/srv/irclogs.ubuntu.com/2014/08/26/#ubuntu-mir.txt

RAOFUrgh. Have I broken CI again?04:14
RAOFThe test-build worked, damnit!04:14
RAOFON CI!04:14
=== LockeAnarchist- is now known as LockeAnarchist
RAOFOOOOooooooooh!05:25
RAOFI have become the destroyer of CI!05:37
* RAOF should not work in branches where the default push is to the default CI runner.05:45
RAOFAlright!06:14
RAOFCI should really, really work now you guys.06:14
RAOFFOR REALZ!06:14
* RAOF goes to pick up his car.06:14
RAOFOk. Now CI's just messing with me.07:27
=== tvoss is now known as tvoss|test
=== tvoss|test is now known as tvoss
=== yofel_ is now known as yofel
anpokyay renaming functions with the c-preprocessor is great09:24
alan_gIt is a powerful technique - and should be used responsibly (i.e. rarely)09:27
anpokhere it was used to link to archives together that use a lot of copied code09:28
anpok*two09:28
anpokhence one is built with a header full of #definef func old_func ..09:28
=== davmor2_ is now known as davmor2
alf_alan_g: Hi! Could you please elaborate on "Indeed wouldn't placing a cache implementation directly in the configuration object be sufficient?". Do you mean that the configuration object should have something like: std::shared_ptr<SharedLibrary> the_platform_library() ?11:22
alan_galf_: I was thinking of a data member SharedLibraryCache with a load_library member.11:24
alan_gBut I guess we only ever load the one, so a std::shared_ptr<SharedLibrary> platform_library member would do.11:26
alf_alan_g: The SharedLibraryCache needs to be shared/survive across configurations and MirConnections, so I am not sure how a SharedLibraryCache as data member would work.11:28
alan_galf_: OK, that answers my "needs info".11:28
=== alan_g is now known as alan_g|lunch
=== Saviq_ is now known as Saviq
eanyxhi12:19
eanyxDoes someone has metrics for 2D/3D throuput of mir?12:22
eanyxthroughput12:23
kdubthere's a glmark built for mir12:24
eanyxDoes the performance are good compared to legacy X?12:24
kdubsure12:26
eanyxIn what proportion?12:26
kdubtoo vague questions12:34
eanyxI'm wanting to write a 2D game, and wanting to know what is the blitter speed ?12:36
kdubwith the in-game stuff, thats up to you... for mir at least12:38
kdubwe bypass fullscreen on mesa12:38
kduband use the overlay path on android12:38
kdubcould everyone take a look at  https://code.launchpad.net/~kdub/mir/frontend-tracking/+merge/23212312:41
kduband indicate which of the two exchange_buffer messages are preferable?12:41
=== alan_g|lunch is now known as alan_g
eanyxis there some vsync for smooth animation?13:08
alan_gkdub: I've looked and don't have a preference13:09
kdubthanks alan_g13:10
eanyxhave to leave you, have a nive day/evening.13:42
alf_kdub: fine with both, although I wonder if allowing the client to indicate surf_id could be used maliciously. Of course we already trust the client library to give us correct info in general, so perhaps this is a non-issue.14:01
kdubalf_, yeah, agree14:02
kdubat any rate, i'll be on later than normal tonight, so I'll just chat it over with RAOF then14:03
dandradera gmock question: is using MOCK_METHODX(....) any good when I want to check a complex parameter of a method?15:09
dandraderI have a parameter like this "const QList<struct QWindowSystemInterface::TouchPoint> &points"15:09
dandraderso I wanna iterate the list and check the values of the TouchPoint structure, etc15:09
dandraderI wonder if using the gmock magic stuff (like declaring the method with MOCK_METHOD() and using things like EXPECT_CALL()... in the test) will be of any use in this case15:11
dandraderor if I'm better of doing the mock implementation manually15:11
* dandrader was told mir developers were the local gmock experts15:11
tvossdandrader, you might want to look into MATCHER_P15:11
tvossdandrader, so yes, gmock is quite useful in your scenario, and you can just describe your matching logic in MATCHER_P15:12
* dandrader googles for MATCHER_P15:13
kdubthere are also some container matchers15:13
kdubhttps://code.google.com/p/googlemock/wiki/CheatSheet#Container_Matchers15:13
kdubthat might be helpful15:13
dandradertvoss, kdub, ok, thanks for the tips. will look into those15:14
=== chihchun is now known as chihchun_afk
josharensongreyback__ when I build qtmir, it keeps failing on a mir header file, even though the file exists...16:40
greyback__josharenson: can I see the build output please? How are you building?16:40
greyback__josharenson: if you're building manually on the phone, you need this: qmake "QMAKE_CXX=arm-linux-gnueabihf-g++-4.9" "QMAKE_LINK=arm-linux-gnueabihf-g++-4.9" "QMAKE_LINK_SHLIB=arm-linux-gnueabihf-g++-4.9"16:41
josharensonas per the readme.. mk-build-deps, qmake, make (getting a pastebin of the output)16:41
greyback__ah, I better update the readme - the 4.9 thing is annoying16:41
josharensoni was going to update it when I got it working :-p16:41
josharensonhttp://pastebin.ubuntu.com/8151394/16:42
greyback__josharenson: so libmirserver-dev is installed? In that case, I'd clean all & try again (qmake not good at seeing dependencies changing): bzr clean-tree && bzr clean-tree --ignore16:44
greyback__then try the qmake command again16:44
josharensongreyback__ I had libmirserver-dev installed, but cleaning the tree seems to have fixed whatever was wrong...16:55
greyback__josharenson: glad to hear it. Usually that happens to me if I run qmake before having all the dependencies installed. Running qmake again after often does not pick up the new header/lib directories16:56
josharensongreyback__ good to know, thanks16:56
greyback__full clean and retry is my usual fix16:56
greyback__np16:56
=== alan_g is now known as alan_g|EOD
=== dandrader is now known as dandrader|afk
=== dandrader|afk is now known as dandrader
=== dandrader is now known as dandrader|afk
=== DalekSec_ is now known as DalekSec
rsalvetiAlbertA2: hey, while I agree that it's good to hold a lock during a voicecall, this same issue will happen next time we have an app using the proximity sensor22:54
rsalvetichecking https://code.launchpad.net/~albaguirre/powerd/avoid-suspend-during-call/+merge/23229022:54
Saviqcamako, is there an actual reason why we need the bump of qtmir dep to > 0.7.0? if not, we should just go for a no-change rebuild?23:01
Saviqcamako, although the -gles dep should definitely be bumped to actually be in sync with the non-gles one, but in general unless there's API breaks we don't bump the -dev dependency, it will still build with the latest available23:02
camakoSaviq we broke the ABI.. So I just updated the dependency23:02
Saviqcamako, right, so we only need a no-change rebuild against the new -dev, debian/control deps should generally reflect the truth (as in what's the minimal mir version with which qtmir will build)23:03
camakoSaviq, do you mean, for instance, if there is no client API break then leave the client dependency alone?23:03
Saviqcamako, basically if there's no code changes required23:03
Saviqcamako, there's no need to touch the deps23:03
camakoSaviq, but we definitely broke the ABI on the server side... client side perhaps I should leave it alone23:04
Saviqcamako, sure, but that will sort itself out through a no-change rebuild23:05
Saviqcamako, the build system will select the highest -dev version during the build anyway, and then encode the real versioned dependency in the runtime dependencies23:05
Saviqcamako, basically, qtmir can build with 0.6.0 still, can it not?23:06
camakoSaviq, Oh I see what you mean... I dunno if it can actually..23:07
Saviqcamako, if you only bumped the debian/control version and it still builds, sounds like it can :)23:07
camakoSaviq, you're right... If mir hasn't changed actual code qtmir (or others) depend on then no need to do what I did23:08
Saviqcamako, dependencies (especially versioned ones) in debian/control should be "the least required"23:08
* Saviq comments on the MP then23:09
camakoSaviq, I see... so how do we do a no change rebuild in the silo?23:09
Saviqcamako, you just need an empty (--unchanged) commit in the MP23:09
Saviqcamako, for the -gles counterpart you still need to sync the version number (and silo)23:10
camakoSaviq, yes Iit's just prelim at this moment23:11
Saviqcamako, yup23:11
camakoSaviq, thanks man... every time I talked to you I'm illuminated :-)23:12
Saviqcamako, always23:12
Saviqcamako, for fast development like we do now it isn't as important, but once Mir becomes part of the default distro, LTS and such, it pays off to not bump deps unnecessarily so that you're not forced to unnecessarily build a chain of dependencies23:14
Saviqwhen someone wants to build on an older distro, that is23:14
camakoSaviq, makes total sense...23:14
=== dandrader|afk is now known as dandrader
Saviqcamako, one thing to note (when not bumping -dev deps), when you build things in silo, you need to make sure of the order things build in the right order23:20
Saviqcamako, because if you just build mir and qtmir in parallel, qtmir will actually build against the distro version23:20
camakoSaviq, ok.. and putting mir MP in the front doesn't help with that right?23:20
Saviqcamako, no, they all get pushed to the PPA at the same time23:20
Saviqcamako, you just need to run the build job with just "mir" first23:21
Saviqor whatever comes first for real23:21
camakoSaviq, ok got it23:21
Saviqcamako, I think there's a circular dependency between papi and mir still, which means you might actually need to build papi, mir, then papi again or something similar23:22
RAOFkdub_: Yo!23:22
camakoSaviq, yeah I've encountered that before and I look out for it23:22
Saviqcamako, ideally this should be resolved if at all possible23:23
Saviqcamako, that would require splitting papi, though23:27
camakoSaviq, yeah it's a nuisance23:27
Saviqcamako, the circular dep might lead into a difficult situation if in a single landing you'll have papi depending on new mir api and mir on new papi, neither will build without the other23:27
Saviqoh and there's even one with mesa too bug #123903723:28
ubot5bug 1239037 in mir (Ubuntu) "circular build-dependency between mesa, mir" [Medium,Triaged] https://launchpad.net/bugs/123903723:28
SaviqI thought I filed one for papi, mir, will do so now23:29
camakoSaviq, yeah I guess we change only one first, build the other then we change the other and build the first.. if that makes sense)23:29
Saviqcamako, sure, that makes sense, but might mean you have to do the changes in stages unnecessarily23:30
camakoSaviq, right can't do in one silo... would have to split the landing23:30
camakowhich would be an even bigger PITA than it already is23:31
camakoSaviq, so it'll look like this right? https://code.launchpad.net/~mir-team/qtmir/devel-mir-next/+merge/23232923:35
Saviqcamako, yes, approved23:41
Saviqcamako, btw, looks like the papi/mir loop was resolved in the end, it's only the mesa one that remains23:42
Saviqcamako, i.e. mir does not build-depend on anything produced out of platform-api AFAICT23:42
camakoSaviq, I see23:42
Saviqcamako, btw, in theory it should be possible to pull lp:qtmir into lp:qtmir/devel-mir-next instead of merging (is what we do with the gles-sync branch)23:44
Saviqwell, not any more, because there were merges and such, but maybe we could make it happen again by overwriting devel-mir-next with trunk at the next landing23:45
Saviqthis way trunk and devel-mir-next actually maintain the same history23:45
camakoSaviq, I see what you mean23:45
camakoSaviq, good idea23:45
* Saviq tries to verify23:45
camakoSaviq, this actually occured to me but I was a chicken to try23:46
Saviqcamako, so yeah, as long as there's no commit *more* on devel than on trunk, you can pull23:47
* camako will do it next time23:48
camakoSaviq another nice thing with the empty commit option is now the MPs will be approved by Jenkins (before they always failed because Jenkins didn't know anything abt the new release before it happened)23:50
Saviqcamako, sure, assuming they will actually build ;)23:51
Saviqcamako, and well. sure, jenkins approved, but with old mir, so not really something you wanted23:51
camakoSaviq, yea I'm hungry so who knows what I did23:51
Saviqcamako, so yeah, even now, if you uncommit your last commit to devel-mir-next, you can just pull from trunk, so yeah, just after the next qtmir landing, pull into devel-mir-next and you'll have history nice & clean23:52
camakoSaviq, yeah but at least we won't see a failure23:52
camakoSaviq, yeah.. same thing with USC and papi23:52
Saviqcamako, yup23:52
Saviqcamako, so yeah, until devel-mir-next has actual source changes, jenkins should be happy - if it stops being happy is when you should actually bump the build dependencies appropriately23:53
camakoSaviq, yea.. it's sort of one extra checkpoint23:54
Saviqcamako, yup23:54
Saviqo/23:56
* Saviq sleepy time23:57
camakoHey Saviq, really thanks man.. good night23:57
Saviqcamako, right back at you23:57

Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!