Announcement

Collapse
No announcement yet.

Red Hat Enterprise Linux 10 Dropping The X.Org Server Except For XWayland

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • mSparks
    replied
    Originally posted by oiaohm View Post

    24 commits on the xserver branch vs over 200 in xwayland.

    And you think that is a "good thing" for xwayland?

    Originally posted by oiaohm View Post
    Lets say we say Redhat staff had to instant stop. Xwayland can progress forwards slower
    xorg-server will continue to work just fine, powering the worlds fastest supercomputers, and providing a flawless desktop experience.
    Wayland and Xwayland will continue to barely work at all, not be used for anything productive, and ignored by mostly everyone.

    Pretty much the same as if RH continue doing exactly what they have been doing.

    Originally posted by oiaohm View Post
    issue is xserver you are only seeing Redhat and Apple personal approving merges.
    The following people are in charge of directing changes
    https://www.x.org/wiki/BoardOfDirectors/
    ?Emma Anholt Google emma AT anholt DOT net Q1 2024
    ?Mark Filion Collabora mark DOT filion AT collabora DOT com Q1 2024
    ?Ricardo Garcia Igalia rgarcia AT igalia DOT com Q1 2024
    ?Arkadiusz Hiler CodeWeavers arek AT hiler DOT eu Q1 2025 Treasurer
    ?Christopher Michael Igalia cmichael AT igalia DOT com Q1 2025
    ?Lyude Paul Red Hat lyude AT redhat DOT com Q1 2025 Secretary
    Alyssa Rosenzweig Valve alyssa AT rosenzweig DOT io Q1 2024
    Daniel Vetter Intel daniel.vetter AT ffwll DOT ch Q1 2025
    Originally posted by oiaohm View Post
    like at moment all CVE notices that is the 90 day heads up of a security problem in x.org xserver or xwayland in fact go to Redhat and nobody else.
    Trend Micro Zero Day Initiative found the last few,



    RH maintain xwayland-server, why are you surprised they are the ones notified about it?



    But those fixes aren'y going in to any of the RedHat distros, only distros that keep their software up to date. FC39 is still on 1.20.14 with all those "CVEs" still active, RH10 will be based on FC40 with is 1.20.14.






    Leave a comment:


  • oiaohm
    replied
    Originally posted by mSparks View Post
    For example?

    Because I only see Peter Hutterer - Senior Software Engineer - Red Hat and, Olivier Fourdan - Senior Software Engineer - Red Hat, José Expósito Red Hat consultant, Michel Dänzer – Principal Software Engineer – Red Hat

    making xwayland/ commits on that link.

    With some "special thanks" to Kenny Levinsen for their help in doing the things they couldn't.
    This you cherry picked again without making sure you were comparing apples with apples.


    Take both look at the commits from 1 Jan 2023 to 13 dec 2023.

    24 commits on the xserver branch vs over 200 in xwayland.

    There at least 8 times more commits into xwayland than xserver branch in the same time frame. So there is a lot more pages to read on the xwayland side to be sure you are getting the correct picture.

    mSparks you did not see any of the Sultan Alsawaf patches did you because you did not scroll back enough or you did not read the merge requests to see who in fact merged the items..
    TearFree with the modesetting driver suffers from inaccurate PresentCompleteNotify event timing, which leads to A/V de-synchronization with clients utilizing Present, among other problems. This turns out...

    Those merges Sultan Alsawaf into Xwayland that get interesting.
    Martin Roukala is not redhat instead is Valve. Yes it was Martin Roukala who is not Redhat who commited Sultan Alsawaf patches into xwayland. The more you look at that over 200 commits the more cases you find new code going into xwayland without going though Redhat staff.

    Yes redhat staff is faster at it most of the time than the alternative like Martin Roukala from Valve. So yes redhat is the dominate party approving merges because they are full time and can do it faster but they are not the only who are merge items into xwayland.

    Issue is xserver you are only seeing Redhat and Apple personal approving merges. Xwayland if you look you find Valve and other companies in the last 12 months approving merges without having to get Redhat rubber stamp..

    Lets say we say Redhat staff had to instant stop. Xwayland can progress forwards slower because there are source code maintainers who are not Redhat with authorization to merge items with the need skilled to review the items correctly. Xserver would in fact grind to a dead stop with only MacOS/Apple stuff getting in.

    Of course there is going to be other problems like at moment all CVE notices that is the 90 day heads up of a security problem in x.org xserver or xwayland in fact go to Redhat and nobody else. MacOS will get notified if CVE is directly about Xquartz.

    This is a big thing no Oracle/SUSE/CIO/Debian/distribution maintainer person get notified about new x11 protocol bugs found or x.org xserver source code issues from CVE. Only redhat core x.org/xwayland people get notified about this stuff.

    To get a reasonable idea what going on with open source project you normally have to look at 12 months worth of commits otherwise you will risk getting a very wrong idea.

    Yes 12 months of xserver commits vs 12 months of xwayland commits the xserver list is very short. xwayland list gets quite lot of reading to be sure you understand what is going on. Mostly due things like who was the party who merged the commit not being correctly shown at the overview.

    Leave a comment:


  • mSparks
    replied
    Originally posted by oiaohm View Post

    Except there is a problem.



    There quite a list of developers committing code to xwayland that are not commiting code to xserver.
    For example?

    Because I only see Peter Hutterer - Senior Software Engineer - Red Hat and, Olivier Fourdan - Senior Software Engineer - Red Hat, José Expósito Red Hat consultant, Michel Dänzer – Principal Software Engineer – Red Hat

    making xwayland/ commits on that link.

    With some "special thanks" to Kenny Levinsen for their help in doing the things they couldn't.

    Leave a comment:


  • oiaohm
    replied
    Originally posted by mSparks View Post
    When I pointed out only redhat has been working on xwayland therefore it is wayland and xwayland that will become adandonware when RH is closed down you went off on a rant about how it isnt only a redhat show didnt you.
    Except there is a problem.



    Lets say we look at the xwayland commits. Back in october there is a commit from a rockchip developer for xwayland but there is no rockchip comments in the xserver branch at all.

    You have Xaver Hugl from KDE you have Gnome developers....

    Its not a rant that xwayland is not a redhat only show. There quite a list of developers committing code to xwayland that are not commiting code to xserver.

    The commit logs of xserver vs xwayland. By the parties commiting code the one that is going to be worse effected if Redhat disappears is xserver. Xwayland could be picked up by the KDE developer who is not paid by Redhat. Who works on the wayland to xwayland backend. Or maybe Kenny Levinsen from wlroots again not Redhat fund but still a full time funded programmer.

    I am not saying xwayland losing redhat is not going hurt it will but no where near as much. But the reality is 60% of the commits to xwayland core are not from Redhat people but xserver since 21.1.1 90% of core engine work is are from redhat people. And the majority of the 10 percent left over that not Redhat with xserver is apple.

    Xwayland losing redhat is survivable because those who write the core code for xwayland 60% of that code was not written by Redhat developers instead by KDE/Gnome/wlroots/rockchip..... Now you move over to xserver the core code 100 percent written by redhat since 21.1.3. Yes xwayland can just swap the maintainer and move forward with skilled developers who know the internals.

    Here is the absolute kicker. Those with the skill sets to maintain xwayland cannot transfer to xserver at the last min because xwayland does not use ddx so has a totally different core design to xserver bare metal.

    xwayland and xserver might be based off the same tree but internal engine is very different.

    Remember you will be depending on Redhat developers who currently working on xserver to be employed by someone else to keep it going if redhat fails or redhat stops support for xserver. Xwayland there is a set of staff outside redhat who can take over that you can see in the commits.

    Xwayland commit layout matches what openoffice looked like before the fork making libreoffice when orcale tried to take openoffice in a direction not wanted so not great but not disaster level if redhat goes away. Xserver commit layout matches projects that have died when backing company goes away so disaster level if something does not change. Yes people with those project that died always claim XYZ parties who have no active developers in the project will take over who end up not doing that as expected because they don't have the skill set to take over.


    Leave a comment:


  • mSparks
    replied
    Originally posted by oiaohm View Post

    Really I said the files being the tar.gz, tar.bz2, tar.xz, and sig at that location you pointed to was created by Redhat developers and it true. Only redhat developer has had permission to add a file to the
    Its one thing to accept changes to a repo, quite another to build an OS, and another thing again to build a cutting edge OS that is also suitable for production. RH can't anymore, dropping X11 is a symptom of that, so is losing all the HPC machines.

    So we see commits by nvidia to xorg server that havent been taken by RH, then wonder why nvidia runs like shite on RH.
    Originally posted by oiaohm View Post
    So you are changing for the distribution I have used for the past 2.5 decades
    Considering Debian 12 was only released 6 months ago, dont see how that can be any more true than the rest of your hyperbole.

    That you joined phoronix Mar 2017 also gives you away.

    Originally posted by oiaohm View Post
    21.1.3 forwards work on xserver has fairly much been redhat/apple show.
    Like it or not what the commits say is many developers were only submitting code to xserver because they were working on xwayland
    When I pointed out only redhat has been working on xwayland therefore it is wayland and xwayland that will become adandonware when RH is closed down you went off on a rant about how it isnt only a redhat show didnt you.
    shrug. full circle.
    Originally posted by oiaohm View Post
    Yes you don't see any Nvidia work on x.org xserver in 21.1.1 or newer x.org server.
    And that is great news, finally a build of Linux that actually works.
    Last edited by mSparks; 13 December 2023, 07:42 PM.

    Leave a comment:


  • oiaohm
    replied
    Originally posted by mSparks View Post
    https://lwn.net/Articles/874152/

    A. Wilcox (unknown)
    Aaron Ma (unknown)
    Aaron Plattner (nvidia)
    Adam Jackson​ (redhat)
    Adam Richter (freelance)
    Alan Coopersmith (Oracle)
    Alex Goins (nvidia)
    Carlos Garnacho (gnome)

    ... OK, bored now, clearly your typical BS
    Really I said the files being the tar.gz, tar.bz2, tar.xz, and sig at that location you pointed to was created by Redhat developers and it true. Only redhat developer has had permission to add a file to the

    location for over a decade now. This statement is true like it or not. What you just point to is that x.org xserver has other people working on it. That does not mean you are going to get release files. In fact Povilas Kanapickas for 21.1.0 did not create the tar.gz/tar.xzfor 21.1.0 that was there it was redhat developer who extracted and made those files from the git repository. Yes the redhat security maintainer working on xserver. This are the little jobs that maintainers need todo that if they are not done will create problems.

    Povilas Kanapickas 21.1.0 release of xserver merges in lots of patches that had not merged 10 May 2018 when redhat removed their primary maintainer. ​ We only have a Redhat security maintainer because no body paid Povilas Kanapickas. You will notice new or reworked features is appearing in xwayland and none is appearing for xserver this is due to the class of maintainer.

    Also since Porilas Kanapickas has stopped being X11 server maintainer the diversity patch authors have dropped. Like not a single nvidia patch into x.org server since he stopped.

    There is a proposed change to X11 protocol to support HDR and SDR applications mixing with each other. Guess what xwayland looks like it going to get it and xserver is going to miss out. Why no maintainer in the class of Povilas Kanapickas to look at new features and merge them.

    Originally posted by mSparks View Post
    It did a bit before this conversation, but from the testing I've done on Debian 12 and fact checking you it doesn't any more. Very much looking forward to upgrading from FC39 to Debian 12 or 13 as soon as I get the chance.
    So you are changing for the distribution I have used for the past 2.5 decades. I have told you before I don't use Fedora or Redhat myself so writing stuff like this thinking it will upset me is another thing.

    How I am aware that one of the security updates xserver broken Nvidia drivers on Debian is because I use Debian. Debian automated build and testing system does not test with closed source drivers mSparks running Nvidia card with Debian can be path to hell. Fedora/Redhat does in fact test with closed source drivers to a point.



    Really you need to look at above mSparks.
    21.1.3 forwards work on xserver has fairly much been redhat/apple show.

    Yes you don't see any Nvidia work on x.org xserver in 21.1.1 or newer x.org server. The 21.1.0 just at the point of the Xwayland split. So developers were submitting to xserver to get alterations for xwayland and their patches do appear going into xserver.

    21.1.1 is after the xserver/xwayland split is formally merged and there is a sudden drop off in the diversity of parties committing to xserver at the 21.1.1 that has not recovered. Xwayland see no such drop off. The gnome developer you mentioned since 21.1.1 xserver has pushed alterations into Xwayland but zero into xserver bare metal.

    Like it or not what the commits say is many developers were only submitting code to xserver because they were working on xwayland so had to and once they did not have to submit to xserver to get alterations into xwayland they stopped working on xserver tree and sending commits for it.

    Leave a comment:


  • mSparks
    replied
    Originally posted by oiaohm View Post
    So this is not true. Reduced staff means cost cutting not loss of the skilled staff capable of building..
    Protip, quality developers can afford morals, only low quality developers stick around whatever they can get regardless.

    Originally posted by oiaohm View Post
    https://www.x.org/releases/individual/xserver/
    Every single one of those files in that location was created by Redhat personal for over the past decade.
    Picking one at random..



    A. Wilcox (unknown)
    Aaron Ma (unknown)
    Aaron Plattner (nvidia)
    Adam Jackson​ (redhat)
    Adam Richter (freelance)
    Alan Coopersmith (Oracle)
    Alex Goins (nvidia)
    Carlos Garnacho (gnome)

    ... OK, bored now, clearly your typical BS

    Maybe someone from AMD can heard those cats.

    Originally posted by oiaohm View Post
    mSparks what is going on should worry you.
    It did a bit before this conversation, but from the testing I've done on Debian 12 and fact checking you it doesn't any more. Very much looking forward to upgrading from FC39 to Debian 12 or 13 as soon as I get the chance.

    Then I can finally get a proper FMOD studio install working, bored of waiting for Fedora/redhat to add support for modern development software.
    Last edited by mSparks; 13 December 2023, 01:57 PM.

    Leave a comment:


  • oiaohm
    replied
    Originally posted by mSparks View Post
    Nope. RH didn't write the code, they did approve some changes as maintainer of the repository.:​
    The handling of appending/prepending properties was incorrect, with at least two bugs: the property length was set to the length of the new part only, i.e. appending or prepending N...

    Except that not true. Redhat developers did write the code. Majority of the new code in xserver is Redhat written code that has not been tested before being commited to xserver to in fact work in the baremetal code base.


    They are current building xwayland right up to current version. So that what Redhat developers have been testing.

    Think about if xQuartz maintainer takes over xserver maintainership would you expect bare metal X11 server code to remain well tested on anything other than Mac OS platform... This is the problem we have. Xwayland maintainer-ship Redhat is interested in. They spun this off xWayland like xQuartz so they would make way for those who had interest in xserver bare metal working. Big problem one person who had interest in the xserver maintainership stepped forwards in total and did not get any funding from any distribution to keep on doing it.
    .
    Originally posted by mSparks View Post
    Probably because most all the redhatters capable of building software left through 2021.
    So this is not true. Reduced staff means cost cutting not loss of the skilled staff capable of building..

    https://www.x.org/releases/individual/xserver/
    Every single one of those files in that location was created by Redhat personal for over the past decade.

    mSparks what is going on should worry you. Redhat maintainer of xserver sees that current Xwayland builds and works so it fine to release a new xserver bare metal untested because we must not have done anything to break xserver. At some point this logic is going to be wrong.

    Originally posted by mSparks View Post
    Debian does all their own testing,​
    Good question why does not Debian Maintainer or some other maintainer doing testing take over the xserver maintainer role. There is a problem Debian does all their testing after package is versioned and released by upstream. To have a quality upstream you want testing before new release is versioned and released to public.
    Originally posted by mSparks View Post
    So does SUSE, Oracle just started with openELA​
    Again after upstream release not before.

    Originally posted by mSparks View Post
    There has been a dozen or so releases since any redhatter tested anything.​
    This is right and wrong. Redhat has been testing xwayland and if that not broken it good enough to send xserver baremetal package into the world for anyone else to deal with.

    Yes years now of the upstream maintainer of xserver not doing proper testing as this saves Redhat money and as you note this has been going on for years.

    The fact Redhat is not maintaining current version of xorg-xserver in fedora should have someone from some other distribution stepping forward to take over the maintainer position. Redhat has allowed a different person to take over with the 21.0 release and the offer for anyone to take over the xserver maintainer-ship at x.org is still open mSparks.

    Also there is another problem you see there is not enough maintainers in xserver or xwayland. You see the lead maintainer signing off his own patches to be commit-ed.

    If you look at a well running open source project package maintainer doing a patch need someone else to sign off that their patch is not broken. xserver and xwayland this not happening. At Weston, Sway, KDE wayland, Gnome Wayland secondary signing of patches is mandatory.

    mSparks; you ware way to fast to cherry pick and not step back and look at the big picture.

    By the way there have been a few breaks of the nvidia DDX module that have traced to patches that work just fine in xwayland that have been merged but in bare metal X11 server disrupt the DDX interfaces. This would have been detected before uses end up with broken Nvidia drivers if upstream had been running the DDX test suite. Debian/Ubuntu package maintainer did not pick that up.

    Redhat has proven they can sabotage upstream Xserver and it go not noticed until the broken server lands on end users. So it kind of important that xserver maintainer at x.org does get replaced someone with interest in keeping it working if everyone is not moving over to xwayland.




    Leave a comment:


  • mSparks
    replied
    Originally posted by oiaohm View Post

    Who did not put .21 though their fedora automated testing servers... That would again be Redhat.
    Nope. RH didn't write the code, they did approve some changes as maintainer of the repository.

    Originally posted by oiaohm View Post
    What you quoted does not nuke my theory. In fact it supports my theory.
    It would be a different matter if xserver .21.9 was in fact released by a debian developer or some other company.
    But the xserver release are coming from Redhat
    Nope.
    Redhat/Fedora only ship 1.20, that they copied from x.org December 2021

    Probably because most all the redhatters capable of building software left through 2021.



    There has been a dozen or so releases since any redhatter tested anything. Debian does all their own testing, its why they have stable/testing/unstable versions.



    So does SUSE, Oracle just started with openELA afaict, although their UEK build is already waaaay more popular than any RH distro.

    that's what I mean by wayland is a symptom not a cause, RH lost all their competent staff, X11 is to hard for the ones that are left - it is not to hard for competent developers, they just work for other companies now.

    Its one thing to accept changes to a repo, quite another to build an OS, and another thing again to build a cutting edge OS that is also suitable for production. RH can't anymore, dropping X11 is a symptom of that, so is losing all the HPC machines.

    RH10 isn't going to attract any money to IBM to pay the salaries of the ones who couldnt find jobs elsewhere. It is over, just not for X11, and definitely not for the HPC that is built on it.

    Plus, IBM is about to lose another CEO
    Last edited by mSparks; 13 December 2023, 01:07 AM.

    Leave a comment:


  • oiaohm
    replied
    Originally posted by mSparks View Post
    CIQ, Suse and Oracle are the main companies that have actually been building the worlds fastest super computers since ~ RH8.

    You know them now as the companies behind openELA.
    Yes and they are the 3 parties who have not been investing in xserver development so leaving the work up to Redhat.

    You find that those 3 companies have depending on Redhat to perform lot of work for them for free. Problem is with the clients Redhat has paying they don't depend on stacks of software that the CIO, Suse and Oracle customers do that Redhat has been maintaining.

    Xserver is just a tip of a very big iceberg. As Redhat has to cost cut removing packages their customers don't need because they don't have customers in that area CIO, Suse and Oracle need to step up.

    Redhat restricting access to their source that a lot of people take as a bad thing is more a here is warning pressure unless you can stand on your own too feet if we die from you taking our market share you are going to be in trouble.

    Originally posted by mSparks;n1427790nb
    redhat is already well behind debian
    v .20 for FC40
    vs
    v .21 for trixie and bookworm


    Except who in fact released the upstream v.21 xserver that would be Redhat right.

    Who did not put .21 though their fedora automated testing servers... That would again be Redhat.

    Originally posted by mSparks View Post
    which pretty much nukes your theory that RH played any role in xorg server in the first place, or that X11 can't survive without them.

    No this is you cherry picking and miss reading the location.

    Redhat has been progressively reducing their quality control on xserver source code releases(this is not the only project). Sooner or latter major bugs will start slipping though.

    Another possibility thinking that Redhat developers are the upstream and FC40 is still on .20 do they know that they have intentionally added a defect into .21.

    What you quoted does not nuke my theory. In fact it supports my theory. The problem you never checked out who in fact upstream released the current xserver version.

    It would be a different matter if xserver .21.9 was in fact released by a debian developer or some other company. But the xserver release are coming from Redhat and then all distributions are building their packages from there. Redhat is the upstream.

    Redhat market-shares says that Redhat should not be x.org xserver upstream developer any more other parties that depend on xserver for their solution to suite their customers should have taken over by now. So far it looks like no one is going to step up so you will at some point be forced to make do with xwayland or at worse no xserver of any form when Redhat fully steps away.

    Leave a comment:

Working...
X