furkan | tjaalton: i just tested out the 20160310 live image, and the bug is reproducible | 01:09 |
---|---|---|
furkan | looking at the diff between the manifests, i don't see any possible culprit other than xorg? | 01:11 |
furkan | tjaalton: looks like this is probably the issue, and there is a fix released in xserver git https://bugs.freedesktop.org/show_bug.cgi?id=94246 | 01:39 |
ubottu | Freedesktop bug 94246 in Driver/glamor "Text shadow is rendered incorrectly on Radeon R7 260X" [Normal,Resolved: fixed] | 01:39 |
furkan | glamor bug | 01:39 |
furkan | if this is it, that should really be backported into a point release | 01:41 |
furkan | these are the 2 patches: | 01:48 |
furkan | https://cgit.freedesktop.org/xorg/xserver/commit/?id=b05ae79ee3bebef9790c97eedc033d1ffb3ec39a | 01:48 |
furkan | https://cgit.freedesktop.org/xorg/xserver/commit/?id=a3e681eafa5355b8bb3b099d47983f14f0d5e197 | 01:48 |
furkan | i'll test them | 01:48 |
furkan | actually it looks like it might depend on some previous patches as well | 01:50 |
tjaalton | furkan: oh right, forgot glamor.. | 05:15 |
tjaalton | furkan: mind filing a bug? | 05:16 |
furkan | tjaalton: sure, will do | 05:17 |
furkan | but i haven't tested the patch | 05:17 |
furkan | did an apt-get source and was manually merging those two patches but noticed some other discrepancies in the code | 05:18 |
furkan | wasn't really in the mood of breaking anything tonight | 05:18 |
furkan | so based on that, it seems there have been other patches to the same functions in question, so it may require a bit of fiddling to backport | 05:19 |
tjaalton | sure | 05:19 |
alkisg | nouveau@riva-tnt2: x11perf -putimagexy500 -eschertilerect500 -repeat 1 | 05:23 |
alkisg | 2000 reps @ 4.4646 msec ( 224.0/sec): 500x500 tiled rectangle (216x208 tile) | 05:23 |
alkisg | 4 reps @ 8389.1916 msec ( 0.1/sec): PutImage XY 500x500 square | 05:23 |
alkisg | ==> how is it possible for a single putimage to need 20 seconds?! | 05:23 |
tjaalton | don't ask me | 05:29 |
tjaalton | where do you get this crap hw? | 05:29 |
furkan | probably from an archaeological dig :D | 05:31 |
alkisg | Greece, that's all we have here :D | 05:31 |
tjaalton | sorry to hear that | 05:32 |
alkisg | For some reason normal operations like menus animations, scrolling windows, watching youtube etc is extremely slow on all the old graphics cards that I've tried | 05:32 |
alkisg | With e.g. windows 2000, it isn't so; it's something specific to the linux implementation | 05:33 |
alkisg | I don't know if it'd be worth it to try and troubleshoot it, to find which code paths are so slow etc... | 05:33 |
alkisg | Difference like from e.g. 1 fps on linux, to 30 fps on windows | 05:33 |
alkisg | Tested in 10 cards or so, so I don't think it's just driver-specific | 05:34 |
furkan | have you tried the proprietary driver (if it's still supported)? | 05:34 |
tjaalton | it's not | 05:34 |
alkisg | Not recently (it would need 8.04), but I did try intel hardware as well | 05:34 |
tjaalton | oh | 05:34 |
tjaalton | nvidia | 05:34 |
furkan | wait what, windows 2000? | 05:35 |
furkan | what kind of CPU do you have? | 05:35 |
alkisg | We're the IT support center of 300+ schools, we have a lot of CPUs available for testing, | 05:36 |
furkan | because i occasionally use windows 7 and 10, and i find that window operations in linux are much more CPU intensive | 05:36 |
furkan | like moving around a window can cause CPU usage to spike to 30-40%, resizing a window can cause 90%+ CPU usage | 05:37 |
alkisg | so currently I'm testing with the windows-2000 -era hardware that they have, we're using them as linux thin clients | 05:37 |
furkan | whereas with windows 7+, CPU usage is negligible | 05:37 |
alkisg | Right, that's what I notice too | 05:38 |
furkan | i think Xorg is just really inefficient | 05:38 |
alkisg | But why? Too many buffers being copied around? | 05:38 |
furkan | i don't know but i think that's part of what Wayland will fix | 05:38 |
furkan | and Mir, on the Ubuntu side | 05:39 |
alkisg | I really can't see why a putimage would need 20 seconds, last I tried implementing that in assembly back in 1995 we could easily do 60 fps... | 05:39 |
furkan | i think there is a build of Ubuntu 16.04 available with Unity 8, maybe you can try it if it's compatible with your hardware | 05:39 |
furkan | Unity 8 uses Mir | 05:40 |
tjaalton | no it won't be | 05:40 |
furkan | but it's a big work-in-progress right now | 05:40 |
furkan | oh, my mistake then | 05:40 |
alkisg | Would e.g. wayland or mir be compatible with intel 855? | 05:40 |
tjaalton | at least I don't think nouveau has any support for ancient hw | 05:40 |
tjaalton | alkisg: no idea | 05:40 |
alkisg | ...or nouveau + geforce 400mx, stuff like that | 05:41 |
tjaalton | i don't know if it needs more than just kms support | 05:45 |
tjaalton | first thing to check is if nouveau is used | 05:46 |
tjaalton | the kernel module | 05:46 |
alkisg | In e.g. 16.04 and geforce 400mx, it is... I'll download a wayland or mir cd to test with... | 05:47 |
tjaalton | with current debian sid on intel skylake & wayland, for instance resizing the terminal is abysmally slow | 05:48 |
tjaalton | it's like 2fps | 05:49 |
alkisg | Meh :) | 05:50 |
alkisg | Another idea could be vnc@directfb to reuse those old clients... maybe it would be faster... | 05:50 |
furkan | alkisg: i wonder if you guys would just be better off buying raspberry pis | 05:50 |
furkan | raspberry pi 3 might be more powerful than your hardware, and consumes much less power | 05:51 |
furkan | not sure how expensive electricity is over there, and how much it would offset the cost | 05:51 |
alkisg | School eletric bills are paid centrally, so they don't care about that | 05:52 |
alkisg | When we do have 100€ per client, we're upgrading P4's with something like ASROCK QC5000M motherboard + 4 GB RAM | 05:52 |
alkisg | (all in one, 1900 passmark score for its kabini cpu) | 05:52 |
alkisg | It's like 100 times faster than raspberry pi 3 | 05:52 |
furkan | ok i see, yeah kabini and 4GB RAM is better than rpi for sure | 05:52 |
furkan | do you use ubuntu with the kabini boxes? | 05:54 |
alkisg | so far we've been using equivalent intel-based boards (with IDE connectors), but they're not available anymore, so we ordered our first kabini one some days ago, it hasn't arrived yet | 05:55 |
alkisg | We'll start testing it in a week or so | 05:55 |
furkan | oh ok, i have a kabini box running ubuntu and it works really nicely | 05:55 |
alkisg | Some teacher reported that his recent amd board didn't work with 64bit ubuntu after 14.10, and he needed 32bit instead | 05:56 |
alkisg | (usb + lan didn't work at all on the live CDs etc) | 05:56 |
furkan | that could be for any number of reasons... i remember i had an issue with an intel machine where the onboard Realtek LAN wasn't working properly, and it's because for some reason the wrong kernel module was being loaded | 05:58 |
furkan | i built the driver from source and it worked fine after that | 05:58 |
alkisg | Haha ok s3virge wins, 4 reps @ 15587.4967 msec ( 0.1/sec): PutImage XY 500x500 square | 06:09 |
alkisg | That's like one putimage per hour :D | 06:10 |
* alkisg is trying to find some ancient graphics card that goes faster in debian wheezy (before xaa was killed) compared to ubuntu xenial... and can't find any!!! | 06:22 | |
tjaalton | furkan: it wasn't too hard to backport the font fixes afterall | 06:48 |
tjaalton | seems to work | 06:48 |
furkan | oh, nice :) | 06:49 |
furkan | did you still want me to file that bug report? i was still going to do it a bit later | 06:50 |
tjaalton | please | 07:05 |
furkan | tjaalton: done https://bugs.launchpad.net/ubuntu/+source/xorg/+bug/1555960 | 07:21 |
ubottu | Launchpad bug 1555960 in xorg (Ubuntu) "Some text (in buttons, dropdown menus) renders incorrectly" [Undecided,New] | 07:21 |
furkan | i could have sworn that somewhere on that page there was an option to indicate that an upstream fix has been committed | 07:21 |
furkan | but i can't find it | 07:21 |
tjaalton | thx | 07:22 |
furkan | np | 07:22 |
furkan | glamor performance is really coming along nicely, it's been getting faster with every xorg release | 07:23 |
tjaalton | i was meant to let skylake fall back to modesetting+glamor but other bugs prevent that | 07:25 |
furkan | i'm looking forward to 1.19, it looks like eric put in some pretty nice improvements while doing his work on the raspberry pi driver | 07:25 |
tjaalton | uploaded | 07:40 |
furkan | to x-staging? | 07:47 |
tjaalton | no | 07:47 |
tjaalton | xenial | 07:47 |
tjaalton | the bugfix i mean | 07:47 |
furkan | hmm maybe it hasn't propagated to my mirror yet | 07:47 |
tjaalton | lucky if it has even built yet.. | 07:48 |
tjaalton | it'll take a day to hit your mirror | 07:48 |
furkan | oh haha | 07:48 |
furkan | btw remember when you said i must have not updated for a quite a while | 07:50 |
furkan | the mirror i was using hasn't been updated since mid-february | 07:50 |
tjaalton | same thing with fi.a.u.c | 07:50 |
furkan | and i think that might be the most popular mirror around here | 07:50 |
furkan | (hosted at a university nearby) | 07:50 |
furkan | come to think of it, i think that might be the mirror being used in our lab, so we might not have been getting security updates for the past month :o | 07:53 |
* furkan ssh's in | 07:53 | |
tjaalton | I hope you use security.u.c | 07:53 |
tjaalton | as everyone should | 07:53 |
furkan | what's that? | 07:54 |
tjaalton | security.ubuntu.com | 07:54 |
tjaalton | see sources.list | 07:54 |
furkan | oh ok | 07:55 |
furkan | checking now, we use an apt-cacher | 07:55 |
furkan | looks like we use the generic ca.archive.ubuntu.com mirror | 07:55 |
furkan | are you suggesting that i add a separate line for security.ubuntu.com? | 07:56 |
tjaalton | no | 07:57 |
tjaalton | should have already | 07:57 |
tjaalton | even if you use a mirror the default sources.list uses security.u.c for security updates | 07:57 |
tjaalton | unless you've fumbled that | 07:57 |
furkan | i'm gonna pastebin this to make sure, i'm not the one who set up this particular server | 07:58 |
tjaalton | i'm talking about the client | 07:58 |
tjaalton | i don't care what you mirror :) | 07:58 |
furkan | oh this server is acting as the apt-cacher for all the computers in the lab | 07:58 |
tjaalton | k | 07:58 |
furkan | http://pastebin.com/M9QwcPJV | 07:58 |
furkan | i don't see "security" anywhere in here | 07:59 |
furkan | well there is trusty-security, but points to ca.archive.u.c | 07:59 |
tjaalton | i don't know how apt cacher works | 07:59 |
furkan | that's just the sources.list file | 07:59 |
furkan | apt-cacher just sets up an HTTP server and serves the cached packages | 07:59 |
tjaalton | then you can hit issues like mentioned | 08:00 |
furkan | ok i see, i'm looking at my sources.list at home and i see the difference | 08:00 |
furkan | deb http://security.ubuntu.com/ubuntu/ xenial-security multiverse main universe restricted | 08:00 |
furkan | i will modify the server, thanks for the heads up | 08:01 |
furkan | basically all the clients have their source.list set up to point to the apt-cacher | 08:01 |
furkan | and the apt-cacher uses its own sources.list to retrieve/cache the packages and serve them to the others | 08:01 |
tjaalton | ok | 08:03 |
tjaalton | I just host a local mirror | 08:03 |
tjaalton | of debian & ubuntu | 08:03 |
=== JanC_ is now known as JanC |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!