=== LockeAnarchist- is now known as LockeAnarchist | ||
=== LockeAnarchist- is now known as LockeAnarchist | ||
duflu | alf__: Is glmark2 designed to finish in a fixed time or vary according to GPU power? | 07:33 |
---|---|---|
=== chihchun_afk is now known as chihchun | ||
alf__ | duflu: fixed time (test parameter 'duration' default is 10 seconds) | 09:07 |
duflu | alf__: Right, so a slow machine won't take it over an hour then | 09:07 |
alf__ | duflu: right | 09:08 |
duflu | Remind me not to try and fix bugs so much. Fixing the fixes takes so much time that the feature development halts :) | 09:32 |
alan_g | Just don't put the bugs in to start with | 09:33 |
duflu | alan_g: None are mine (?). I'm trying not to victimize their creators | 09:34 |
duflu | Mostly not mine | 09:34 |
alan_g | Indeed, we all review the MPs | 09:34 |
duflu | Oh and then some devious test reboots my phone | 09:35 |
duflu | alan_g: I remember you suggested making people responsible, back at the sprint. I had thought the same before. The real issue is the person desiring a fix the most is not usually the person who broke it | 09:36 |
alan_g | I think people should be aware of how they broke something so that they can learn. But one can't be made responsible one has to be it. | 09:41 |
duflu | OTOH, if someone is a repeat offender, we would be wise to throttle them with bugfix tasks instead of feature development | 09:44 |
duflu | But that would be mean too | 09:44 |
duflu | More phone testing... | 09:45 |
duflu | alan_g: OK I can reproduce the GLMark2Test hang. It's a visible freeze. Will need to continue tomorrow | 09:56 |
duflu | Visible on mako | 09:57 |
alan_g | duflu: that's great - I couldn't reproduce (admittedly I didn't spend very long on it) | 09:57 |
duflu | alan_g: I suspect we're not matching up number of buffer replies with number of requests. Both server and client are starved and idling | 09:58 |
alan_g | Then you've got processes to poke around in? Even better. | 09:59 |
alan_g | I'll be glad to see that one gone | 10:00 |
duflu | alan_g: Yeah. Annoyingly seems related to what I'm fixing, but more needs fixing obviously | 10:00 |
alan_g | duflu: hopefully that means there will be a unified fix that will be obviously correct. (Once you've thought of it.) | 10:02 |
duflu | alan_g: Or many | 10:02 |
* alan_g hopes not | 10:02 | |
duflu | Oh it's going to be a bug fix week | 10:03 |
duflu | alan_g: Can you track down any more MPs that made it worse? | 10:05 |
duflu | Although I've got two that do it reliably | 10:06 |
alan_g | I've not noticed any others. But I'll keep track. | 10:07 |
duflu | Hmm, framedropping policy has moved into GlibMainLoop recently | 10:08 |
=== dandrader is now known as dandrader|afk | ||
=== dandrader|afk is now known as dandrader | ||
AlbertA | ogra_: the mir release is stuck in proposed: http://people.canonical.com/~ubuntu-archive/proposed-migration/update_output.txt | 14:22 |
AlbertA | http://pastebin.ubuntu.com/9250453/ | 14:22 |
AlbertA | ogra_: it looks like ubuntu-touch package has a dep on libmirplatform3driver-android | 14:23 |
AlbertA | ogra_: and looking at the README it seems it is based on the seeds: http://bazaar.launchpad.net/~ubuntu-branches/ubuntu/vivid/ubuntu-touch-meta/vivid/view/head:/README | 14:23 |
AlbertA | ogra_: any ideas on how to move forward? | 14:24 |
ogra_ | by dropping it ;) | 14:25 |
ogra_ | its a bit awkward that we have to update the seeds for such a change ... that means new -meta packages for every ABI bump you guys do | 14:26 |
ogra_ | would be nicer to have a higher level libmir meta package that the seed could depend on | 14:26 |
AlbertA | ogra_: right that's why this new mir release introduces mir-graphics-drivers-android | 14:27 |
ogra_ | (similar to linux vs linux-image vs linux-image-$(uname -r) | 14:27 |
ogra_ | ah, right, so lets land this then :) | 14:27 |
AlbertA | so do we manually modify the dep lists for ubuntu-touch and land it together with the mir release? | 14:29 |
AlbertA | ogra_:^ | 14:30 |
ogra_ | no | 14:30 |
ogra_ | ubuntu-touch-meta gets generated from the seeds ;) | 14:30 |
AlbertA | ogra_: ok | 14:31 |
ogra_ | i just merged your seed change ... takes a bit to generate the package | 14:31 |
ogra_ | i'll upload asap | 14:31 |
AlbertA | ogra_: ok thanks! | 14:31 |
alan_g | alf__: I'm looking at MirScreencastTest and thinking that the tests that depend on the mock_screencast are really integration tests. (And that the remaining tests don't really prove the feature.) Do you concur? | 15:24 |
alf__ | alan_g: looking | 15:24 |
alf__ | alan_g: I mostly agree. This (like other tests) would like to be true end-to-end tests, but unfortunately that's not possible with our current CI infrastructure. | 15:31 |
alan_g | yes | 15:33 |
alf__ | alan_g: Also what is an acceptance vs integration tests in our current architectur is somewhat fuzzy. The "hexagonal architecture" definition of acceptance vs integration is the clearest I have encountered. | 15:34 |
* alan_g goes to google | 15:34 | |
alf__ | (hexagonal == ports and adapters) | 15:36 |
alan_g | From a quick skim this looks like what I intended to describe at the initiation sprint | 15:37 |
alf__ | alan_g: Makes sense. It's also what the example in the GOOS book finally organically ends up with. Here are some nice diagrams from one of the authors: http://www.natpryce.com/articles/000772.html | 15:42 |
alan_g | yep | 15:43 |
alf__ | alan_g: I like that acceptance tests (tests again the domain model) are different from end-to-end (system) tests. | 15:44 |
alan_g | Sure. And it does fit with our classification. (What system tests we have are in benchmarks or examples and are not very good at reporting pass/fail) | 15:47 |
alan_g | Anyway, for the current instance, you're happy if I shuffle most of the tests over to integration tests? | 15:48 |
alf__ | alan_g: Sure. If we can easily change some of them to remain in the acceptance tests (and it makes sense to do so), then I would prefer that of course. | 15:50 |
alan_g | I'm taking reliance on mock_screencast (as opposed to having it around by accident) as the discriminator. | 15:52 |
alf__ | alan_g: Moving them will create some acceptance test gaps (e.g. for the mir_screencast api), but let's fix them properly. | 15:53 |
alan_g | It just makes the gap more visible | 15:54 |
=== chihchun is now known as chihchun_afk | ||
alf__ | alan_g: The MirScreencast test in particular could be probably made to work as acceptance test, if we use our MockGL/EGL infrastructure, which makes sense since EGL/GL is our integration point. We don't have an interface for GL/GLES but we can just use the link-time seam as usual. | 16:01 |
alf__ | alan_g: ... or introduce an interface for it | 16:02 |
alan_g | Yes, but I don't want to take that on right now | 16:03 |
alf__ | alan_g: Understood. | 16:05 |
AlbertA | greyback_: mir has released, so no lock on qtmir anymore :) | 17:09 |
greyback_ | AlbertA: yay, thanks | 17:09 |
* DalekSec waves to olli_. | 19:49 | |
olli_ | camako, DalekSec is representing xubuntu and was helping to gather some xmir compatibility test results back in the day | 19:52 |
olli_ | camako, he had a question if http://unit193.net/wiki/doku.php?id=wiki:mir is still of help | 19:52 |
camako | olli_, first time I'm seeing it... It could be useful if it were up to date. | 21:06 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!