IRC logs for #buildstream for Thursday, 2019-12-12

« Wednesday, 2019-12-11 Index Friday, 2019-12-13 »
*** slaf has quit IRC01:46
*** slaf has joined #buildstream01:50
*** slaf has joined #buildstream01:50
*** slaf has joined #buildstream01:51
*** slaf has joined #buildstream01:51
*** slaf has joined #buildstream01:51
*** slaf has joined #buildstream01:51
*** slaf has quit IRC01:51
*** slaf has joined #buildstream01:52
*** slaf has joined #buildstream01:52
*** slaf has joined #buildstream01:52
*** slaf has joined #buildstream01:52
*** slaf has joined #buildstream01:53
*** slaf has joined #buildstream01:53
*** slaf has joined #buildstream01:53
*** slaf has joined #buildstream01:54
*** slaf has joined #buildstream01:54
*** slaf has joined #buildstream01:54
*** slaf has joined #buildstream01:54
*** slaf has joined #buildstream01:55
*** slaf has joined #buildstream01:55
*** slaf has joined #buildstream01:55
*** slaf has joined #buildstream01:56
*** slaf has joined #buildstream01:56
*** slaf has joined #buildstream01:56
*** slaf has joined #buildstream01:56
*** slaf has joined #buildstream01:57
*** slaf has joined #buildstream01:57
*** phildawson_ has joined #buildstream05:11
juergbiMarge still complains even with the remote execution test passing :-/ "It is taking too long to see the request marked as merged!"06:36
gitlab-br-botjuergbi merged MR !1762 (jjardon/fdsdk-190805->master: .gitlab-ci.yml: Use latest freedesktop-sdk 19.08.5) on buildstream https://gitlab.com/BuildStream/buildstream/merge_requests/176206:37
*** phildawson_ has quit IRC07:35
*** benschubert has joined #buildstream08:33
*** bochecha has joined #buildstream09:07
*** bochecha has quit IRC09:08
*** bochecha has joined #buildstream09:09
*** bochecha has quit IRC09:20
*** bochecha has joined #buildstream09:21
*** santi has joined #buildstream09:45
*** toscalix has joined #buildstream09:51
*** jonathanmaw has joined #buildstream09:58
*** tiagogomes has joined #buildstream10:04
*** tme5 has joined #buildstream10:06
*** traveltissues has joined #buildstream10:09
*** traveltissues has quit IRC10:09
*** toscalix has quit IRC10:17
tlater[m]juergbi: Oh, maybe someone needs to kick marge then10:28
* tlater[m] checks her github page10:29
juergbitlater[m]: have you seen my comments on https://gitlab.com/BuildStream/buildstream/commit/5164eed10b66e1bf5e25fa5960e94c20ea414086 ?10:48
*** traveltissues has joined #buildstream10:49
tlater[m]juergbi: yeah, I just spotted them10:50
* tlater[m] agrees with both and will get that fixed up10:50
*** benschubert has quit IRC10:53
tlater[m]juergbi: Does this look good? https://gitlab.com/BuildStream/buildstream/commit/c4facc4b8bf46803df6866ebedfe296dc07373d110:58
tlater[m]If so, want me to push that to your branch?10:59
tme5!1765 is ready again :)11:11
gitlab-br-botMR !1765: Make Git source plugin clone submodules recursively https://gitlab.com/BuildStream/buildstream/merge_requests/176511:11
tme5at the moment, it seems the 'tags' configuration of the git source is used to recreate a .git database after the old .git from the clone is deleted. I would have thought that the clone .git was complete and so didn't need rebuilding. What am I missing here?11:29
*** lachlan has joined #buildstream11:32
juergbitlater[m]: yes, looks good to me, ta. sure, you can push it to my branch or I'll cherry-pick it11:35
tlater[m]juergbi: I suppose if you cherry-pick we won't end up needing to keep it in sync with master anyway11:36
tlater[m]I'll just keep around the branch then11:36
juergbiok, will pick it11:36
juergbitraveltissues: I see the remote execution CI job succeeding again, e.g., https://gitlab.com/BuildStream/buildstream/-/jobs/37648200611:39
juergbiwhat issue are you still seeing?11:39
*** lachlan has quit IRC11:40
tlater[m]juergbi: I think we're running into this: https://github.com/smarkets/marge-bot/issues/10111:47
tlater[m](Regarding marge)11:47
juergbitlater[m]: but that's about CI failing11:48
juergbimarge complained with CI all green11:48
juergbihttps://gitlab.com/BuildStream/buildstream/merge_requests/176211:49
tlater[m]I read that as marge failing whenever the "automatically merge when pipeline succeeds" option is set11:49
juergbithat was not set11:49
juergbiI merged it manually after Marge complained11:50
tlater[m]Hrm11:50
* tlater[m] wonders what the timeout for that message is11:50
tlater[m]tpollard: Any idea where we could check marge's configuration?11:51
tpollardI don't tlater[m]11:51
juergbiMarge's message was 5 minutes after the pipeline succeeded, afaict11:52
tlater[m]juergbi: Timeout is 15 minutes by default, that pipeline took 45 minutes11:52
*** lachlan has joined #buildstream11:52
tlater[m]... and it's changed recently11:52
tlater[m]It's possible someone updated marge but didn't realize that new timeout is in effect11:53
juergbiI think we've configured Marge to a 2h timeout or so for CI to complete11:53
juergbishe complained 5 minutes _after_ the pipeline completed, though, so I don't see how this could be the issue11:53
tlater[m]juergbi: The flag that sets the timeout has changed, if the config is still the same as before that change, we'd be seeing timeouts11:54
tlater[m]It's possible the comment setting is just delayed or somesuch11:54
juergbiand I don't think we ever saw that message when the CI failed11:54
juergbiI'm fairly certain Marge correctly saw the pipeline passing11:54
juergbiand it's an issue with the actual merge11:54
tlater[m]:|11:55
tlater[m]Well, I guess we should wait and see if we have the same behavior on another passing MR11:56
juergbiyes, although the failing remote execution test shouldn't have caused that issue anyway. in which case we've seen this issue tons of times on passing MR11:56
* tlater[m] hopes both things caused this problem :D11:57
gitlab-br-bottpollard approved MR !1766 (juerg/job-pipe->master: job.py: Replace message queue with pipe) on buildstream https://gitlab.com/BuildStream/buildstream/merge_requests/176611:59
tlater[m]Guess there's the next candidate12:00
jjardonHi, is it possible to disable the remote cache through the command line, even if it's configured in project.conf?12:35
tlater[m]jjardon: --pushers 0 will have that effect12:36
tlater[m]I agree it's a little unintuitive, we have a UI rework issue for that.12:36
jjardontlater[m]: does this work for bst 1.x ?12:36
tlater[m]It should do12:37
tlater[m]We've not changed how that works since then12:37
jjardonok, thanks tlater[m] !12:37
tpollardthat won't disable it trying to pull though right?12:37
tlater[m]tpollard: Oh, right, I think --pullers 0 also exists?12:38
tlater[m]--fetchers 0, in fact12:38
tpollardin master you can give build --remote12:38
tpollardbut I don't think that was backported12:38
tlater[m]jjardon: ^ You need --pushers 0 and --fetchers 0 to completely disable remote caches in 1.x12:38
jjardonrigth, thanks!12:39
coldtomyeah, on master to disable remote caching i used `bst -r /dev/null build foo.bst`12:40
coldtom/dev/null can actually just be anything that doesn't get resolved though12:40
traveltissuesjuergbi, yes re seems to work again, i was looking at issues like https://gitlab.com/BuildStream/buildstream/-/jobs/37523850912:47
traveltissuesthere are still some issues with overnight12:48
traveltissuesi restarted the wsl runner last night since that seemed to have fallen over12:49
juergbitraveltissues: right, that job was shortly before the new buildgrid docker image was built12:53
juergbieverything should be working since yesterday Wednesday12:53
juergbidon't know about the WSL runner12:53
traveltissueswsl was timing out, it's working on the last merge12:55
jjardontlater[m]: mmm, seeems it's actually pulling: https://gitlab.com/freedesktop-sdk/freedesktop-sdk/-/jobs/37742722612:58
tlater[m]jjardon: :(13:00
tlater[m]That's probably a bug, or we didn't make --fetchers generic back then13:00
jjardonOk, is there any other way?13:01
tlater[m]You'd need to override the remote like coldtom suggests, I'm not sure there's a way of doing that from the CLI.13:02
jjardonOk, thanks13:06
*** bochecha has quit IRC13:13
*** tpollard has quit IRC13:18
*** tpollard has joined #buildstream13:19
*** santi has quit IRC13:50
*** benschubert has joined #buildstream13:52
*** tpollard has quit IRC14:00
*** tpollard has joined #buildstream14:05
*** santi has joined #buildstream14:05
*** lachlan has quit IRC14:42
*** lachlan has joined #buildstream14:44
*** phildawson_ has joined #buildstream14:52
*** phildawson_ has quit IRC15:03
*** phildawson_ has joined #buildstream15:04
gitlab-br-bottpollard opened MR !1767 (tpollard/shellbuildtree->master: _frontend/cli.py: Make show() --use-buildtree respect pull semantics) on buildstream https://gitlab.com/BuildStream/buildstream/merge_requests/176715:10
*** phildawson_ has quit IRC15:29
*** lachlan has quit IRC15:29
*** bochecha has joined #buildstream15:34
*** phildawson_ has joined #buildstream15:37
*** lachlan has joined #buildstream15:44
*** phildawson-ct has joined #buildstream15:49
*** phildawson_ has quit IRC15:50
*** traveltissues has quit IRC16:01
*** lachlan has quit IRC16:04
*** phildawson-ct has quit IRC16:05
*** lachlan has joined #buildstream16:07
*** phildawson-ct has joined #buildstream16:12
*** lachlan has quit IRC16:12
*** phildawson-ct has quit IRC16:15
*** bochecha has quit IRC16:17
gitlab-br-botBenjaminSchubert opened issue #1239 (BuildStream tries to push incomplete artifacts and fails) on buildstream https://gitlab.com/BuildStream/buildstream/issues/123916:25
*** lachlan has joined #buildstream16:26
*** lachlan has quit IRC16:29
*** lachlan has joined #buildstream16:37
*** santi has quit IRC16:44
*** lachlan has quit IRC16:45
*** lachlan has joined #buildstream16:50
*** tpollard has quit IRC17:05
*** santi has joined #buildstream17:09
*** phildawson-ct has joined #buildstream17:11
*** mohan43u has quit IRC17:25
*** mohan43u has joined #buildstream17:26
benschuberttlater[m]: #1238 the last comment, you had that and fixed it right?17:33
gitlab-br-botIssue #1238: Unable to build examples (Python 3.8) https://gitlab.com/BuildStream/buildstream/issues/123817:33
tlater[m]benschubert: I had the beginnings of a patch for that17:34
tlater[m]benschubert: Bug is still open though: https://gitlab.com/BuildStream/buildstream/issues/116417:35
benschubertthanks!17:36
*** lachlan has quit IRC17:38
*** lachlan has joined #buildstream17:43
*** lachlan has quit IRC17:51
*** lachlan has joined #buildstream17:55
*** tiagogomes has quit IRC17:58
*** tme5 has quit IRC18:03
*** santi has quit IRC18:07
*** lachlan has quit IRC18:13
*** mohan43u has quit IRC18:15
benschuberttlater[m]: are you around and have a minute?18:19
tlater[m]benschubert: Yep and yep18:19
benschubertAbout !1739, I've got my initial benchmarks results. The 'show' is indeed faster, even for our very simple debian-stack benchmark18:20
gitlab-br-botMR !1739: WIP: Optimize consistency and state handling https://gitlab.com/BuildStream/buildstream/merge_requests/173918:20
tlater[m]How much faster? :D18:21
benschubertHowever, the build is slightly longer (2-5%) which I am not 100% sure why. One thing I _could_ imagine is because of the new changes in `fetch_done`.  I was also therefore wondering if we should handle that part in the job itself (last part of 'fetch') and forcefully mark the thing as cached in fetch_done, trusting the child?18:22
benschubertfor the base-files, it was like 10% faster18:22
benschubertfor internal projects, I went from 1m30 to 15sec18:22
benschubertI'll also try to run benchmarks on freedesktop-sdk for that18:22
benschubertwould trusting the child seem something sensible there?18:24
tlater[m]benschubert: I thought about that when I was working on the element side of this as well18:24
tlater[m]Let me take a skim through the code again18:24
tlater[m]benschubert: If we're going with that, could we also move tracking_done?18:26
tlater[m]No, we need to update resolved state18:27
tlater[m]I think that makes sense, at this point we're really just checking after the child, and doing no computation at all in `fetch_done`18:27
tlater[m]benschubert: Have you tried benchmarking something without fetching?18:28
benschubertNot yet, I'll have to save some uninterrupted time to dig back on this :)18:29
benschubertFor track, we could do part of it in the child, part in the parent if needed18:29
benschubertah actually no18:29
benschuberttlater[m]: but if we were to gain more out of this, you would not see any reason for not doing this in the child right?18:33
*** mohan43u has joined #buildstream18:33
tlater[m]benschubert: No, in fact I think I'd prefer that for readability18:34
benschubertGreat, I'll test all of this then :)18:34
benschubertthanks!18:34
tlater[m]benschubert: I sure hope it doesn't all just fall apart now ;p18:35
benschubertyep :)18:35
benschubertI'll push that tomorrow probably :)18:36
benschubertand I might have a couple following PRs to clean more stuff18:36
benschubertI want to see whether we can get rid of the 'ready_for_runtime' and such now that we don't track while building18:37
*** jonathanmaw has quit IRC18:41
benschuberttlater[m]: Actually I believe something like https://gitlab.com/BuildStream/buildstream/merge_requests/1739/diffs?commit_id=49df4cad9bf83731cc26887e7a9dd07432b5456a might be quite neat :)18:48
juergbibenschubert: close and None ok for you as well for the second case (instance variables)? https://gitlab.com/BuildStream/buildstream/merge_requests/1766#note_25955867218:58
*** phildawson-ct has quit IRC20:04
gitlab-br-botjuergbi merged MR !1766 (juerg/job-pipe->master: job.py: Replace message queue with pipe) on buildstream https://gitlab.com/BuildStream/buildstream/merge_requests/176620:21
*** mohan43u has quit IRC21:36
*** mohan43u has joined #buildstream21:37
*** mohan43u has quit IRC21:52
gitlab-br-botalbfan closed issue #1238 (Unable to build examples (Python 3.8)) on buildstream https://gitlab.com/BuildStream/buildstream/issues/123822:04
*** mohan43u has joined #buildstream23:08
« Wednesday, 2019-12-11 Index Friday, 2019-12-13 »

Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!