| *** mohan43u has quit IRC | 02:35 | |
| *** mohan43u has joined #buildstream | 02:38 | |
| *** mohan43u has quit IRC | 03:03 | |
| *** tiagogomes has joined #buildstream | 03:26 | |
| *** tiagogomes has quit IRC | 03:33 | |
| *** tiagogomes has joined #buildstream | 03:34 | |
| *** tiagogomes has quit IRC | 03:35 | |
| gitlab-br-bot | BenjaminSchubert approved MR !1722 (valentindavid/compose-integration-commands->master: compose: Allow compose elements to run integration commands) on buildstream https://gitlab.com/BuildStream/buildstream/merge_requests/1722 | 09:36 |
|---|---|---|
| *** phildawson has joined #buildstream | 10:07 | |
| benschubert | Anyone waiting on reviews from me? Or needing a review? | 10:07 |
| *** lachlan has joined #buildstream | 10:29 | |
| *** lachlan has quit IRC | 10:40 | |
| tlater[m] | benschubert: I assume you have no new comments on !1718 | 10:44 |
| tlater[m] | ? | 10:44 |
| * tlater[m] is about to translate the benchmarking | 10:46 | |
| *** lachlan has joined #buildstream | 10:46 | |
| benschubert | tlater[m]: just interested in the benchmarks results, but otherwise everythign seems sane and will help me for my next improvement :) | 10:47 |
| tlater[m] | Oooh, what's the next improvement? | 10:47 |
| aevri | benschubert: if you're in the market for reviews, this guy could do with a final approval - !1638 | 10:48 |
| gitlab-br-bot | MR !1638: cascache: refactor, extract CASDProcessManager and CASDConnection https://gitlab.com/BuildStream/buildstream/merge_requests/1638 | 10:48 |
| benschubert | tlater[m]: removiing `get_consistency` and having `is_cached` only, and assume `consistenyt means has_ref`, which then allows to stop calling `get_consistency` in every part of the pipeline that is not workspace operations, which could save 2/3rd of the loading time in git cases (and potentially others) | 10:51 |
| tlater[m] | Yup, that sounds about right | 10:52 |
| benschubert | tlater[m]: stable is bst 1.4 right? | 11:18 |
| tlater[m] | 1.2 | 11:18 |
| tlater[m] | I believe the benchmark target for stable hasn't been updated in a while. | 11:19 |
| *** lachlan has quit IRC | 11:19 | |
| benschubert | tlater[m]: i'm surprised there is no benefits? Or is this just because we havent started the optimisations there? | 11:20 |
| tlater[m] | benschubert: You mean compared to stable? | 11:20 |
| benschubert | to master :) | 11:21 |
| tlater[m] | stable is a lot quicker than master, yeah | 11:21 |
| tlater[m] | Oh, we're about 5 seconds faster on startup time | 11:21 |
| benschubert | I mean master vs your branch | 11:21 |
| * tlater[m] thinks the changes don't affect build time much | 11:21 | |
| benschubert | Then I'm not sure how to read the benchmarks | 11:22 |
| tlater[m] | Startup time (a) for master is 0.37 but 0.33 for track-cleanup | 11:22 |
| tlater[m] | It'd help figuring out what that means if I knew what (a) is, but the results don't really tell me :( | 11:23 |
| tlater[m] | But yeah, not a severe improvement to overall build time. | 11:23 |
| benschubert | wait, .05 s not 5s right? | 11:25 |
| tlater[m] | Oh, you're right | 11:25 |
| * tlater[m] isn't sure how his mind went to 5s there | 11:26 | |
| benschubert | and the big show is slower | 11:27 |
| tlater[m] | I made no changes after that benchmark, so those numbers are what we have. | 11:27 |
| tlater[m] | I'm not sure about the last one, stddev 3x what it is on the previous run | 11:27 |
| tlater[m] | So I'd suspect something environment related. We don't do a warmup run there, so the data is pretty noisy | 11:28 |
| tlater[m] | No, not 3x | 11:29 |
| *** lachlan has joined #buildstream | 11:29 | |
| * tlater[m] clearly hasn't properly woken up yet | 11:29 | |
| tlater[m] | But it's certainly within stddev | 11:29 |
| * tlater[m] could benchmark locally and hope for slightly more accurate results | 11:30 | |
| benschubert | If you have time that might be better, I'm surprised that the initial thing would be quicker but not that :) | 11:32 |
| tlater[m] | The initial benchmark wasn't quicker either | 11:33 |
| tlater[m] | That one was, in fact, slower, because I accidentally regressed a previous optimization | 11:34 |
| tlater[m] | But it was within measurement error, as well... | 11:34 |
| * tlater[m] wonders if our benchmarks are just too generic | 11:34 | |
| benschubert | possibly | 11:37 |
| benschubert | ok, it cleans up the code anyways, let's go for it, if we can optimize let's do after? | 11:37 |
| tlater[m] | That makes sense, yeah | 11:37 |
| tlater[m] | traveltissues won't respond today, but we discussed that comment and I think I can tick that one off too, so yeah, let's merge! | 11:38 |
| tlater[m] | benschubert: Mind approving on gitlab so I don't look too narcisstic? ;) | 11:39 |
| benschubert | sure | 11:40 |
| gitlab-br-bot | BenjaminSchubert approved MR !1718 (tlater/track-cleanup->master: Simplify the codebase using post-tracking removal simplicity of the pipeline) on buildstream https://gitlab.com/BuildStream/buildstream/merge_requests/1718 | 11:40 |
| *** lachlan has quit IRC | 11:56 | |
| *** lachlan has joined #buildstream | 12:09 | |
| *** bochecha_ has joined #buildstream | 12:09 | |
| *** bochecha has quit IRC | 12:12 | |
| *** bochecha_ is now known as bochecha | 12:12 | |
| jjardon | Can I have review of https://gitlab.com/BuildStream/buildstream/merge_requests/1727/diffs please? | 12:37 |
| gitlab-br-bot | jjardon opened (was WIP) MR !1728 (jjardon/fedora-31->master: Use fedora31 instead older versions) on buildstream https://gitlab.com/BuildStream/buildstream/merge_requests/1728 | 12:56 |
| jjardon | also ^ | 12:58 |
| *** lachlan has quit IRC | 13:01 | |
| gitlab-br-bot | marge-bot123 closed issue #1199 (Follow-up from "skip tracking elements without trackable sources") on buildstream https://gitlab.com/BuildStream/buildstream/issues/1199 | 13:03 |
| gitlab-br-bot | marge-bot123 closed issue #1194 (Remove `Element.__tracking_scheduled`) on buildstream https://gitlab.com/BuildStream/buildstream/issues/1194 | 13:03 |
| gitlab-br-bot | marge-bot123 closed issue #1195 (Resolve FIXME from "skip tracking elements without trackable sources") on buildstream https://gitlab.com/BuildStream/buildstream/issues/1195 | 13:03 |
| gitlab-br-bot | marge-bot123 closed issue #1193 (Split `Stream._load` for target types) on buildstream https://gitlab.com/BuildStream/buildstream/issues/1193 | 13:03 |
| gitlab-br-bot | marge-bot123 merged MR !1718 (tlater/track-cleanup->master: Simplify the codebase using post-tracking removal simplicity of the pipeline) on buildstream https://gitlab.com/BuildStream/buildstream/merge_requests/1718 | 13:03 |
| *** traveltissues has joined #buildstream | 13:41 | |
| benschubert | jjardon: regarding !1727, that would mean dropping our last tests for python 3.5. I personnally would be happy to stop supporting 3.5, but do we still need it? If yes, then we should add some python3.5 as part of the same MR :) (Maybe named explictely 3.5?) | 13:51 |
| gitlab-br-bot | MR !1727: .gitlab-ci.yml: Use Debian10 instead Debian9 https://gitlab.com/BuildStream/buildstream/merge_requests/1727 | 13:51 |
| jjardon | benschubert: happy to keep Debian9 and add Debian10 if you want | 13:53 |
| jjardon | but remember this is for the current master, stable branch will keep testing with debian 9 | 13:53 |
| *** phoenix has joined #buildstream | 13:58 | |
| gitlab-br-bot | jjardon opened (was WIP) MR !1728 (jjardon/fedora-31->master: Use fedora31 instead older versions) on buildstream https://gitlab.com/BuildStream/buildstream/merge_requests/1728 | 14:04 |
| benschubert | jjardon: right, but do we expect bst2 to be distributed on debian 9? or want it? | 14:04 |
| jjardon | benschubert: no, it will not happen | 14:05 |
| jjardon | not even debian 10, actually, as it is stable now | 14:06 |
| benschubert | Oh ok, then let's drop python 3.5 on master :) | 14:06 |
| benschubert | Does it need a ML message? | 14:06 |
| jjardon | mmm, we still have centos rigth? what version of python centos uses? | 14:07 |
| benschubert | 3.6 | 14:08 |
| jjardon | benschubert: I can improve docs etc with the MR to clarify we depend on python 3.6 from now on, to make it more explicit? | 14:13 |
| jjardon | mmm, why did we remove the installation instructions from master? the installation instructions differ quite a lot between versions .... | 14:22 |
| * jjardon ponders adding them back | 14:23 | |
| *** phildawson has quit IRC | 14:28 | |
| *** phildawson has joined #buildstream | 14:30 | |
| juergbi | jjardon: what is missing on the website? buildbox install instructions for master are linked | 14:50 |
| juergbi | iirc, the idea was to have separate install instructions for the different branches on the website, but not readd them to the repo | 14:51 |
| *** phoenix has quit IRC | 14:53 | |
| juergbi | having install instructions both in the repo and on the website is not ideal, imo, as one of the two will likely get outdated | 14:53 |
| juergbi | unless they are automatically synced | 14:53 |
| jjardon | juergbi: absolutely agree, but why not leave them in the repo where natureally is going to be better maintained as they are close to the code? | 14:55 |
| jjardon | we even have this separation here now: https://docs.buildstream.build/ | 14:57 |
| juergbi | right, maybe we can include it in the already versioned documentation | 14:58 |
| juergbi | but we should make sure it's easy to find from the main web page | 14:58 |
| jjardon | actually, we have installation instructions in 3 different locations? https://docs.buildstream.build/ , https://buildstream.build/install.html and https://buildstream.gitlab.io/buildstream/ XD | 14:59 |
| jjardon | I'd suggest make https://buildstream.build/install.html point to https://docs.buildstream.build/ are we are done | 15:00 |
| juergbi | ah, in 1.4 it's already in docs | 15:00 |
| juergbi | but not in master | 15:00 |
| jjardon | yeah, it got removed, for some reason | 15:01 |
| juergbi | https://gitlab.com/BuildStream/buildstream/merge_requests/872 | 15:01 |
| juergbi | some discussion here https://gitlab.com/BuildStream/buildstream/merge_requests/835 | 15:01 |
| *** lachlan has joined #buildstream | 15:03 | |
| jjardon | well, maybe the intentions were good but the result is way more confusing IMHO, and incorrect as well. buildstream dependencies in the stable branche are very different from master | 15:04 |
| *** lachlan has quit IRC | 15:09 | |
| gitlab-br-bot | tlater approved MR !1728 (jjardon/fedora-31->master: Use fedora31 instead older versions) on buildstream https://gitlab.com/BuildStream/buildstream/merge_requests/1728 | 15:12 |
| *** phildawson has quit IRC | 15:19 | |
| *** phildawson has joined #buildstream | 15:22 | |
| *** swick has quit IRC | 15:24 | |
| gitlab-br-bot | marge-bot123 merged MR !1728 (jjardon/fedora-31->master: Use fedora31 instead older versions) on buildstream https://gitlab.com/BuildStream/buildstream/merge_requests/1728 | 15:24 |
| jjardon | tlater[m]: are you going to do that mailing announcement or am I suppose to do it? | 15:25 |
| *** swick has joined #buildstream | 15:25 | |
| tlater[m] | jjardon: You're proposing the change, though I can do it if you like | 15:26 |
| jjardon | I proposing update the docker images, not sure I can announce we do not support python 3.5 anymore :) but I do not mind doing it | 15:27 |
| tlater[m] | jjardon: Hrm, debian 9 doesn't look to be EOL yet | 15:28 |
| tlater[m] | jjardon: In that case, I think I'd rather you keep the debian 9 jobs and we'll think about deprecating debian 9/python3.5 when that rolls around | 15:29 |
| jjardon | tlater[m]: but why? bst2 will never be packaged in debian 9, not even debian 10 | 15:29 |
| tlater[m] | But then you know more about debian release cycles than I do ;) | 15:30 |
| tlater[m] | It will be installed there | 15:30 |
| * tlater[m] knows several people who install BuildStream despite its lack of distro support | 15:30 | |
| tlater[m] | If we drop support for the platform now we'll have a couple of months of confused people attempting to use master but failing | 15:31 |
| * tlater[m] doesn't like that | 15:31 | |
| *** lachlan has joined #buildstream | 15:33 | |
| jjardon | Only people using debian old-stable though . But sure I will keep it if you think is going to cause a lot of disruption | 15:35 |
| jjardon | thanks for the reviews! | 15:39 |
| tlater[m] | jjardon: Marge didn't like merging both your branches ;p | 15:41 |
| tlater[m] | Ta for the patches :) | 15:41 |
| jjardon | IS the diagram at https://docs.buildstream.build/master/arch_remote_execution.html up-to-date now that casd is a depency? | 15:46 |
| *** lachlan has quit IRC | 15:47 | |
| *** lachlan has joined #buildstream | 15:48 | |
| tlater[m] | jjardon: pretty much. The "Local artifact cache" is now just a buildbox-casd instance. | 15:55 |
| tlater[m] | Oh, actually... | 15:55 |
| jjardon | I see | 15:55 |
| jjardon | but is not buildboc who talk with the remote? | 15:56 |
| jjardon | buildbox* | 15:56 |
| tlater[m] | Hm, no, still correct. | 15:56 |
| tlater[m] | I don't think so | 15:56 |
| tlater[m] | But juergbi would know better | 15:56 |
| juergbi | if we consider buildbox-casd to be 'Local artifact cache' it looks correct | 15:57 |
| juergbi | that should probably be made explicit, though | 15:58 |
| tlater[m] | ta for the work on CASDProcessManager aevri | 16:05 |
| tlater[m] | Shame marge is being useless | 16:05 |
| *** lachlan has quit IRC | 16:09 | |
| *** cs-shadow has quit IRC | 16:12 | |
| *** bochecha has quit IRC | 16:12 | |
| *** bochecha has joined #buildstream | 16:16 | |
| *** lachlan has joined #buildstream | 16:25 | |
| *** lachlan has quit IRC | 16:35 | |
| *** rdale has joined #buildstream | 16:38 | |
| *** rdale has quit IRC | 16:41 | |
| gitlab-br-bot | marge-bot123 merged MR !1727 (jjardon/debian-10->master: .gitlab-ci.yml: Use Debian10 instead Debian9) on buildstream https://gitlab.com/BuildStream/buildstream/merge_requests/1727 | 16:51 |
| benschubert | ^ :( I was looking forward to no more python3.5 | 16:59 |
| tlater[m] | benschubert: Yeah, it's almost time for that though | 17:03 |
| tlater[m] | EOL happens ~2020, so another month or two :) | 17:04 |
| benschubert | Then we could drop it now | 17:04 |
| benschubert | master is not meant to be release before then | 17:04 |
| benschubert | https://www.irccloud.com/pastebin/7qR9gSAt/ | 17:04 |
| tlater[m] | That's true. I'd feel dirty doing so, though, given that we don't drop support for fedora-30 either | 17:05 |
| tlater[m] | Might be worth it for the async work. | 17:05 |
| *** cs-shadow has joined #buildstream | 17:09 | |
| benschubert | well, debian 9 will never update buildstream, I don't know for fedora :) | 17:10 |
| *** phildawson has quit IRC | 17:11 | |
| *** wberrier has joined #buildstream | 17:25 | |
| *** bochecha has quit IRC | 17:30 | |
| *** lachlan has joined #buildstream | 17:54 | |
| aevri | tlater: Welcome! Thanks for trying to set Marge back on track | 18:00 |
| *** swick has quit IRC | 18:22 | |
| *** lachlan has quit IRC | 18:33 | |
| gitlab-br-bot | marge-bot123 merged MR !1638 (aevri/casdprocessmanager->master: cascache: refactor, extract CASDProcessManager and CASDConnection) on buildstream https://gitlab.com/BuildStream/buildstream/merge_requests/1638 | 18:40 |
| *** swick has joined #buildstream | 18:47 | |
| *** lachlan has joined #buildstream | 19:07 | |
| *** lachlan has quit IRC | 19:11 | |
| *** bochecha has joined #buildstream | 21:03 | |
| *** bochecha has quit IRC | 21:42 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!