RAOFI'll fire up the ati system again.00:00
bschaeferRAOF, thanks, kgunn just wanted another working xmir data point, so for 1 is failing and 2 are working00:01
bschaefers/for/far ... geez00:02
* robert_ancell -> lunch00:08
kdubmy 1tb harddrive is 80% full of compiled mir code... bzr -_-00:23
kgunnRAOF: can you please ping didrocks when he comes online & try to help resolve his ati issue...its our last blocker00:24
RAOFkgunn: Sure.00:24
RAOFkdub: cd ~/.bazaar/plugins; bzr branch lp:bzr-removable ; cd ~/Devel/Mir ; for I in $(bzr removable trunk ) ; do rm -r $I ; done ☺00:25
kdubRAOF, hmm, might give it a try!00:26
RAOFkdub: bzr-removable adds the "removable" command; in a repository, you give it a branch and it'll tell you all the branches in the repository which have been merged into that branch.00:27
kdubah, that would be useful00:27
RAOFIt also adds the "unremovable" command, which does the reverse, and tells you why it's not removable00:28
* RAOF might get around to adding a "-d" option to it, so that ‘bzr removable -d trunk’ would delete all the branches that have been merged into trunk00:29
* RAOF books travel to Boston00:30
RAOFbschaefer: Yup; My ati box still works with unity-system-compositor00:31
bschaeferRAOF, thanks00:32
RAOFNow; does this XMir populate xrandr data?00:38
bschaeferRAOF, where would one check for this data?00:42
bschaeferi see this in my syslog: Aug  7 16:30:51 bschaefer-GA-870A-UD3 colord: Device added: xrandr-XMIR-100:42
RAOFbschaefer: Oh, your XMir definitely doesn't. I'm just hooking up the new Mir multi-head API.00:45
bschaeferoo cool, though im not sure what that is :)00:45
RAOFxrandr will report the full set of modes your outputs can do, rather than just the single mode Mir set on startup :)00:48
bschaeferRAOF, oo nice, yeah, i've just been stealing mir_connection_get_display_info ... and using the width/height for what I think mir can display00:49
RAOFYeah. You'll need to switch to mir_connection_create_display_config() at some point ☺00:50
bschaefersweet, yeah I saw that in the egl examples yesterday, now that makes much more sense :)00:50
RAOFYeah. Make sure you check the most recent egl examples; I fixed a thinko in them yesterday :)00:51
TheDrumsSorry for the question, but are there any major changes since the last PPA build in this one?  Drivers, Mir, Xorg u-s-c?  And lastly, I'd guess 0.0.9 is still out several weeks?00:52
bschaefercool, yeah I need to go back and update some of my branches... I think the ABI might have changed as well00:52
RAOFbschaefer: The client ABI shouldn't have changed. If it did, please hit us with a stick.00:53
RAOFThe server ABI is... not yet ☺00:53
bschaeferRAOF, haha, will do00:53
bschaeferI think the last time was ... the swap buffers call00:53
bschaeferit use to be mir_connection_next_buffer or something?00:54
bschaeferbut that was a while ago...00:54
RAOFAh, yes. We did change that.00:55
RAOFThat was a while ago.00:55
RAOFWoot! We have xrandr info.00:55
bschaeferyes it was, I just haven't re-compiled my branch in sometime :)00:55
RAOFAlthough I should probably try it on something that isn't this laptop, as it only supports 1920x1080, which makes the mode list somewhat short!00:55
bschaeferthat could help haha00:57
robert_ancellduflu, RAOF etc, can you review https://code.launchpad.net/~robert-ancell/mir/vt-switch-keys/+merge/179067?01:10
robert_ancellwould like to knock out the VT issues today01:10
RAOFrobert_ancell: Looks like you've still got some debug printfs in there?01:13
robert_ancellRAOF, ah, missed one01:13
RAOFQuite a few, it seems?01:13
RAOF51, 87, 126...01:14
robert_ancellRAOF, ah, no I just didn't push the last commit01:14
robert_ancellthe "clean this up for merging" commit01:14
RAOFAlso, what is KEY_L?01:15
duflurobert_ancell, yep01:15
robert_ancellRAOF, also debug and also removed01:15
RAOFGood, good.01:15
robert_ancellRAOF, I needed to check I wasn't hitting the existing alt+ctrl+Fn keys01:15
RAOFYeah, fair suck of the saz.01:15
RAOFrobert_ancell: Hm. Why are you explicitly initialising a std::initialiser_list?01:32
robert_ancellRAOF, because std::make_shared seems to get confused01:33
RAOFAh, if you just do ...({vt_filter})?01:33
robert_ancellRAOF, because the it treats it as a std::initializer_list<std::shared_ptr<VTFilter>> not, std::initializer_list<std::shared_ptr<mi::EventFilter>>01:34
robert_ancellspent a lot of time trying to understand the error message01:35
RAOFexport CC=clang; export CXX=clang++ in ~/.bashrc makes that much easier :)01:38
=== jono is now known as Guest576
dufluHey, Ctrl+Alt+Backspace is a much cleaner shutdown than Ctrl+C was. I wonder how much we should still worry about our signal handling...01:59
RAOFrobert_ancell: Hm. That doesn't VT switch correctly for me - or, at least, it does vt switch but then immediately switches back.02:02
RAOFrobert_ancell: Does it require https://code.launchpad.net/~mlankhorst/mir/setsid/+merge/177800 ?02:02
robert_ancellRAOF, yeah, I saw something similar. It goes away with setsid02:02
duflurobert_ancell: I knew I would have to test with setsid. Does a prereq somewhere make sense?02:04
robert_ancellduflu, I don't think it's strictly required, though I would land it next02:04
duflurobert_ancell: I'm finding similar issues with three of my branches right now... Behaviourally they are quite dependent, but diff-ly they are separate :/02:05
RAOFAs long as we land both at approximately the same time it's ok.02:07
RAOFIdeally we'd land setsid and vt-switch-keys atomically, though.02:07
dufluRAOF: Atomicity can only be guaranteed by telling people not to pull from trunk for a while, I guess02:10
RAOFNo, it can also be guaranteed by merging the two branches and landing them as one?02:11
dufluRAOF: Other than that, which clearly we're trying to avoid02:11
dufluThat said, I'm already building/testing them together02:11
RAOFIt's not clear to me that we're trying to avoid that?02:11
RAOFAnyway, +1 on vt-switch-keys and +1 on setsid.02:12
duflurobert_ancell, hangout?02:56
robert_ancellduflu, syre02:57
robert_ancellthomi, weird thing - https://launchpadlibrarian.net/147042471/buildlog_ubuntu-saucy-i386.mir_0.0.8%2B13.10.20130807.3bzr943saucy0_FAILEDTOBUILD.txt.gz03:55
robert_ancellFails to build (I saw it locally too). But doesn't occur when the CI builds occur03:56
robert_ancellAnd the error seems quite straight forward - just missing #import <iostream>03:56
thomithat is odd03:57
thomiyou'd expect them both to fail03:57
robert_ancellyeah. When I build locally with ./native-compile.sh it works, but not when using debuild03:58
robert_ancellMaybe debuild sets some cflags and normally it's just  a warning, but for debuild it's an error03:59
thomirobert_ancell: perhaps locally one of the other header files #include's iostream, but the version in the package build env is older, and doesn't?03:59
robert_ancellhah, looks like I broke it in a previous merge. The CI just didn't pick it up04:00
robert_ancellactually it was alf, when he finished off my alt+ctrl+backspace branch04:00
thomirobert_ancell: should we add that flag to the builds?04:03
robert_ancellthomi, We probably should if you can work it out.04:03
thomirobert_ancell: or rather, can you add it to the standard compile flags?04:03
robert_ancellthomi, don't the builds run debuild anyway?04:03
thomirobert_ancell: yeah, I'm not sure what's going on04:03
thomihmmm.. perhaps the pkg builds override CXXFLAGS?04:03
robert_ancellthomi, can you review https://code.launchpad.net/~robert-ancell/mir/missing-import/+merge/17909204:04
* robert_ancell shrugs04:04
robert_ancellguess we wait until it happens a second time before spending too much time fixing it04:04
thomirobert_ancell: BTW, did you want a copy of my travel plans for Boston -> AKL so you have a travelling companion on the way home?>04:05
thomior are you so sick of me now that you'd like my flight details so you can deliberately plan a different route? :P04:05
robert_ancellheading out - be back in an hour04:06
robert_ancellthomi, sure, forward them04:06
thomirobert_ancell: OK, jobs are all kicked off again for the next test run04:10
tvoss_good morning :)04:28
thomihi tvoss_, how's life?04:29
tvoss_thomi, pretty good :) how is life on your side?04:29
thomitvoss_: still getting over the jetlag, but otherwise good04:29
thomithinking about investing in a pottery wheel... ;)04:29
tvoss_thomi, having a pottery wheel around is always a good idea04:31
thomigotta get some practise in, before the international competition04:31
thomiI gotta head out, will be back later for late calls.04:33
RAOFHm. There is no problem that can't be fixed with another layer of indirection!04:34
tvoss_RAOF, for sure :)04:34
Mirvhmm, I wonder if this https://launchpadlibrarian.net/147032204/buildlog_ubuntu-saucy-armhf.mir_0.0.8%2B13.10.20130808-0ubuntu1_FAILEDTOBUILD.txt.gz is something already taken care of?04:41
Mirvdemo_inprocess_surface_client.cpp:58:27: error: 'cerr' is not a member of 'std'04:41
RAOFMirv: robert_ancell was just talking about that04:44
Mirvah, I see04:45
Mirvand approved, nice, I'll rerun the stack when it has been merged04:46
Mirvok, rerunning05:00
Mirvwell, unity got there first, might take some time..05:02
RAOFthomi: Could you kindly turn off the autolanding for mesa into the staging ppa.05:21
RAOFdidrocks: Yo!05:33
didrocksRAOF: hey!05:33
RAOFdidrocks: I understand you've got a kernel backtrace for the ati system of death?05:34
didrocksRAOF: I don't have a kernel backtrace AFAIK. All the traces were pasted on IRC.05:36
tvoss_didrocks, jibel pasted the original kernel tr05:37
tvoss_didrocks, RAOF browser-history ftw: http://paste.ubuntu.com/5959277/05:38
RAOFHm. That's interesting.05:43
RAOFdidrocks, tvoss_: So, that backtrace has Xorg as the controlling process of the CPU that wedged; it's calling drm_mode_setcrtc, which means that it's not running under unity-system-compositor.05:44
tvoss_RAOF, hah05:45
didrocksRAOF: right, so we had one run with u-s-c05:45
didrockswhich starts, but compiz never finished its init05:45
didrocks(blocked on opengl plugin)05:45
didrocksthen, we restart without u-s-c/Mir, under raw Xorg05:45
didrocksand this is what happens ^05:45
tvoss_didrocks, this is a very wild guess, but could we update the machine's bios?05:46
didrockstvoss_: that doesn't change that the previous run failed05:46
didrocksthe machine blocking is a consequence05:46
robert_ancellcould we accidentally drop that machine out the window?05:46
RAOFdidrocks: How do you restart without u-s-c? Is a reboot involved, or a tweak of lightdm.conf.d + a lightdm restart?05:47
* didrocks wants to accidentally drop doko out the window05:47
didrocksRAOF: new install, other stack to tests05:47
RAOFdidrocks: So, a reboot?05:48
didrocksRAOF: without any reboot05:48
didrocksit's a lxc container05:48
RAOFAh, ok. Right.05:48
didrocksso the previous run (with u-s-c failing) was ended05:49
didrocksthen, another one for another stack with regular Xorg05:49
RAOFSo that would mean the blocking problem is likely to be a failure to clean up u-s-c properly on lightdm shutdown.05:49
RAOFAnd then there's the problem of u-s-c failing in the first place.05:49
didrocks(only on ati)05:50
tvoss_robert_ancell, +105:50
didrocksintel with the same packages works perfectly05:50
robert_ancellRAOF, could I ask you a favour? I need to do family things before coming back tonight - can you watch https://launchpad.net/~mir-team/+archive/staging/+build/4859513 and when that completes copy (without rebuild) mir 0.0.8+13.10.20130807.3bzr944saucy0  from the staging PPA to the system-compositor-testing PPA?05:50
RAOFrobert_ancell: Ok.05:50
robert_ancellThen follow up with u-s-c rev 40 when that autolands (with rebuild)05:50
robert_ancellI'm in fear that by the time I come back and new revision will have landed and wipe out the build :/05:51
RAOFdidrocks: I don't suppose it's possible to log into that box while it's running the test?05:57
didrocksRAOF: we can, just not now05:57
didrocksdoko broke Qt5…05:57
didrocksneed to fix that first05:57
* RAOF goes back to xrandr hotplug05:59
dufluRAOF: I suspect I had the Xdamage/Compiz argument backwards. Some Compiz plugins actually expect and require that they be able to spuriously render outside the damage region, and that it only looks *right* if you show the damage region (hide overdraw).06:08
dufluSo that was mostly fixed when we enabled page flipping. And would be more fixed by using regional logic again06:09
RAOFduflu: You mean they render *incorrectly* outside the damage region, and expect any rendering outside the damage region to be invisible?06:16
dufluRAOF: Yes06:17
RAOFThat's messed up.06:17
dufluRAOF: Most of the obvious occurrences were fixed last year, but not all06:17
dufluRAOF: unityshell too :P06:17
RAOFAlso, XMir will display any rendering they do.06:17
RAOFBecause XMir sees the damage that *compiz* does.06:18
RAOFOh. I guess unless compiz does partial frontbuffer updates.06:18
dufluRAOF: Yes, that's just a CCSM tickbox away06:18
RAOFIn which case XMir will just see the bits of the frontbuffer you updated.06:18
dufluRAOF: Just untick everything except sync_to_vblank in CCSM>OpenGL06:20
RAOFFrankly I'd prefer to just show the broken rendering and fix it.06:20
dufluRAOF: I don't think you will see any bugs outside of obscure plugins06:21
dholbachgood morning06:57
RAOFWow, Mir takes a while to build on arm.07:04
ogra_RAOF, locally ? or on the buildd07:05
RAOFOn the builld.07:05
ogra_(we got new buildds yesterday ... )07:05
RAOFI'm pretty sure I could have had this finished in a qemu schroot in the hour it's taken on the buildd, including the time taken to set up the armhf chroot :)07:05
ogra_is that a distro buildd or PPA ?07:06
ogra_would be worrying if its distro07:06
RAOFI understand we have shiny new caldexa nodes for our distro buildds.07:06
RAOFThat do things like build Mir in less than an hour :)07:06
ogra_on the distro buildera firefox build takes ~5h now ... was between 20 and 30 before07:07
mlankhorstRAOF: ping07:11
RAOFmlankhorst: Pong.07:11
RAOFGah. Did I fail to push the fixed patch for that07:11
mlankhorstit has come to this07:12
RAOFmlankhorst: I fixed that in one of the pre-7.2.0 uploads, but I may not have pushed all of those to alioth :(07:12
mlankhorsthm reverse debdiff it? :P07:13
mlankhorstseems you're right07:14
RAOFgit merge to the rescue.07:14
RAOFrobert_ancell: Whoops. Sorry - the armhf build didn't finish before the i386 & amd64 build for the *next* revision. No copy-with-binaries available for us!07:15
RAOFmlankhorst: I'll fix that up, push to git, and upload.07:16
mlankhorstgit merge is awesome ;P07:17
mlankhorstdo you have dpkg-mergechangelogs set up for git merges too?07:19
RAOFIndeed I do.07:28
didrocksRAOF: ok, let me find an archive I can restore07:29
didrocksare you free to debug Mir?07:29
RAOFdidrocks: Uuur, I'll be doing Zoë's bath in a 10 minutes or so.07:29
RAOFI can be available later, or if the machine won't be free then I guess I could get Sam to handle Zoë's bath.07:30
alf__RAOF: quick question: can an output have more than one preferred modes?07:33
RAOFalf__: I don't believe that makes sense, no.07:33
didrocksRAOF: should be freed, how long will that take you?07:34
didrocksto know if I'm going out for exercise now or later07:34
didrocksor just prepare and block the machine and ping you with the details07:34
RAOFProbably prepare and block the machine & ping me with the details would be best for me.07:35
RAOFI'll probably be ready to start debugging in ~1hr or so.07:35
didrocksRAOF: ok, doing that07:35
* RAOF prepares to set up the VPN on his new laptop07:36
robert_ancellRAOF, damn!07:41
thomiRAOF: would you like it turned off forever?07:45
robert_ancellAnd the reason mir rebuilt? - didrocks daily landing bumped debian/changelog :)07:46
didrocksrobert_ancell: sorry, was it a question? ;)07:46
robert_ancelldidrocks, :P07:46
thomiRAOF: I disabled the jenkins job - let me know if/when you want it turned on again07:47
robert_ancellduflu, have you given up with mir+raring? I want to disable all the raring packages from the staging PPA07:48
duflurobert_ancell: No, still using it for now. But I know I lost that argument07:48
robert_ancellduflu, well, it hasn't built for raring in some time, so you must only be building it locally right?07:48
duflurobert_ancell: Yes, always locally. I guess I need the PPA only for the Mesa/Intel changes07:49
robert_ancellthomi, can you disable mir and unity-system-compositor raring builds in the staging PPA? (no rush)07:49
robert_ancelland unity-greeter too I guess07:50
thomirobert_ancell: can get to that tomorrow, sure.07:50
thomirobert_ancell: just raring, right>07:50
thomiok, made a note. Will do that first thing tomorrow.07:51
dufluping alf__08:06
robert_ancellcd bzligd/sea08:40
RAOFdidrocks: Ping? New laptop has a new ssh key :(08:45
seb128RAOF, he went out for exercice, he should be back in half an hour08:46
RAOFseb128: Ah. Are you able to shove another ssh key on the box we ssh into?08:47
seb128robert_ancell, hey, if "jetessrter" is a password of yours, change it :p (you typed that and a bunch of commands in IRC)08:47
seb128RAOF, I'm afraid I don't know how to do that, maybe jibel can help though08:47
seb128jibel, ^08:47
jibelRAOF, which box?08:48
RAOFWhatever is.08:49
RAOF(In the QA lab, I think.)08:49
jibelRAOF, you don't seem to have an account on this box, which user do you use?08:49
seb128RAOF, how is your new laptop btw? you got one of the new system76 ones right? they look quite nice08:51
RAOFseb128: They are quite nice.08:51
robert_ancellseb128, heh, must fix that08:52
seb128robert_ancell, ;-)08:52
RAOFNot quite the same build quality as a high-end thinkpad, but servicable, with a nicer screen and cheaper :)08:52
seb128RAOF, it's renewal time for me so I'm looking around, though I'm still happy with my 3 years old latitude ... I might go for an ultrabook (the xps13 looks nice, but glossy screen ...)08:53
RAOFMatte screen on the galago :)08:54
jibelRAOF, I re-imported your ssh keys, can you try again?08:54
RAOFjibel: Works, thanks!08:55
RAOFOh, hello.09:03
RAOFWhy is mir_wait_for blocking indefinitely, I wonder?09:03
RAOFWow. unity-system-compositor has 17 threads.09:05
RAOFduflu: You were playing around in the mir_wait_for pool before - any ideas why mir_wait_for would be blocking indefinitely?09:06
dufluRAOF: Assuming you don't have a logic error, I do recall the design of mir_wait_for is actually not thread safe :/09:07
dufluRAOF: Perhaps mir_wait_for_one() ... http://unity.ubuntu.com/mir/group__mir__toolkit.html#ga4f9ee1ace58423c5482e9b301806025209:09
RAOFHm. That's probably not it, because we only mir_wait_for in the xserver mainloop09:09
alf__duflu: pong, sorry lost in console land09:12
alf__alan_g: @internal_surface, how come we can call as_internal_surface() without the namespace?09:13
duflualf__: I will defer bothering you while I suspect I might have caused the issue :)09:13
alf__duflu: ok09:13
alan_galf__: ADL09:14
alf__alan_g: ok09:17
RAOFduflu: Why “while ((!expecting && !received) ” in wait_for_all?09:19
RAOFduflu: This seems to be a race condition?09:20
dufluRAOF: Don't know. I thought that was well tested...09:20
dufluRAOF: It's in a unique_lock, so can't race. Only be misused by callers...09:21
RAOFIt seems to me that if you call mir_wait_for(mir_connection_do_something()); and your timeslice runs out between mir_connection_do_something and mir_wait_for then it's possible for the request to be processed before mir_wait_for *acquires* that lock.09:21
dufluRAOF: It looks like the kind of situation wait_for_one was invented to solve. The problem with wait_for is that it is unsafe to ever have more than one thread call it09:22
duflu... cos if you do, the first will win and starve subsequence threads09:22
RAOFBut this *does* only ever have one thread call it!09:22
dufluRAOF: Is your received < expecting?09:23
dufluWhat is expecting?09:23
RAOFNo; both are 009:23
dufluRAOF: That means no result_received yet09:24
RAOFduflu: Surely it doesn't?09:24
dufluRAOF: It's pretty clear in mir_wait_handle.cpp09:24
RAOFI would have thought expecting > received was no result_received.09:24
RAOFOh, no; you're quite right.09:26
dufluRAOF: Any result_received is a result. It's just a matter of how many you expect (are committed to wait for)09:26
RAOFIf it had been received then there'd be non-zero numbers there.09:26
dufluRAOF: I'm not saying all uses of MirWaitHandle are bug-free, but it looks likely the class itself is.09:38
RAOFWhich suggests that the request just hasn't been processed.09:44
RAOFdidrocks: If you need that box back I've probably got enough to think about.09:46
didrocksRAOF: ok, rebooting it to unscrew it now :)09:46
didrocksthanks RAOF09:47
RAOFBah. Where was that bug again? I should update it.09:50
alf__RAOF: it seems that KMS supports multiple preferred modes...09:53
RAOFHuh. You live and learn.09:53
alf__RAOF: but xrandr hides this, probably selects the first/highest preferred09:54
alf__RAOF: I wonder how it gets this from EDID information, I thought it had only one preferred mode field?09:55
RAOFI knew that I could technically set as many "preferred" modes as I liked; it's just a flag I can apply to a mode. I didn't think it would ever have more than one, though.09:55
RAOFalf__: Likewise.09:55
RAOFalf__: When you say kms supports multiple preferred modes, what do you mean?09:58
RAOFOr, rather - have you seen it *return* multiple preferred modes?09:58
alf__RAOF: I mean that on both my laptop and desktop I get multiple modes with the preferred flag set, yes09:58
RAOFI'm surprised you get more than one mode *at all* on your laptop, frankly :)09:59
alf__RAOF: actually you are right, lvds screen, gives me just one, it's the external screen that supports more there09:59
RAOFOk. It's not clear to me what the hell's happening there.10:30
RAOFIf anyone's got a few cycles spare to be perplexed, https://bugs.launchpad.net/mir/+bug/1204939 is really odd.10:31
ubot5Launchpad bug 1204939 in Mir "Unity doesn't start on ATI test machine (fails to find GL acceleration) (logind fails to track session?)" [Critical,Triaged]10:31
tvossalan_g, ping10:39
alf__mlankhorst: Hi! Any idea if it's normal for KMS to return multiple "preferred" modes (DRM_MODE_TYPE_PREFERRED set in mode->flags)? A quick survey of the drm kernel code (e.g. for radeon) indicates that this shouldn't happen (unless I missed something).10:55
mlankhorstalf__: why do you ask?10:59
mlankhorstfirst preferred mode would be the correct one, if there is more for the same connector it's probably a bug11:01
duflualan_g: The timing failure on N4, is that all the time or intermittent?11:10
alan_gduflu: happened twice - am increasing sample size11:11
duflualan_g: OK, yeah. I think the test is more at fault than the code. I'm going to loosen the test today and work on improving it another day11:12
alan_gduflu: failed 3 out of 1011:14
duflualan_g: What was the highest number reported of the 3?11:14
alan_gduflu: all 1 vs 111:15
duflualan_g: Right, so worst case was ~1% hiccups. That's the test being too strict, or not running long enough (since I shortened everything today)11:15
duflutvoss: Done, methinks11:17
dufluAt least that took long enough that the meat I was meant to cook is fully defrosted for sure11:18
alf__mlankhorst: on both radeon and intel I get more than one preferred mode for outputs11:18
* alan_g wonders why http://unity.ubuntu.com/mir/using_mir_on_android.html still says "stop" - didn't duflu correct that?11:18
duflualan_g: Yes but I don't know how the web gets updated11:19
alan_gduflu: it ought to be generated from trunk11:19
duflualan_g: Wait, no. That's a doc file I forgot to fix11:19
alan_gNot sure how often11:20
alan_gduflu: np - go eat11:20
alan_gtvoss: How do we zap surface flinger in the current image?11:24
alan_gNM I found https://wiki.ubuntu.com/Touch/Testing/Mir#Switch_from_SurfaceFlinger_to_Mir11:53
=== alan_g is now known as alan_g|lunch
mlankhorsthmz any open bugs I can look at?12:21
kgunnmlankhorst: if you'd like to pick up where RAOF has left  off https://bugs.launchpad.net/mir/+bug/120493912:29
ubot5Launchpad bug 1204939 in Mir "Unity doesn't start on ATI test machine (hang in mir_wait_for())" [Critical,Triaged]12:29
didrockskgunn: I noticed that unity_support_test tool is stuck in fact12:32
didrocks(it's launched by compiz)12:32
didrocksso I guess it's the first gl app which is blocking12:32
mlankhorstkgunn: what is that machine btw?12:57
=== alan_g|lunch is now known as alan_g
kgunnmlankhorst: its "otto" which didrocks or jibel can help get you access13:01
mlankhorstdidrocks/jibel: ok, I have no idea what's going on from the logs so if I could get access^13:03
didrocksmlankhorst: kgunn: the machines are under heavy use right now13:03
didrockskgunn: once we'll have asac's 3h daily release, we won't be able to all to give back access13:03
didrockswhich we are going to plug soon13:03
didrocksso it's the last calls, this time, please really looks at the logs, later will be too late :)13:03
didrocksmlankhorst: in ~1h, I should be able to block it13:04
didrockswill you be around?13:04
didrocksmlankhorst: you have the VPN access right?13:04
mlankhorstI'll be here in 1h, I'm on some vpn to qalab13:04
didrocksmlankhorst: what's your ssh key I should import?13:05
kgunndidrocks: understand....but, we've been runnning xmir fine on other ati machines so the problem seems specific to this one13:10
didrockskgunn: we know what happens with "specific to this one" which then happens to multiple configs, I'm not that idealist ;)13:10
didrocksif on one of the 2 machines we have, it happens (and traditional Xorg is working with acceleration), there is probably something that can affect other13:11
kgunndidrocks: we tried yesterday with 3 others with no problems13:11
kgunn3 other ati that is13:11
didrockskgunn: as Unity7, I never uploaded a Unity7 that was screwed on my machines at home13:11
didrocksthen… you see that even testing on 4 with the 3 cards were not enough :p13:11
didrocksso I wish we can settle that down13:11
seb128didrocks, I don't think anyone is trying to deny there is an issue, it's just harder to debug when the hardware you have local access to doesn't reproduce the issue13:12
kgunndidrocks: nvmd, our exchange doesn't seem helpful13:12
didrocksseb128: that exactly why I'm proposing access for the past 2 weeks13:12
kgunndidrocks: again not helpful13:13
seb128kgunn, what would be helpful to debug, out of access to the machine we have which hit the issue?13:13
asackgunn: didrocks: i guess you might want to align the exact execution environment so you see the same thing13:14
kgunnseb128: unfortunatley, we're spanning time zones (and we weren't helping ourselves with xorg churn...but we're past that)...so now we need to iterate13:14
kgunnasac: we've verified multiple environ's on other machines....it could even be the specific card/family at this point13:15
asaconce you are sure you do the exact same thing, investigating difference in behavioru becomes relevant13:15
asackgunn: right it could. just want to ensure that we are sure its the different card/family13:15
asacand not something else we do different still13:15
didrocksasac: it's passing with the same packages and setup on the intel machine13:15
kgunnasac: were you on this channel yesterday?13:16
asacthat doesnt mean you run stuff in the exact same way that kgunn is running them13:16
asackgunn: i was in this channel, but not following13:16
didrockskgunn: wasn't the same with the race issue? you never reproduce it before we spot it as dailies and after that we discovered that a lot of devs got it but juts reboot? (we had the same argument of "we never saw that, it's only you")13:16
didrockssorry for being extra cautious, but I think we should understand what's happen before pushing u-s-c13:17
kgunndidrocks: ok, let's try it otherway around....would you mind setting up lexington machine via robotfuel to be exactly like otto ?13:17
kgunnthis way there is no more questions about setup ?13:17
kgunndidrocks: that would be helpful....what do you think?13:17
kgunndidrocks: to be clear...i am not asking you to push u-s-c....i am trying to repro the problem with ott13:18
didrockskgunn: first, as we still don't have those 3h-dailies, I'll be able to give access to mlankhorst  and block every stacks on this13:18
didrockskgunn: I don't know if they have another ati machine with the same card?13:18
mlankhorstwhat is that machine btw?13:18
kgunndidrocks: got it...but i am anticipating we still won't have it solved13:18
* kgunn is a realist13:18
didrocksmlankhorst: the one I pasted the other day, you will have access13:18
didrockskgunn: let's hope we know what happens first, RAOF spent more than 1h on it this morning13:19
robotfuelkgunn: I don't think we have a system with the same card13:19
didrockskgunn: then, yeah, we need a more dynamic QA lab for stuff running on real hardware13:19
kgunndidrocks robotfuel ....why not make sure the one in lexington lab has the same steup/run of a trademark didrocks run ??13:20
kgunnwho cares if the card is different....13:20
kgunnthis goes back to your post a moment ago about teams claiming that stuff isn't broken only to find out it is13:20
didrocksjibel: can you install otto on that one? as it seems kgunn and asac may think otto is the cause of this?13:21
didrockskgunn: I just wonder as we would have the same issue with intel, or even unity won't even run on traditional Xorg on ati13:21
didrockswhich it does ~20 times a day for multiple hours for the past 5 months13:21
kgunndidrocks: jibel .....i am more suspicious of it being a potential ati-issue13:21
didrockslike that driver/card in particular? (completely possible)13:22
mlankhorstmy ati works just fine13:22
kgunndidrocks: could be a combo - your specific setup + your specific test run + ati driver (general even)13:22
mlankhorstoh wait PROVE_LOCKING is disabled due to some regression, grrr13:22
kgunnmlankhorst: yeah...i'm just trying to nail the exact environ + test run eliciting the behavior13:23
didrockskgunn: no test are run at this point as the session never successfully fully comes up, but yeah, setup + ati driver13:24
jibeldidrocks, install otto on which one? if there is a box remotely accessible I can deploy the test setup on it of course13:24
didrocks(knowing that it's working on Xorg + unity/compiz on that card)13:24
didrocksrobotfuel: ^13:24
mlankhorsthmz :p13:25
mlankhorst    drm: Don't pass negative delta to ktime_sub_ns()13:26
mlankhorst    13:26
mlankhorst    It takes an unsigned value. This happens not to blow up on 64-bit13:26
mlankhorst    architectures, but it does on 32-bit, causing13:26
mlankhorst    drm_calc_vbltimestamp_from_scanoutpos() to calculate totally bogus13:26
mlankhorst    timestamps for vblank events. Which in turn causes e.g. gnome-shell to13:26
mlankhorst    hang after a DPMS off cycle with current xf86-video-ati Git.13:26
robotfueljibel: didrocks you can use ps-radeon-hd6870-he13:28
mlankhorstmakes me wonder if it only happens to blow up on i386 or not ;)13:28
alan_galf__: any ideas on how best to progress https://code.launchpad.net/~alan-griffiths/mir/fix-1208774/+merge/178925?13:31
alf__alan_g: I would be good to find out what the problem is (probably just a linker bug), but I think the workaround is ok for now.13:33
kgunnrobotfuel: jibel didrocks mlankhorst ....the key is to setup & "start" the run on ps-radeon-hd6870-he in exactly the same manner13:34
kgunnas otto13:34
alan_galf__: Feel like voting then?13:35
alf__alan_g: sure13:35
jibelkgunn, undertood13:36
mlankhorsthow do I connect to that one btw?13:36
kgunnrobotfuel: ^ ?13:37
robotfuelmlankhorst: https://wiki.canonical.com/UbuntuEngineering/QA/Lab ctrl-f and search for ps-radeon-hd6870-he13:38
kgunnjibel: mlankhorst robotfuel ...thanks for all this....heroes you are!13:40
kgunnor will be :)13:40
mlankhorstis something supposed to happen when i hit connect?13:43
mlankhorstoh there we go13:44
mlankhorstkgunn: hm default xmir works?13:50
kgunnmlankhorst: ....can we ensure our boot and attempt to programatically run are the exact same to didrocks on otto, e.g. he ran unity_support_test13:54
didrocks(compiz does in fact)13:54
didrocksmlankhorst: the machine will be available in 5 minutes13:54
=== chihchun_afk is now known as chihchun
kgunndidrocks: does it run that every boot regardless? (e.g. we don't need to enable any boot script or something)13:58
=== greyback_ is now known as greyback|latelun
didrockskgunn: if you start an unity session, it's ran14:00
kgunndidrocks: :-/....damn, back to otto14:01
kgunndidrocks: what is the specific card on that again (i'll note it in the bug)14:02
didrockskgunn: one sec, launching the tests ASAP and noting it on the bug14:03
mlankhorstcuriously is otto a machine without its outputs connected to anything?14:05
mlankhorsti suppose I could test that setup locally14:10
=== chihchun is now known as chihchun_afk
=== chihchun_afk is now known as chihchun
mlankhorsthm my machine died without output connected14:19
tvossmlankhorst, kgunn, didrocks any more insight into the ati issue?14:19
kgunntvoss: right now...suddenly seems to work....didrocks suspecting hybris14:20
kgunngetting solved14:20
mlankhorst<4>[   60.988003] other info that might help us debug this:14:20
mlankhorst<4>[   60.988003]  Possible unsafe locking scenario:14:20
mlankhorst<4>[   60.988003]14:20
mlankhorst<4>[   60.988003]        CPU014:20
mlankhorst<4>[   60.988003]        ----14:20
mlankhorst<4>[   60.988003]   lock(&mm->mmap_sem);14:20
mlankhorst<4>[   60.988003]   <Interrupt>14:20
mlankhorst<4>[   60.988003]     lock(&mm->mmap_sem);14:20
mlankhorst<4>[   60.988003]14:20
mlankhorstbut no idea wtf is going on there14:20
kgunnmlankhorst: did i speak too soon....is that otto or your local  ?14:21
didrockssecond run, works…14:21
didrocks(well started)14:21
didrocksthe machine didn't die14:21
mlankhorstno idea wtf is going on there14:21
kgunntvoss: then didrocks gonna check diff between yesterday & today14:22
kgunntvoss: ....enjoying your afternoon off :)14:22
mlankhorstnew mir patch for ati at least14:22
mlankhorstmeh list corruption in nfs here, grrr14:23
mlankhorstwhy does nfs have to add a new corruption every time I update my kernel :/14:25
didrockskgunn: didn't yet look at it, before yesterday-today: http://paste.ubuntu.com/5962760/14:26
=== chihchun is now known as chihchun_afk
mlankhorstdidrocks: yeah ubuntu2 ati, might have been what fixed it14:28
didrocksmlankhorst: interesting14:28
=== alan_g is now known as alan_g|tea
mlankhorstbut no idea really what was going on14:29
didrockscan be that or Mir itself14:30
mlankhorsthm fun, another crash at a random place in the kernel14:32
mlankhorstI'm definitely hitting some nasty fd corruption on this linux/master + airlied/drm-fixes kernel14:42
=== alan_g|tea is now known as alan_g
mlankhorstother machine is fine though :/14:48
didrocksok, so can't reproduce anymore14:50
didrocksafter 3 runs14:50
didrocksand then one run to regular Xorg with another stack14:50
didrocksno hang14:50
didrocksno machine getting crazy14:50
* didrocks just removes a flacky test from the list and will push u-s-c to universe14:50
mlankhorstok I think it's radeon's fault on my machine at least, I swapped out the card for another one and that machine works14:51
kgunnjibel: robotfuel didrocks mlankhorst ...thanks for all the help14:52
didrockskgunn: yw14:52
* kgunn 's beer/food list is getting long14:52
mlankhorstmy 6570 hangs really badly, 5570 works14:53
mlankhorstgetting really bad kernel memory corruption with the 6570, so I'll have to look at that later14:53
kgunnmlankhorst: thanks...14:54
jibelkgunn, yw, thanks to whoever fixed this crash14:54
kgunnrobotfuel: was there a Northern Islands (HD 6xxx) Series ati in lexington working ok ?....just tieing this back to mlankhorst complaint about 657014:56
robotfuelkgunn: we have a 6870, but it's not being used on the new server. openarena worked on it before with xmir14:57
mlankhorstI'm close to upstream though, hm perhaps even ahead of upstream, I'll try without some patches :P14:58
robotfuelkgunn: it's the one jibel is testing with14:58
mlankhorstoh i was already testing without those14:58
kgunnrobotfuel: cool....yeah, that's definitely same family14:58
mlankhorstbut in my case it's definitely a kernel corruption14:59
mlankhorstno way userspace would screw up this badly :P15:00
smartboyhwmlankhorst, in http://unity.ubuntu.com/mir/using_mir_on_pc.html it says in Running Mir Natively a "some-mir-client" command.15:03
smartboyhwWhat does that mean?15:03
smartboyhwolli_, ^15:05
olli_smartboyhw, in a meeting, kgunn^15:06
kgunnsmartboyhw: like for instance mir_demo_client_egltriangle15:12
smartboyhwkgunn, oK15:12
smartboyhwkgunn, whoa nice!15:13
kgunnsmartboyhw: there's several there15:15
smartboyhwkgunn, you should provide a list to us:P15:15
=== greyback|latelun is now known as greyback
alan_gsmartboyhw: ls mir_demo_client*15:34
smartboyhwalan_g, great, thanks15:34
kgunnracarr: can you join us ??15:35
=== om26er is now known as om26er|away
=== greyback is now known as greyback|dinner
=== alan_g is now known as alan_g|EOD
=== om26er|away is now known as om26er
kgunnracarr: ping18:07
kgunnracarr: ping (sorry...i kept rebooting i know)18:26
racarrkgunn: pongish18:42
racarrHey wait I'm supposed to be racarr|dentist18:42
racarrwhat's up?18:42
kgunnracarr: no worries...hopefully its a "good" dentist experience & not a "painful" one18:45
racarrmedium lol18:45
kgunnracarr: was just curious...we were wondering on surface configurator if mir is to control the osk visibiltiy with the minimize attribute18:45
racarrkgunn: Yeah! I think that's the idea.18:46
racarrwe need to hook it up to qtubuntu18:46
kgunnracarr: cool greyback|dinner ...dang he's at dinner ^18:46
racarrbut dandrader was telling me mallit just calls like QWindow show/hide18:46
racarrso, the surface configurator actually does two things for the OSK:18:46
racarr1. Let's it recognize the overlay surface type (and approve/dissaprove setting this) with select_attribute_value18:46
racarr2. Implement minimize/restore with the attribute_set interface18:47
kgunnracarr: that's perfect!18:47
kgunnthanks for pushing that...18:47
racarrI have to go again soon. I failed to bring proof of residence to the DMV yesterday so trying again today18:51
racarrDrivers license expires tomorrow so if I don't do it today I have to take a driving test ><18:51
racarrI think, based on kdubs branch though, I can submit a much simpler version of client-focus-notifications18:51
racarrhopefully this evening or tomorrow18:51
=== greyback|dinner is now known as greyback
greybackkgunn: ack, all sounds good to me19:11
kgunnbschaefer: thanks for the help yesterday!....glad it "resolved itself"20:05
bschaeferkgunn, haha np, as I am...and I hope it stays that way20:06
bschaeferand yay u-s-c is in main :)20:06
kgunnbschaefer: i know....feels naughty!20:06
robert_ancellthomi, hey, did you forward that itinery?21:10
racarrnext time you guys come to oakland I can not recommend the california DMV21:10
racarrfor a fun afternoon21:10
kdubracarr, in lexington next month, i'll show you my license with my dmv picture21:11
kdubpretty funny how evidently angry i am in it :P21:11
robert_ancellkdub, ha!21:12
racarrapparently when the person yesterday told me I needed a copy nof my birth certificate21:12
racarrthey meant in particular, the original copy21:12
thomirobert_ancell: we're having trouble finding flights from bos -> SFO or LAX that don't involve a huge stopover. Should have final flights tomorrow21:12
robert_ancellthomi, k21:13
robert_ancellthomi, is it faster to go via NYC?21:15
thomirobert_ancell: NYC is kind of in the wrong direction21:20
thomiI have faith that BTS will sortit out. Worst case is 10 hours in SFO - maybe time to do a bit of sightseeing :)21:20
racarrthomi: That's just long enough to get a drivers license at the DMV!21:23
thomiracarr: I already ahve two drivers licenses, why would I want a third?21:23
racarrthomi: :)21:24
jonohey guys21:24
jonoFYI: to track flavor bugs I am going to ask flavors to tag their bugs with 'flavormirbug'21:25
jonothe resulting search with these bugs is at https://bugs.launchpad.net/bugs/+bugs?field.searchtext=&orderby=-importance&field.status%3Alist=NEW&field.status%3Alist=CONFIRMED&field.status%3Alist=TRIAGED&field.status%3Alist=INPROGRESS&field.status%3Alist=FIXCOMMITTED&field.status%3Alist=INCOMPLETE_WITH_RESPONSE&field.status%3Alist=INCOMPLETE_WITHOUT_RESPONSE&assignee_option=any&field.assignee=&field.bug_reporter=&field.bug_commenter=&field.subscriber=&fie21:25
jonokgunn, ^21:25
jonois there any chance we could get #1208250 on a priority list?21:25
thomihaha, nice URL21:25
jonooh, it is Critical21:25
jonobut seems not assigned21:26
jonothomi, seriously21:26
robert_ancelljono, cheers21:35
jonorobert_ancell, np21:35
jonothey are choosing whether to go with Mir on the 22nd21:35
jonoso I think if we can resolve most of their issues by then, it should encourage the Xubuntu community to ship Mir21:36
thomirobert_ancell: robotfuel is still hitting this issue in the xmir tests: https://bugs.launchpad.net/xmir/+bug/120900022:11
ubot5Launchpad bug 1209000 in XMir "radeon hd7850 fails to load driver from system-compositor-testing ppa with gbm: failed to open any driver (search paths /usr/lib/x86_64-linux-gnu/dri:${ORIGIN}/dri:/usr/lib/dri)failed to load driver: radeonsi" [Undecided,Fix released]22:11
robotfuelthomi: I still need to check if that's the case with the new drivers,  this is the bug I was talking about https://bugs.launchpad.net/ubuntu/+source/xserver-xorg-video-ati/+bug/120939722:12
ubot5Launchpad bug 1209397 in xserver-xorg-video-ati (Ubuntu) "[radeonsi] radeonhd "southern islands" 3d hardware acceleration" [Undecided,Confirmed]22:12
thomirobotfuel: ahhh ok22:13
robotfuelthomi: it looks like they are related22:13
thomirobert_ancell: so that card is listed on the xmir go/no-go acceptance criteria22:13
thomirobert_ancell: so it's needed before we turn xmir on for 13.1022:13
thomirobert_ancell: I wonder if you can use your team-lead super powers to poke someone to fix this for us?22:14
robert_ancellthomi, otp22:14
robotfuelthomi: there is actually a new bug, x crashes now with /usr/bin/X: symbol lookup error: /usr/lib/xorg/modules/drivers/radeon_drv.so: undefined symbol: exaGetPixmapDriverPrivate22:15
thomirobotfuel: OK, can you please make sure that all the relevant bugs are linked in the SS, so I can keep bugging people until they get fixed?22:16
robotfuelthomi: yes everything but that bug is to date22:17
thomirobotfuel: OK, so there's 3 issue, the two bugs in the SS (both linked above), and the third that you've just mentioned?22:18
robotfuelthe 1209000 isn't happening anymore,  now it's undefined symbol: exaGetPixmapDriverPrivate22:20
robotfuelthomi: I'll ping you when apport is done uploading the bug22:20
robert_ancellthomi, robotfuel, ah, RAOF said that one looked like exa support wasn't loaded22:21
robert_ancellplease assign to RAOF for triaging once its up22:21
robert_ancellWhat are the other two issues?22:21
racarrkdub: Have been going through connect-display-request btw. I like it :)22:22
kduboh, great :)22:22
robotfuelrobert_ancell: it's one other issue not 222:22
racarrwill rework client-focus-notifications on the refactoring there to get rid of the event sink changes22:22
robotfuelrobert_ancell: this is the other issue https://bugs.launchpad.net/ubuntu/+source/xserver-xorg-video-ati/+bug/120939722:22
ubot5Launchpad bug 1209397 in xserver-xorg-video-ati (Ubuntu) "[radeonsi] radeonhd "southern islands" 3d hardware acceleration" [Undecided,Confirmed]22:22
racarrI am starting to wonder though...22:22
racarr(not for this branch, just in general)22:22
robert_ancellrobotfuel, that card has issues on traditional X right?22:23
racarrperhaps SessionManager isn't mf::Shell22:23
robotfuelrobert_ancell: yes https://bugs.launchpad.net/ubuntu/+source/xserver-xorg-video-ati/+bug/1209397 is a traditional x bug22:23
racarryou know? There are kind of different roles between this handle_surface_created, and open_session type stuff22:23
racarrand I'm more inclined to call the handle_surface_created, etc, the 'shell'22:24
kdubracarr, yeah, should make picking out right where the focus occurred a bit easier22:26
kduband should also let you write an implementation of focus that depends on events22:26
kdubso it can be like,  z-order based in a simple case, or as complicated a class you can think up22:26
kdubbeing in this part of the system for a bit pointed out to me how the shell's grand data structure22:27
kdubis the SessionContainer22:27
kdubwith mir's grand data structure being the SurfaceStack22:27
kdubthings are starting to shape up along those lines :)22:27
robotfuelthomi: spreadsheet is up to date22:28
racarrAlan thinks the session assosciation should be in the SurfaceStack (SceneGraph)22:28
racarrand I'm not totally sure yet.22:28
thomithanks even22:28
racarron one hand it might make certain synchronization issues, and exposing stuff out to the shell, etc, easier and safer22:29
robert_ancellthomi, did you disable the raring builds from the testing ppa?22:29
racarron the other hand 1. I'm not sure how to do it and 2. I kind of like the way that the shell functions as22:29
robotfuelthomi: I am going to mark https://bugs.launchpad.net/xmir/+bug/1209000 fixed released because it's not happening anymore22:29
racarrsomething on top of the SurfaceStack22:29
ubot5Launchpad bug 1209000 in XMir "radeon hd7850 fails to load driver from system-compositor-testing ppa with gbm: failed to open any driver (search paths /usr/lib/x86_64-linux-gnu/dri:${ORIGIN}/dri:/usr/lib/dri)failed to load driver: radeonsi" [Undecided,Fix released]22:29
robotfuelheh it was already marked fixed release :P22:31
kdubracarr, i'm not sure either about that, but thats an initial reaction22:31
racarrkdub: Perhaps part of the session manager is really part of the session container22:44
racarrand the rest is mf::Shell22:44
racarri.e.open_session/close_session anyway22:44
racarrthen the shell gets on_session_opened, on_session_closed sort of stuff22:44
kdubwell, the session manager is being reduced to the session factory22:46
kdubcreate_surface_for can be eliminated if we have a "SessionMediator-like" class for our internal clients22:47
RAOFGah. Why is BTS 50% more expensive than flights I could book myself?22:48
robert_ancellRAOF, yeah, I always get that too23:06
robert_ancellRAOF, we really should have an Australasian travel agent23:06
RAOFAlso. UA - just say no.23:07
racarrkdub: mm. maybe there is a SessionMediator that can work for internal and external client sort of object23:08
racarrand the RPC also has a "socket mediator" or something23:08

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