/srv/irclogs.ubuntu.com/2013/12/16/#ubuntu-mir.txt

=== duflu_ is now known as duflu
dufluEeek, brown-out01:52
RAOFWin!01:57
RAOFThe mediumtests-trusty-touch and mediumtests-runner-mako CI jobs are just failing at the moment, aren't they? That's not an issue with my code?02:10
dufluRAOF: They are indeed just failing :/02:15
dufluin 80-90% of cases...02:15
dufluSo you can retry :)02:16
=== chihchun_afk is now known as chihchun
=== chihchun is now known as chihchun_afk
dufluRAOF: Could you kindly give the changelog and version proposals an honorary approval?03:27
=== chihchun_afk is now known as chihchun
RAOFThere you go.03:29
dufluRAOF: Ta03:47
RAOFI know *why* we have debian/ in the repository, I just continue to think it's a bad idea ;)03:48
dufluRAOF: If we continue with the two separate branches then I agree. But while the goal is to ideally have one, then I disagree03:48
dufluHmm, we *could* hide debian/ in dev and just quietly maintain it in lp:mir :)03:49
RAOFYes, we could.03:50
dufluAlthough that makes code reviews of packaging changes impossible03:50
dufluUmm, less convenient03:50
RAOFNot really? We just review on lp:mir03:50
RAOFI don't think it's a bad idea to have packaging changes reviewed in a different place to code changes; our developer skill sets are almost entirely disjoint, too.03:51
dufluSurely debian/ as a subdirectory is designed that way so you can keep it with your code without stepping on any toes03:51
RAOFHistorically it has not been; we're being an unhelpful upstream by including it.03:52
RAOF(This is mostly fixed with deb src 3.0, though)03:52
dufluRAOF: I hate to say it, but code reviews are even more scarce on lp:mir than dev.03:52
RAOFAre there any reviews to *do* on lp:mir?03:53
dufluI mean, if you rely on the subset of developers with the skills to review packaging changes, then you don't have enough reviewers03:53
RAOFAh, but we can get potentially get didrocks and other distro people to review those changes.03:54
dufluBecause he loves us so much already03:54
dufluAnd rightly so03:54
dufluI mean that both ways03:54
RAOFNo one from distro is going to regularly check merges that are almost all code changes unrelated to distro work; if we separate out those quite-infrequent changes it's more likely we can shanghai people to do them.03:55
=== sfeole` is now known as sfeole
RAOFduflu: What's the typo?06:58
dufluRAOF: construtor ?06:58
RAOFAHA!06:59
dufluOr is it copy-construe-er?06:59
dufluIf all else fails, copy and paste into an edit field with spell-checking07:00
RAOFHm. I do appear to have dropped into my native braces style a bit there, don't I?07:02
RAOFAlso, I'm pretty sure our style guide suggests 120 characters ;)07:04
dufluRAOF: On the theme of readability, not style compliance :)07:06
dufluThough I've already spent enough of my life trying to explain to people why a limit of 80 cols is important. I won't waste my breathe (much) more07:07
duflubreath too07:08
* RAOF wonders if we should set up clang-format as a precommit hook.07:13
dufluRAOF: I suspect we have too many exceptions we'd like to keep :)07:22
dufluIncluding 3rd party, which we wish to not touch where possible07:23
tvossRAOF, +1 on clang-format07:44
anpokis there a rule for curly braces placement .. egyptian vs on the same column?08:38
RAOFIs egyptian08:39
RAOFif (foo)08:39
RAOF{08:39
RAOF   hello();08:39
RAOF}08:39
RAOF?08:39
RAOFIf so, yes, we use Egyptian :)08:39
anpoknope08:39
anpokif (foo) {08:39
anpok   hello();08:39
anpok}08:39
anpokhttp://www.codinghorror.com/blog/2012/07/new-programming-jargon.html08:40
alf_anpok: http://unity.ubuntu.com/mir/cppguide/index.html08:48
tvossolli, ping08:51
anpokalf_: yes, but it only states it for function definitions.. and some examples use { on the same line..08:51
alf_anpok: they are wrong :)08:52
anpokobviously08:52
dufluAwesome. Now I can code like an Egyptian and party like it's 1986 simultaneously08:54
RAOFalf_: I'd like a second opinion on GLContext { void make_current() const; }; duflu has suggested that having it be const is an artefact of using non-C++ types, and I can see that, but I vacillated when writing the commit message for unconstifying it.08:57
RAOFalf_: cf: https://code.launchpad.net/~raof/mir/check-for-surfaceless/+merge/19851008:57
dufluRAOF: I'm happy to keep const. Just expect it to be non-const if the implementation changes08:58
RAOFduflu: Ok then. I think conceptually const makes more sense, and it's an implementation detail if it isn't.09:00
dufluYeah const is always a maintenance risk in that way09:00
dufluThere's a reasonable argument for not having const in the language :)09:01
dufluAlthough I think having const with clearly defined policy is better09:01
anpokit is always simpler to explain the usage of const_cast / mutable if it is to fulfill a const in the problem domain, than because of a possible trivial implementation09:14
anpokduflu: any suggestions on the pixel format encoding? should I only take RGB based formats into account now..09:16
anpokor look for something that also adresses yuv ..09:16
dufluanpok: I suggest sticking with what you have (don't change MirPixelFormat for now). It's nice and forward compatible to just define query functions like you have09:17
dufluFitting all the info we need into 32-bits is quite difficult actually09:17
anpokSDL2 does something similar .. but they dont encode shift values directly09:18
dufluanpok: The panacea is an accurate set of instructions for interpreting/encoding any pixel :)09:20
dufluWithout switch/case09:20
RAOFWe should probably look at gallium's format descriptor infrastructure.09:21
RAOFanpok: That's what your merge proposal triggered for me.09:22
anpokthe one in gbm.h09:22
anpok?09:22
dufluI think fourcc is kind of OK, but fails to resolve the bug really. You have to switch on all known fourcc's. It's ugly09:23
RAOFanpok: No... let me hunt it down.09:23
dufluHmm SDL_pixels.h is interesting09:23
RAOFhttps://github.com/RAOF/mesa/tree/egl-platform-mir/src/gallium/auxiliary/util - u_format.{c,h,csv}09:26
RAOFA bunch of that is driver-implementation-specific, but the general gist could be useful.09:28
anpokhm there are also compressed formats09:32
anpokhm those would not be needed ...09:32
RAOFRight. There are a huge number of formats there, most of which we wouldn't need.09:34
RAOFAlso, we (probably?) don't need all that description; block width and height, channel swizzle, layouts that aren't ‘plain’09:35
anpokthat reminds me of https://github.com/laanwj/etna_viv/blob/master/doc/hardware.md09:36
anpokif I am not mistaken the device can render to super tiled frame buffers.. and the optional composition gpu can do scanouts on those09:37
anpokthat might be relevant again with hwc on android09:48
anpokdifferent topic in some code parts DisplayConfigurationPolicy is named "initial" conf.. why only initial?09:49
anpoki initially assumed that.. if changes to the available displays happen the policy would querried again09:50
RAOFanpok: No; that's only used for initial configuration, and something else is assumed to handle configuration on hotplug events.09:53
* RAOF EODs10:00
* duflu too. Overdue for cooking a healthy meal10:26
mlankhorstRAOF: grrr! fix your trailing whitespace11:45
=== RoddieKieley1 is now known as RoddieKieley
mlankhorstRAOF: anway xorg-server for saucy-proposed, pixman and libdrm for all supported releases in the release queue:)12:31
=== chihchun is now known as chihchun_afk
=== chihchun_afk is now known as chihchun
anpoki am a bit concerned about the conversion operator usage13:19
anpokin EGLDisplayHandle and similar... not widely used. it seems odd to have conversion operators for typedefs of void*13:19
=== dandrader is now known as dandrader|afk
=== dandrader|afk is now known as dandrader
kgunnbregma: did RAOF get you the mesa you needed for nested mir on desktop ?15:16
kgunnnever saw a follow up15:16
=== dandrader is now known as dandrader|lunch
anpokkgunn: iirc nested mir on desktop also needs a small change in mesa platform which is currently in review15:22
kgunnanpok: exactly, this is what i was asking about15:23
kgunnanpok: i looked at raof's github but didn't see any reference15:23
kgunnbut maybe i don't know where to look exactly ?15:23
kgunncan you share if you do ?15:24
anpoksorry my wording was ambigiuous ..15:25
anpokmir/mesa needed another change15:25
anpokthere is some utility class that needs surfaceless context, but creates a pbuffer surface which is not supported by mesa.15:25
anpokhttps://code.launchpad.net/~raof/mir/check-for-surfaceless/+merge/198510 thats the change15:26
=== penghuan_ is now known as penghuan
xnoxDoes mir at all can export / has api to query detected screen resolution and the DPI setting?15:40
anpokyou can querry that through the mir client api - take a look at demo_client_display_config.c in the examples folder15:44
=== Saviq_ is now known as Saviq
=== chihchun is now known as chihchun_afk
kdubyay https://lists.ubuntu.com/archives/ubuntu-devel/2013-December/037904.html16:09
xnox_=)16:10
kdubi was just looking up who to +1... i'll see  if  i can drop my amateur cmake cross compile file from mir16:11
=== Saviq_ is now known as Saviq
=== dandrader_ is now known as dandrader
kgunnkdub: so did you say if you put in some logic to "ignore" the second "on"...display turns on but then hangs ?16:26
kdubkgunn, right16:27
kdublike, the problem is16:28
kgunnkdub: damn...that's just weird...so not only does powerd send it...it meant to send it :)16:28
kdubright16:28
kdubi'm still poking around powerd trying to understand16:28
kdubricmm, ping16:29
greybackkdub: anything I could do to help?16:35
kdubgreyback, don't know of anything at the moment16:36
greybackkdub: ok16:37
kdubunity-mir's role seems to be pretty straightforward16:37
=== chihchun_afk is now known as chihchun
=== FunnyLookinHat__ is now known as FunnyLookinHat
=== chihchun is now known as chihchun_afk
* mterry pokes ricmm about libhybris17:33
kdubmterry, what about hybris?17:34
mterrykdub, that libhybris fix that ricmm identified during the sprint, to allow two different users (root and phablet) to access hw on maguro17:35
mterryIt's the last fix needed for nested mode to land17:36
kdubah, okay17:37
kdubmterry, is screen on/off with nested working okay? i have some concerns about that17:37
kdubeh, actually, i'd expect it to work17:38
kdubits just funny that we make the power request to a nested client who asks the host server17:39
kdubtoo many hops17:39
mterrykdub, no, UCS is actually the one that will host the dbus interface17:40
mterrykdub, the unity-mir code to do it is still there, but will gracefully fail once UCS already owns it17:40
mterrykdub, and then we can remove that codee17:40
kdubmterry, great, sounds good17:40
=== dandrader is now known as dandrader|afk
=== dandrader|afk is now known as dandrader
=== fginther` is now known as fginther
=== csslayer_ is now known as csslayer
kgunnkdub: so i was just trying your instructions...i'm on N4 with todays devel-proposed image...after i install mir-test-tools20:57
kgunnthe only mir test i see is mir_stress20:58
kgunnany ideas?20:58
kgunnhttps://docs.google.com/a/canonical.com/document/d/1oxXDxehRQ35rYRwkuBCy1ZNmB_IBpg5UhE3jvNHEk_0/edit20:58
* kdub looks21:00
kdubthey were added in rev 125621:07
kduband it looks like 1248 was the last merge to lp:mir21:07
kdubkgunn^^21:08
kgunnkdub: heh21:09
kgunnfell victim21:09
=== slangase` is now known as slangasek

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