributors and the Ubuntu equivalents seem to attribute openness the value it deserves more than most do.
I agree with your sentiment, but I think there is something to be said for small, light software, and that most OSS projects begin small and light for obvious reasons and then mature into huge, bloated pigs like Firefox. People who like small, light software are forced to continuously downgrade to newer, shittier software. This even happened to Scheme, which started out as a programming language whose spec could be printed on a handful of pages but which recently ballooned into three or four documents adding up to about 200 pages. If you believe removed code is debugged code, what's added code?
That said, the main thing I don't understand about the big distros is why they do so much patching in the first place. You often hear about kernel instabilities caused by distro maintainers applying patches that weren't accepted by the kernel team for good reason. Why monkey with stuff you don't understand? This is another reason I've preferred Gentoo and Arch, and, when possible, FreeBSD over Linux.
With Debian Stable I have to install lots of software from source because the supported versions are so pathetically outdated. This causes the system to be... less stable.
LTS release schedules are more stable and less work to maintain because they typically have all the software I need in their supported repositories.
A slashdotter who didn't build his own computer is like a Jedi who didn't build his own lightsaber.
my trust in OSS.
I've been an open source user and developer since long before there was a Linux. And, I've been a Linux user for a long time. Used Redhat, Debian, and now Ubuntu. I've been using Ubuntu since 5 something. I like Ubuntu. It is easy to install, gets easier all the time. It works, which is really nice. And, it has very good support for things like Flash and proprietary graphics card drivers. You can complain that it doesn't have some detail covered that is critical to you, but that's OK. I've been very happy with Ubuntu.
Well, I was. I always try to test the alpha and beta releases. In the early days I could down load the first alpha and it would work. It might get a little weird, but it would work. In the worst case I can remember the computer would at least boot up to the command prompt. That is until the 10.4 release. That just plain wouldn't boot until we got to alpha 3. It wouldn't even install. It has been awful ever since. I don't know if it is a problem with X.org, but every time I type in the search field on firefox I get a black screen. After a few seconds the login screen comes up and I can login. The machine did not reboot. It looks like typing in the search field on firefox is crashing the X server. Now, back in the early '90s I helped get a little program called xcrashme written and distributed and after that was around for a few years the X server was damned near bullet proof. What did they do to mess it up so badly? I went to file a bug report. It turned out to be a duplicate. Seems a lot of people have reported the problem. I haven't seen any action on it.
Then there is the little thing about the user interface in 10.4. Nobody in their right mind, at least no body who had any respect for their users, would change something as basic as the location and order of the window buttons. But, Shuttleworth has done just that. The reason? To make room for a "cool" something that will appear in a later version of Ubuntu. The only discussion involved in the decision was the coolness of the feature and the vague technical argument that somehow it reduces mouse movement, because the buttons are now on the same side of the screen as the menus. Oh, yeah, like the amount of time anyone spends opening new apps is worth retraining your hands to find the new buttons. On the bug discussion list Shuttleworth would not even admit that human factors might have some validity in the discussion. Only the coolness and the bullshit argument about mouse movement were treated as worthy of consideration. Shuttleworth even posted data showing his own mouse movement. The data did not support moving the buttons. But, he claimed it did. He saw what he wanted to see. After all, the new thing is so cool we should all be grateful for the inconvenience.
Why doesn't Ubuntu care about the effect the change will have on their customers? Because they have no customers. They are in it to be cool and to score techie points with other people who do not understand why proprietary software actually tries not to piss off their customers. If you don't believe me ask a human factors engineer why purple is an awful background color for a GUI and then ask what percentage of the public can read light gray text on a dark gray background. Then look at the new Ubuntu default theme. It sure is "cool". I used ssh -Y to log in from a computer with a different theme so I could work select a readable theme and move the buttons back to where I'm used to having them.
The backlash from the users has been astonishing. Even more astonishing is Shuttleworth's "I'm to cool to care" attitude.
At least for now you can move the buttons back and choose another theme. What happens when he puts his uber cool new feature into the UI? I guess I am looking for a new Linux distribution.
That was bad enough... But, then I ran into OO.o Issue #956 (http://qa.openoffice.org/issues/show_bug.cgi?id=956). Have you heard about this one? It was filed May 25, 2001. For comparison current issue numbers for OO.o are now above 110,000.
Still, my original question was appropriate. I’m not asking why they sent that to their developers; I never was. I’m asking why it’s being reported on Slashdot before anything really is known about the problem.
To answer my own question, and this is something I only realised after reading the other comments on this thread... it appears that this is news and it’s being reported on Slashdot primarily because of speculation that it has the potential to delay their hard-and-fast 6-month release cycle date. I’m pretty certain there’s a certain amount of schadenfreude motivating it...
Alexander Peter Kristopeit bought his basement from his mommy for one dollar.