X.Org 7.4 / X Server 1.5 is running months behind schedule, but now holding up its release is Mesa 7.1 or there the lack of. However, Adam Jackson has pushed out a new X Server 1.5 development release with a few more changes while waiting for this updated Mesa release.
Later this week at OSCON 2008, Intel's Keith Packard will be talking about the Linux desktop when it comes to X.Org, Mesa, and related areas. Unfortunately, it doesn't look like the talk will be that technical (unless of course you haven't been staying up to date on our graphics articles), but today he has provided an in-depth blog posting. In this post, Keith provided his take on the current status of X when it comes to output technologies.
Sparking a heated Sunday afternoon debate, NVIDIA's Aaron Plattner had commited a trivial change to the X Server that resulted in several key open-source X developers becoming disgruntled. Ultimately, this NVIDIA-spawned patch ended up being recalled just hours later.
In May we shared that Multi-Pointer X (or MPX for short) was entering the mainline X server. While it was merged to master that month, X Server 1.5 was already branched out and therefore it won't appear in X.Org 7.4, but it will appear in X Server 1.6 (X.Org 7.5) until next year.
One of Google's Summer of Code projects this year is to bring hardware-based video acceleration to Linux with Gallium3D. The advantage of this design is that the implementation is designed to be universal to any driver using Gallium3D, which for now is largely just the Nouveau driver and an experimental Intel version.
As we mentioned yesterday when X Server 1.5 RC4 was released (only to be outdone by 1.5 RC 5 that same day), this new version of the X Server (and therefore X.Org 7.4) requires building against the latest Mesa code that ultimately will become known as Mesa 7.1. In addition to needing the latest-and-greatest from Mesa, a new version requirement for libdrm is also in place. The Direct Rendering Manager library is now at version 2.3.1 and that or a newer version (to come) will be needed for X Server 1.5 / X.Org 7.4. David Airlie's change-log for libdrm 2.3.1 just says "Stuff changed - you need this for Mesa 7.1, and Xorg 1.5 Deal with it." No GEM or TTM code is present in this DRM point-release. However, Intel will be rolling the Graphics Execution Manager into libdrm 2.4 very shortly.
The much-delayed X Server 1.5 RC4 (v220.127.116.114) release made it out today, but it's already being replaced by a newer version. No, there wasn't a big development marathon today or anything spectacular, but X Server 18.104.22.1684 got pushed out with a show-stopping bug. The XInput ABI (for the non-programmers, the Application Binary Interface) number went with version 3.1, when it was supposed to be 2.1. Keith Packard had also squeezed in a change for wrapping AddTraps in EXA and Damage.
The development cycle for X.Org 7.4 has stretched on for months longer than anticipated, and earlier this month it sounded like X.Org 7.4 would end with a quick release, but still it's being dragged out longer. This morning the release manager for X.Org 7.4, Adam Jackson, had announced the release of X Server 22.214.171.1244. This is another development release -- equivalent to a Release Candidate 4 -- before X Server 1.5.0 is ready for X11R74.
The last day and a half has been filled with X.Org articles at Phoronix... Yesterday was X Server 1.4.1, today was X Server 1.4.2, and in just a few days we will now see X Server 1.5.0. X Server 1.5.0 and X.Org 7.4 were expected to be released in tandem earlier this year, but its February then May release date had slipped with no official word of when a release would be out. There were a total of four development releases planned prior to X Server 1.5.0, and we've only had two of them so far. However, Adam Jackson has made it known on the X.Org mailing list that X.Org 7.4 will soon be released.
It took more than 200 days for X Server 1.4.1 to be released, but X Server 1.4.2 is coming just about 24 hours after yesterday's release 1.4.1. This release is coming along quickly as Matthieu Herrb had shared multiple vulnerabilities in the X Server extensions. Three of these CVEs have to do with the RENDER extension while the other two fixed in X Server 1.4.2 are for memory corruption with the RECORD and Security extensions and an MIT-SHM arbitrary memory read. The X Server 1.4.2 release announcement can be read on the xorg mailing list.
Earlier this month we reported at Phoronix that Multi-Pointer X was going mainline with the planned merge to master date being the last week of May. Peter Hutterer has remained on track with this goal and as of last night, Multi-Pointer X (or MPX for short) can now be found in the X Server master branch. However, as Daniel Stone was busy with the Debian-OpenSSL fiasco (and maybe working on X Server 1.4.1), the new version of XKB has yet to reach master.
Amid the on going discussion right now surrounding the TTM memory manager, Intel's Keith Packard has announced another new project that picks up another acronym. The Graphics Execution Manager, or gem for short, that as Keith explains, "It takes the lessons we've learned from TTM and constructs just the API we need to implement the dri_bufmgr interface." That may not sound interesting to an end-user, but through the use of this Intel Graphics Execution Manager, the Intel i915 Linux performance has been improved by 50% in OpenArena and glxgears is running 60% faster. However, this Intel gem doesn't take advantage of features found in newer Intel IGPs (GMA 3000 / i965 series) quite yet. In Keith's mailing list announcement he goes on to explain the technical workings of the Graphics Execution Manager as well as its API.
If you're interested in the internal workings of X.Org and Linux graphics drivers, you may want to read the latest discussion going on the DRI mailing list that concerns the TTM memory manager. Thomas Hellstrom of Tungsten Graphics had asked what is stopping TTM from going in the mainline kernel, which led David Airlie to chime in on its current lack of open-source drivers utilizing this memory management system (really just the Intel driver at this point) and thoughts from other developers. Some feel TTM is too oriented towards its usage on Windows and they have other technical hard-feelings towards this Tungsten Graphics creation. Read the thread on dri-devel.
The 2008 X Developers' Conference wrapped up today at the Googleplex in California. With it being the last day of the conference, there isn't much information to share judging by the XDC Notes, compared to the first two days. There is, however, tentative information regarding this year's X Developers' Summit. XDS 2007 was in Cambridge (UK), but XDS 2008 will be taking place at the conference facilities found inside the Edinburgh Zoo. This X.Org conference is tentatively scheduled from the 10th to 12th of September. Fitting well with Linux developers, this Scottish zoo does have both King and Gentoo Penguins.
The second day of XDC 2008 was filled with quite a bit of technical information (as you can see from the Wiki notes page) ranging from the DRI2 infrastructure and talking about its event ring buffer and simpler DDX driver requirements to discussing video playback APIs. However, there are a few bits of information that are relevant to the community at large.
As was mentioned yesterday, the X Developers' Conference (XDC) is taking place this week at the Googleplex. There's only around 50 developers in attendance, but for everyone else there still is a flow of information to gather. A few Google employees had recorded each of the talks from the first day, however, they aren't yet available on You Tube or Google Video. Unfortunately, it doesn't look like Google will be recording the X videos for the second and third days of this conference. Red Hat's Adam Jackson, however, has been summarizing the highlights of each talk in near real-time and has been uploading it to the X.Org Wiki.
The 2008 X Developers' Conference (XDC) is kicking off this morning at the Googleplex in Mountain View, California. The XDC (not to be confused with XDS, the X Developers' Summit) is made up of a variety of presentations and lightning talks surrounding X.Org and upcoming work done by this important free software project.
Back on the 4th of February, Kristian Høgsberg began merging his new DRI2 components. This initial DRI2 (Direct Rendering Infrastructure 2) work was made up of the DRI2 module, glxdri2 to GLX, and DRM/Mesa patches. Today Kristian has merged his last major part of DRI2 and that is the direct rendering support. With this new code, it's now possible to do directed rendering to redirected windows! This directed rendering to redirected windows support even works with Compiz and other OpenGL window managers using the GLX_EXT_texture_from_pixmap extension. However, to reach this state the DRI interface was broken. Kristian's announcement can be read on the DRI devel mailing list. The only X.Org video driver using DRI2 right now is the Intel batchbuffer branch for xf86-video-intel.
Thanks in large part to the release leadership of Adam Jackson, a new development X server is now available that ultimately will become known as X Server 1.5.0. This new xorg-server release, v126.96.36.1991, has a plethora of changes ranging from fixing memory leaks to EXA improvements. In total there are hundreds of changes to be found within xorg-server 188.8.131.521, with the complete list being available in the X.Org mailing list. More information on the plans for X.Org 7.4 can be found in this article.
With Red Hat's Adam Jackson being interested in shipping X.Org 7.4 for Fedora 9, he has stepped up as becoming the release manager of this next X.Org release and has put together a schedule that will allow this to become possible.
While the XvMC (X-Video Motion Compensation) extension is reliable for offloading MPEG video decoding to the GPU, its limitation is that it only supports MPEG video formats and nothing more. We had expected XvMC not to be around much longer, since Intel has been devoting resources in creating a new video extension for X.Org. This new video work of Intel's is known as VA-API, or Video Acceleration API, and is still quite early in development. VA-API, however, will be able to handle offloading more tasks along with support for all of the latest video standards (MPEG-4, H.264, VC-1, etc). VA-API is not based upon XvMC but is written from scratch.
Remi Cardona is currently talking at FOSDEM 2008 in the X.Org development room about Metisse. For those that don't know, Metisse is an open-source X Window System that was conceived as a French research project (and is incorrectly considered competition to Compiz). Remi's talk is on the features of Metisse, various desktop demonstrations, and bringing Metisse and X.Org together. Some of the items he shared for the Metisse road-map include adding a configuration interface, Expose-like features, FreeDesktop.org integration, and then X.Org integration. This research project is also looking at collaborative features such as window sharing between computers or other mobile devices, etc.
Not that you really would have expected X.Org to ship next month as planned, but X.Org 7.4 looks like it may now be released in May. The original plan for X.Org 7.4 was to release it in February, but that was when X.Org 7.3 was planned for August though it didn't make it out the door until September. With a six month release cycle, March was the new strike point for this X.Org update. We're now running into late February and X Server 1.4.1 hasn't even been released yet, which is a maintenance update that was supposed to arrive last November. There are still five bugs blocking X Server 1.4.1. Red Hat's Adam Jackson has updated the 7.4 Wiki page to reflect that X.Org 7.4 is now planned for a May release. Hopefully that time-frame is met, but that could turn into a June release. If meeting their ideal six month release cycles, X.Org 7.5 would then be released in November or December, but because of the holidays that is likely a Q1'09 release.
Kristian Høgsberg, the creator of AIGLX, has announced on the X.Org mailing list that he is beginning to merge his new DRI2 components. This work includes the new DRI2 module, glxdri2 to GLX, and DRM/Mesa patches. The DRI2 work is ready for early adopters, but is able to safely reside next to XF86DRI. With an Intel DDX patch in the intel-batchbuffer git branch, the xf86-video-intel driver will have a DRI2 option to manually enable the DRI2 module.
For those not satisfied by RandR 1.2 or just wish to live on the cutting-edge of X.Org developments (like us), this week on the X.Org mailing list has been a discussion among driver developers surrounding RandR 1.3. RandR 1.3 is the next update to the Resize and Rotate (RandR) extension that allows for resizing, rotating, and reflecting of the X screen. With the RandR 1.2 update it had introduced display hot-plugging support. When it comes to features, RandR 1.3 is most notably expected to introduce GPU object support. This GPU object support is another layer between the X screen and the CRTCs. Ultimately, this should allow multiple GPUs to be merged into a single X screen.
It's coming a month late, but today X Server 1.4.1 has been pre-released. The xorg-server is now at version 184.108.40.206 and contains over two dozen fixes since X Server 1.4.0. The X Server 1.4.1 pre-release notes and download link is available from the mailing list announcement. On the xorg-server 1.4.1 blocker bug are still five outstanding bugs.
Sascha Hlusiak has announced the release of xf86-input-joystick v1.3.1, which is for providing joystick input support on X.Org 7.3. What makes this joystick driver release newsworthy is that it now provides evdev integration. If you're not familiar with evdev, it's the generic input driver for X.Org, and commonly is used for accessing USB input devices. This new joystick driver supports evdev but has fall-back support for the Linux joystick interface. Thanks to the evdev support, the xf86-input-joystick driver can now be auto-loaded with hot-plug support through HAL. A few other changes also make up the xf86-input-joystick v1.3.1 release, which is detailed in the release announcement.
There's still one bug left before the release of X Server 1.4.1, but Daniel Stone has announced on the X.Org mailing list that X Server 1.5 will no longer contain any "input hotness". Specifically, XKB 2 and Xi 2 were planned but that has been postponed until at least X Server 1.6. This delay has occurred because these changes will require quite a bit of work with how input events are processed and related MPX changes. Expect new X.Org input hotness then towards the end of 2008.
X Server 1.4.1 was originally slated for release on the first of November, then Daniel Stone had pushed the release back to November 11, but that didn't happen either. The X Server 1.4 Wiki page still reflects a release date of November 11, but that's passed by nearly a month.
855 X.Org news articles published on Phoronix.