On Mon, 2012-09-24 at 17:59 -0400, Adam Jackson wrote: > On Mon, 2012-09-24 at 13:30 -0700, John Reiser wrote: > > > Do you have any old or new graphic hardware, working or not? Join this test week and help us to hunt down driver bugs before Fedora 18 Beta release! > > > > There is doubt about whether this particular test day is worth the time. > > You're free to doubt whatever you like, but the experiences you appear > to be citing as evidence are not especially relevant. > > > At a minimum, don't bother with any Radeon card that is less than a Radeon 9600. > > Fedora pays no attention to bugs on such hardware, not even saying "Sorry, this > > hardware is too old; we will increase the minimum hardware requirement" or > > "This is hardware bug, and a software workaround is not feasible." For instance: > > radeon_rendercheck 152752 failures; pci:1002:5157 Radeon RV200 QW [Radeon 7500] > > https://bugzilla.redhat.com/show_bug.cgi?id=638758 > > This card still is usable in XFCE. > > a) You're calling out a test you ran once nearly two years ago. Pretty > sure we've fixed at least one RV200 bug since then. > > b) rendercheck is just a test battery. Passing all of its tests is not > necessarily a prerequisite for correct rendering of any desktop; if you > don't hit cases that fail, you'd never notice they were broken. > > c) I'm fairly sure zero of the tests you've shown to fail there _do_ > matter. They all happen when you do a Render operation directly to a > window, and nobody does that. Drawing straight to a window is > unbelievably flickery, that's why everybody buffers things up to an > offscreen pixmap and then blits across to the window (using core > CopyArea, not Render). We added the rendercheck test to the list just to provide the data for you folks (X devs) to look through if you thought it might be useful. I'll see if I can do something (more) on the test day pages to indicate that failures in rendercheck aren't always bugs per se and not to worry too much if some of the tests fail, and that it's not usually necessary to open a bug just for rendercheck failing. > > Gnome3 is not putting any effort into fallback mode. So my Radeon 9250 > > purchased new in 2006 as a mid-life upgrade for a box with 1.6GHz Pentium 4 > > (well above the Fedora minimum CPU) also runs only XFCE. Gnome3 forces > > fallback mode, where some Desktop features are *dropped* instead of emulated. > > Why do you feel this is relevant? X bugs are X bugs. Right, this is mostly out of scope. The actual bug here is well-known, btw - there's a blacklist for adapters that do 3D but are known not to be capable of rendering Shell properly. Currently, if your card is on that blacklist, you wind up with fallback mode, but what we really want to happen is for it to fall back to software rendering, like it does on systems where there's no 3D capability at all. That's the bug here. As ajax says it's nothing to do with X or the X test days, it's a bug in the GNOME fallback logic. Ajax replied to your specific points, but there's an underlying general point, and here it is: not all bugs reported as part of Test Days will be fixed. Even some 'valid' ones will probably wind up withering. It's unfortunate but for zillions of reasons - each individual case has an explanation, as the specific examples cited here show - it happens. We don't expect 100% response rate for Test Day bugs, it'd be unrealistic. I'd only be worried if the rate was extremely low or dropping fast. Since X Test Week is a large and regularly repeated event it's actually possible to look at those numbers, and indeed we do: I post a statistical breakdown of various things, including a look at the outcome of the filed bugs, after each Test Day. Look through the archives for these posts: "Very belated 2011-09 Graphics Test Week recap" - Wed, 30 Nov 2011 11:41:20 -0800 "2011-02 Graphics Test Week recap" - Thu, 03 Mar 2011 03:22:49 +0000 "2010-09 Graphics Test Week recap" - Tue, 05 Oct 2010 14:50:19 -0700 "2010-04-13 to 2010-04-15 Graphics Test Week recap" - Tue, 20 Apr 2010 15:11:55 -0700 In the 2011-09 one you can see the numbers for what's happened to all the bugs reported in all the test days from Fedora 11 cycle through Fedora 15 cycle. They go up and down, but there isn't any clear worrying trend there, and a decent number of bugs was fixed in almost every cycle. As long as that's happening, the event has value. From the F15 cycle, in total across all three test days, 123 bugs were reported, 34 were closed as 'fixed'...fixing 34 bugs isn't a bad outcome from such an event, I don't think. (if you're wondering where the F16 numbers are - they would have been in the recap for the f17 X test week, but there was no f17 X test week because I never quite got around to running it. I'll do the f16 numbers in the f18 test week recap.) -- Adam Williamson Fedora QA Community Monkey IRC: adamw | Twitter: AdamW_Fedora | identi.ca: adamwfedora http://www.happyassassin.net -- test mailing list test@xxxxxxxxxxxxxxxxxxxxxxx To unsubscribe: https://admin.fedoraproject.org/mailman/listinfo/test
