=== mvo_ is now known as mvo | ||
Plagman | Hi. | 14:42 |
---|---|---|
Plagman | I'm looking for someone who knows anything about 107_fedora_dont_backfill_bg_none.patch | 14:44 |
Plagman | I suspect it's only there to help with Xgl, which is largely unused nowadays | 14:44 |
Plagman | Pretty much every driver provides its own texture_from_pixmap implementation | 14:44 |
Plagman | This patches introduces graphical corruption with compositing enabled; we've had quite a bit of users complaining about it | 14:45 |
Plagman | s/patches/patch | 14:46 |
superm1 | Plagman, I can't comment on the patch particularly, but the people who would be responding to it later would probably ask, we being Ubuntu users? | 14:46 |
tjaalton | Plagman: it's still shipped by fedora, and was added to the package a long time ago, so no-one here knows the details why it was added | 14:47 |
Plagman | we equals NVIDIA X driver team | 14:48 |
Plagman | People like to say it's our bug as well. :/ | 14:49 |
tjaalton | Plagman: any way to reproduce with 177.80? | 14:50 |
Plagman | sure | 14:50 |
Plagman | you'll need a compositing manager running | 14:50 |
tjaalton | check | 14:50 |
Plagman | basically, any newly created window should briefly flash black | 14:51 |
Plagman | (or random colorful corruption if you're using InitialPixmapPlacement=2) | 14:51 |
Plagman | if you start firefox, you should get a black square before the first expose | 14:51 |
superm1 | i've seen that particularly with bigger windows like FF on my nvidia cards but never thought much of it | 14:51 |
tjaalton | right, me too with FF | 14:52 |
Plagman | people don't usually complain about it with default IPP settings | 14:52 |
Plagman | but with InitialPixmapPlacement=2, the graphics memory isn't initialized so they get whatever was there at the time | 14:52 |
Plagman | whereas the contents of the backing pixmap should really be initialized to whatever is behind it | 14:53 |
Plagman | using that block that's removed by the patch I'm talking about | 14:53 |
tjaalton | it was added to the fedora xserver version 1.0.99.2-1 | 14:54 |
tjaalton | by krh | 14:54 |
Plagman | oh, okay | 14:55 |
Plagman | I'll just email him then | 14:55 |
superm1 | tjaalton, but it didn't actually land upstream? | 14:56 |
tjaalton | superm1: nope | 14:56 |
Plagman | of course not :/ | 14:56 |
Plagman | it removes functionality | 14:56 |
Plagman | as I see it, it's mostly an ugly hack to make Xgl faster at the cost of correctness | 14:56 |
tjaalton | actually, I did drop it from 2:1.2.0-3ubuntu1, but re-enabled because people complained about performance regressions | 14:56 |
superm1 | with what drivers did they see regressions? | 14:57 |
tjaalton | heh, can't remember.. it was 1,5 years ago | 14:57 |
tjaalton | and the changelog entry doesn't help either | 14:58 |
tjaalton | munckfish: hey, the daily ps3 image doesn't work :) | 14:59 |
superm1 | perhaps it will be worthwhile to test via an SRU for regressions against a handful of drivers then after Gold | 14:59 |
munckfish | tjaalton: the desktop one doesn't nope - I just tested it this weekend | 15:00 |
munckfish | but the alternate one does | 15:00 |
superm1 | and if perfromance regressions are still present just drop the SRU. | 15:00 |
tjaalton | munckfish: damn, another 2h download it is then :) | 15:00 |
Plagman | I'll contact Kristian about the original intent of the patch and its meaningfulness as of today | 15:00 |
munckfish | usplash is knackered for some reason and the desktop installer didn't have it disabled | 15:00 |
munckfish | last time it was built | 15:00 |
tjaalton | Plagman: that would help | 15:00 |
Plagman | If we agree that it doesn't make sense to keep it, who do I copy here to make what you said happen? | 15:00 |
tjaalton | munckfish: but shouldn't "install" run the text-mode installer? | 15:01 |
munckfish | tjaalton: other than that once it's up and running it's really quite nice | 15:01 |
munckfish | tjaalton: no it runs ubiquity I think | 15:01 |
tjaalton | Plagman: me and bryce for instance | 15:01 |
tjaalton | munckfish: bah | 15:01 |
Plagman | sounds good | 15:02 |
Plagman | thanks for the quick feedback | 15:02 |
tjaalton | np | 15:02 |
Plagman | damn, krh is off on vacation | 15:19 |
tjaalton | you could try ajax then | 15:23 |
Plagman | I guess I could | 15:24 |
Plagman | I'm not really familiar with the Ubuntu release/update management process; even if this patch was proved to be harmful, would there be a realistic hope of the X server in final 8.10 not having it? Or maybe in a subsequent update? I realize it's pretty late for that right now. | 15:28 |
tjaalton | no chance to get in the final release | 15:29 |
tjaalton | but an update is possible if dropping it doesn't break anything | 15:29 |
Plagman | I see. | 15:31 |
tjaalton | afk for 4h -> | 15:31 |
bryce | heya | 16:03 |
Plagman | well, ajax doesn't want the patch gone | 16:12 |
Plagman | apparently most drivers suffer from this | 16:13 |
Plagman | it's not a performance problem on our side, though; we'll discuss options internally | 16:14 |
Plagman | thanks for the help | 16:14 |
=== james_w` is now known as james_w | ||
jcristau | tjaalton: welcome :) | 18:41 |
wgrant | Hrmm. X doesn't like having key repeat settings set with HAL managing devices. | 21:56 |
tjaalton | jcristau: thanks :) | 22:27 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!