Fatdog64-700 beta1

[jamesbond]

Please do not update xorg. Updating xorg means you need to update a few other libraries at the same time: xf86-*, libdrm, and a few others. The new xorg will be in the next release. May be I need to "lock" the xorg package so it is not update-able.

@GKM - you're doing the right thing, but your "append" line in syslinux is wrong (the first = is not necessary). Please tell me the content of /proc/cmdline.

@Ted Dog - theoritically there is no limit.

@James C - probably kernel problem. We're going with updated kernel on next release, so please hang on ...

[Gobbi]

Nice job with qemu-vm-manager 1.1 package , jamesbond

I don't know if it would be not so hard to add an option to the GUI to asign one of the USB devices ( keyboard-mouse kit ) from the host machine to the VM ( guest ) from a list of connected devices provided in Fatdog , wouldn't it

Also , mikeb has an intent to add a sound option to Qemu Launcher , one more step to more a complete interface for qemu ...

[LateAdopter]

Hello jamesbond

I've played this linux media player dependency sudoku before. They seem to want every conceivable network protocol just to play an mpeg video.

For MPV 0.6.0 I'm stuck for one dependency: libcurl-gnutls.so.4.
I seems to need to match the exact version of libgnutls.so.28 that Fatdog already has, or the subsequent dependencies go parabolic.

If it's not too difficult, would you be able to make one that works in Fatdog, please? If it's not convenient, I'll wait for VLC 2.2 to arrive in Fatdog.

Thanks.

EDIT1: I found this post which implied that it was just libcurl built with a --with-gnutls option.
http://forums.kleientertainment.com/top ... ntry488142

I just made a symlink called libcurl-gnutls.so.4 to libcurl.so.4.3.0

That satisfied ldd. MPV complains about missing version information, but does run and play a video

EDIT2: A minor anomaly that I noticed is that libgsm1 is missing but the package manager says the package is installed. I looked in the package but the lib doesn't seem to be there.

[stemsee]

@Gobbi

Assigning a particular keyboard+mouse to a qemu vm is not so easy. I have researched it quite a bit (I even wrote to the original programmer and members of the current development team without making progress) . It is hard to find any detailed documentation, and there doesn't seem to be anyone responsible for coding that. I guess that is why at the moment qemu grabs the entire xinput 'Master' sections with their devices without distinguishing. Perhaps it is possible to create a substitute xinput for qemu to grab.

[Gobbi]

stemsee - Thank you for interest and explanations .
I understand the issue it's more complex at the moment . A taylored xinput maybe it's the way to try ...
jamesbond gave interesting explanations on FD 630 - FD631 thread , on page 18 , but I remember it was about sandbox , at the time .
For the multi-seat , at least VBox allow general asigning USB input for the guest VM . But things are also evolving in qemu...
Sound it's easy to get for qemu from the terminal.
In qemu I never had the cursor issues I often have in VBox ( stopping at some imaginary barrier on some distance from the side of the desktop ) .

[LateAdopter]

Hello jamesbond

I've briefly tested hardware decoding with MPV. VDPAU is berserk, BUT VAAPI WORKS! even though the MPV man page says Intel only for VAAPI.

Decoding 1080i50 H.264 without deinterlacing I get the following results for cpu load:

30% at 2x 2800MHz with software decoding

3%-10% at 2x 800Mhz with --hwdec=vaapi --vo=vaapi

8%-10% at 2x 800MHz with --hwdec=vaapi --vo=opengl

Note --vo=vaapi is not recommended because the onscreen controls do not display.

This looks quite promising for anyone who has an AMD laptop with a RS780 or RS880 chipset.

Now I need to see if I can get any deinterlacing.

[stemsee]

@Gobbi

this is the guy who is expert and maintains the code for spice/sdl in qemu. He wrote a document explaining how to configure multiseat/multihead in a single instance of qemu. I am pretty sure that he could give guidelines how to acheive a qemu instance with one xinput master section.

Gerd Hoffmann <kraxel@redhat.com>

[GKM]

@jamesbond

Thanks, I looked in proc/cmdline, and indeed it contains usbcore.autosuspend=-1 along with where to find my savefile etc, although I did correct another = in the boot parameters. However, I now have reason to believe the cause of the problem may be something else entirely. I have not yet pinpointed the problem, but small modifications of the communication software can reproduce the problem on my other system too. It is probably some timing issue in the code which I need to solve. Thanks for the support though.

Regards,

GKM

[TJK]

I was trying to install e17 and I messed up (corrupted) my fatdog64 700b2 instalation.

maybe the latest xorg update is the problem - it certainly breaks my laptop in similar manner you describe ... just an idea

Reinstalled FD 700b1 and everything was running fine after rebooting many time. Then, I did an upgrade a few minutes ago (for the first time) -- now it won't boot. It appears not able to connect to the X server. So you must be right, rodbrogden,

[kirk]

rodbrogden wrote:
TJK wrote:
I was trying to install e17 and I messed up (corrupted) my fatdog64 700b2 instalation.


maybe the latest xorg update is the problem - it certainly breaks my laptop in similar manner you describe ... just an idea


Reinstalled FD 700b1 and everything was running fine after rebooting many time. Then, I did an upgrade a few minutes ago (for the first time) -- now it won't boot. It appears not able to connect to the X server. So you must be right, rodbrogden,

Don't do that. If you replace xorg-server, lots of other things need to change too. The versions that are in the repo right now are from the development process. Once final get's released, we'll get that all cleaned up. Sorry about that guys.

I've briefly tested hardware decoding with MPV. VDPAU is berserk, BUT VAAPI WORKS! even though the MPV man page says Intel only for VAAPI.

Xine-lib/xine-ui supports vaapi, I think the ones in the repo now are built with that enabled, if not it will be the next time we upload. I tested Xine with vaapi enabled on my laptop with Intel gpu, CPU use actually went up. But maybe you'll have better luck.

[chapchap70]

rodbrogden wrote:
TJK wrote:
I was trying to install e17 and I messed up (corrupted) my fatdog64 700b2 instalation.


maybe the latest xorg update is the problem - it certainly breaks my laptop in similar manner you describe ... just an idea


Reinstalled FD 700b1 and everything was running fine after rebooting many time. Then, I did an upgrade a few minutes ago (for the first time) -- now it won't boot. It appears not able to connect to the X server. So you must be right, rodbrogden,

Don't do that. If you replace xorg-server, lots of other things need to change too. The versions that are in the repo right now are from the development process. Once final get's released, we'll get that all cleaned up. Sorry about that guys.

I had noticed that since 630, my X acts funny on this netbook. I guess others have more serious issues. I think I boot into xvesa but my screen is black after I boot. I close my netbook lid, the computer suspends, then X works normally when I open the lid and come out of suspend mode. I didn't mess with anything but could a solution to other issues be to revert to old X files say from version 620?

[kirk]

I didn't mess with anything but could a solution to other issues be to revert to old X files say from version 620?

That's possible, but not easy. What kind of GPU do you have (Intel/ATI/Nividia)?

[chapchap70]

Code: Select all

# lspci
00:00.0 Host bridge: Intel Corporation Atom Processor D2xxx/N2xxx DRAM Controller (rev 03)
00:02.0 VGA compatible controller: Intel Corporation Atom Processor D2xxx/N2xxx Integrated Graphics Controller (rev 09)

Edit: Was just looking around and there seems to be a bug that pertains to my particular netbook model; I boot into some puppies and it hangs just as described. The fact that X works only after the netbook suspends is weird. Not a big deal for me but just figured you might want to know.

https://bugs.launchpad.net/ubuntu/+sour ... bug/944929

[Gobbi]

I confirm the behavior described by chapchap70 - on my daugther's netbook , an Acer Aspire One : Intel Atom N2800 (x86_64) & Intel GMA 3600 .
I only observed that even I can boot various puppys including Lighthouse and Fatdog - up to 620 , on Fatdog 630 the screen turned black . Closing and opening the lid makes X start , indeed .

[LateAdopter]

Hello kirk

Thanks for the Xine suggestion, but suggesting Xine-ui was a bit cruel, I discovered Gxine after a while which is much saner and gives you the engine console.

Xine-lib/xine-ui supports vaapi, I think the ones in the repo now are built with that enabled, if not it will be the next time we upload. I tested Xine with vaapi enabled on my laptop with Intel gpu, CPU use actually went up. But maybe you'll have better luck.

I've tested it and I think I can explain... But since the clever people who invent this video stuff only speak C++ and never tell you what they have done in plain language, the following contains a lot of inference and speculation, so I may be wrong about any or all of it.

Xine uses VDPAU read back or vaapi-copy, so that it can do postprocessing. This means that it has to copy the images from video RAM to system RAM and that carries a significant processing overhead. The Xine implementation seems to be particularly inefficient as it takes more CPU power than decoding in software.

There two showstoppers with UVD on my hardware currently:
- There is no deinterlacing - presumably due to the frame based output of these older GPUs, but there should be a workaround eventually.
- The GPU locks up frequently when using VDPAU decode with VDPAU output. It locks up sometimes when using VAAPI with VAAPI output. With DPM and UVD enabled the messages show the GPU sometimes locks up at the end of the boot process, presumably when the X-server starts, but still gets to a desktop somehow.

EDIT: I was wrong about lockup during boot. I hadn't realised that the bluetooth stuff initialises long after everything else has finished. Plenty of time to start playing a video.


Here is more of what my experiment has found:

With my hardware Gxine says it is using VDPAU:

Code: Select all

vo_vdpau: vdpau API version : 1
vo_vdpau: vdpau implementation description : G3DVL VDPAU Driver Shared Library version 1.0
vo_vdpau: maximum video surface size for chroma type 4:2:2 is 8192x8192
vo_vdpau: maximum video surface size for chroma type 4:2:0 is 8192x8192
vo_vdpau: maximum output surface size is 8192x8192
vo_vdpau: hold a maximum of 10 video output surfaces for reuse
vo_vdpau: using 3 output surfaces of size 1600x1200 for display queue
vo_vdpau: this hardware doesn't support vc1.
vo_vdpau: this hardware doesn't support mpeg4-part2.
.
.
.
vdpau_set_property: property=0, value=0
vo_vdpau: deinterlace: none
vdpau_set_property: property=0, value=0
vo_vdpau: deinterlace: none
vo_vdpau: set_scaling_level=0
vo_vdpau: disable noise reduction.
vo_vdpau: disable sharpness.
vo_vdpau: skip_chroma = 0
vo_vdpau: background_color = 0

When using VDPAU with Xine-UI there is a lot of image corruption but with Gxine the images are OK so I wonder whether it is really doing it in software.

Gxine fails with VAAPI (and so does MPV with --vo=vaapi-copy)

Code: Select all

libva info: VA-API version 0.36.0
libva info: va_getDriverName() returns 0
libva info: User requested driver 'gallium'
libva info: Trying to open /usr/lib64/dri/gallium_drv_video.so
libva info: Found init function __vaDriverInit_0_35
libva info: va_openDriver() returns 0
gtkvideo: video driver vaapi failed.

Libva opens the driver correctly but gtkvideo fails. I guess vaapi-copy is not implemented in Gallium VAAPI.

VLC 2.1 only does VDPAU read back, and it does not offer VAAPI. The images have a lot of corruption with VDPAU.

MPV is quite comprehensive in the modes it supports:

--hwdec=vdpau --vo=opengl ... offered but does not work with my GPU because it requires field based output.
--hwdec=vdpau --vo=vdpau ... works with low CPU load but locks up frequently.

--hwdec=vaapi --vo=opengl ... works quite well with low CPU load
--hwdec=vaapi --vo=vaapi ... works with very low CPU load but locks up sometimes
--hwdec=vaapi --vo=vaapi-copy ... offered but fails to initialise with my GPU

OpenGL output is what the developers are working on, because it can do postprocessing without copying back to system RAM, but it's in its early stages.

My system is an ordinary desktop PC with a 60Hz monitor, but the scaler chip will accept 50Hz input. I am testing with 1080i50 H.264 video and, if I set the PC video output to 50Hz using XRANDR, there is much less image corruption.

Conclusion: When kernel 3.18, mesa 10.4 and VLC 2.2 are all released, I'll test again.

[kirk]

Gobbi, chapchap70,

I have a Acer Aspire One with an Atom n470, but unfortunately it works fine with that. Fatdog64-620 was the last release to use the "stable" Intel driver. I put stable in quotes because it's been over a year since there was a stable release and anything made in the past 2 or 3 years isn't well supported by the "stable" driver. I've built the last "stable" driver and made a package here:

http://distro.ibiblio.org/fatdog/broken ... 6_64-1.txz

If you want to test it you might want to use a new savefile in case things go wrong. If it helps let me know. It could also be the kernel driver, in which case we would need an older kernel.

Thanks for the Xine suggestion, but suggesting Xine-ui was a bit cruel,

Ha, ha, it's not that bad. When you go to the Xine-ui settings, on the first tab you'll need the select "Master of the known universe" as your experience level to be able see all the options. Probably won't help though . Actually Xine is quite good, if they had a UI that was nearly as good as VLCs we'd probably dump VLC.


EDIT: Forgot to mention, to install a package that you've downloaded, right-click on it and select Install Package.

[chapchap70]

Gobbi, chapchap70,

I have a Acer Aspire One with an Atom n470, but unfortunately it works fine with that. Fatdog64-620 was the last release to use the "stable" Intel driver. I put stable in quotes because it's been over a year since there was a stable release and anything made in the past 2 or 3 years isn't well supported by the "stable" driver. I've built the last "stable" driver and made a package here:

http://distro.ibiblio.org/fatdog/broken ... 6_64-1.txz

If you want to test it you might want to use a new savefile in case things go wrong. If it helps let me know. It could also be the kernel driver, in which case we would need an older kernel.

I made a new savefile for testing and tried replacing files with files in the above package and running your install script. Restarted X like instructed. No change.

However, since my wireless light stays on after booting into 630 and 700, I tried typing "xorgwizard" (without the quotes, of course) even though I was at the black screen then hit the enter key. I then shut the lid until the netbook suspended then reopened. I was not in X and the xorgwizard actually appeared on my screen. Frame buffer (fbdev, I think) and modesetting were the only ones to get screens although the display said the frame buffer was at 0 hz. I don't think I have ever gotten an intel driver to work with this netbook. I am pretty sure 620 was modesetting too and not xvesa like I thought. Anyway, no more black screen when rebooting or restarting X in Fatdog.

[Gobbi]

No change in the previous behavior , here .
The screen is still black after booting , and closing/opening the lid makes X start . Also on "Restart X" the screen remains black until the same closing/opening the lid . Then again I can see the desktop ...

[Sage]

Sorry, bit late into this one and apologies for not reading all the foregoing - busy!
Could not get my bog-std PS/2 Logitech mouse to work, not even with CntrlPnl. Ditto sound for onboard SiS SI7012. Both working with everything else on test at moment.

[L18L]

Some stuff for translators to other languages here:
http://www.murga-linux.com/puppy/viewto ... 1&start=42

Thank you, james and kirk