| *** tristan has quit IRC | 02:46 | |
| *** tristan has joined #buildstream | 02:58 | |
| *** ChanServ sets mode: +o tristan | 02:58 | |
| tristan | nanonyme, yes, but there is no incentive to start worrying about it until we can get more serious about a 2.0 timeline | 03:00 |
|---|---|---|
| tristan | I think that the buildbox & friends part is currently "mostly stable", except for buildbox-common, so there might be a question of whether to distribute that dynamic or do static linking or whatnot | 03:01 |
| * tristan not 100% clear but that is the story afair | 03:01 | |
| tristan | juergbi, fwiw, the orphaned casd instances from one test run I ran last night are still lingering on today | 03:17 |
| tristan | I'll see if I can figure out which tests are causing this | 03:20 |
| * tristan is wondering if he can consolidate more code across the AssetCache derived classes... | 04:19 | |
| tristan | I wonder what is the justification for such wildly diverting APIs | 04:19 |
| tristan | SourceCache has commit(), how come ArtifactCache has no commit() ? | 04:21 |
| tristan | It looks to me that the only thing these subclasses do that is special, is the translation of data model objects to protos and back | 04:22 |
| tristan | and asides from this, the base AssetCache class can take care of loading/storing to local CAS and pushing/pulling from remotes | 04:22 |
| tristan | Also seems strange that ElementSources is taking care of putting the Source.get_kind() into it's own ElementSources.get_unique_key() implementation, rather than the Source._get_unique_key() wrapper | 04:26 |
| tristan | We would not otherwise need Source._get_source_name(), we could just use Source._get_unique_key() for the keys in SourceCache | 04:27 |
| *** tristan has quit IRC | 07:29 | |
| *** tristan has joined #buildstream | 07:31 | |
| *** ChanServ sets mode: +o tristan | 07:31 | |
| tristan | juergbi, I have a nice idea on how to abolish *both* of those obnoxious junction attributes which were used to say "download from junctioned project caches" and "upload artifacts to this project's server even if they come from junctioned projects" | 07:59 |
| tristan | Here, https://lists.apache.org/thread.html/rf2da9830e2fa918357f99a6021e55fc43df876f0b19d43f68802f083%40%3Cdev.buildstream.apache.org%3E I suggest removing "pull-subproject-artifacts" completely, since user config can make a blanket statement | 08:00 |
| tristan | But I think an elegant way to have full straight forward control, is to change the "push" attribute | 08:00 |
| tristan | So instead of a boolean "push", we could have an enum "action", and it's value can be "push"/"pull" or "both" | 08:01 |
| tristan | With this, user configuration is very straight forward, you can decide globally where to push and pull, and also override it on a per-project-name basis in the buildstream.conf | 08:02 |
| * tristan wonders if he forgot something | 08:02 | |
| tristan | my branch currently untangles all the mess of AssetCache poking it's nose in the parsing/decision making process, and results in removing about 250 loc | 08:03 |
| tristan | Hmmm indeed I think I missed something, I think there is a use case where we want to specify push remotes in the buildstream.conf without discarding fetch remotes recommended by project configuration | 08:04 |
| tristan | Maybe that is just one switch | 08:07 |
| tristan | "augment-or-override project recommended cache servers when building this project as a toplevel" | 08:07 |
| *** santi has joined #buildstream | 09:46 | |
| *** tristan has quit IRC | 11:25 | |
| *** tristan has joined #buildstream | 11:37 | |
| *** ChanServ sets mode: +o tristan | 11:37 | |
| *** tristan has quit IRC | 12:19 | |
| *** tristan has joined #buildstream | 12:19 | |
| *** ChanServ sets mode: +o tristan | 12:19 | |
| *** tristan has quit IRC | 12:41 | |
| *** tristan has joined #buildstream | 12:41 | |
| *** ChanServ sets mode: +o tristan | 12:41 | |
| *** tristan has quit IRC | 12:58 | |
| *** lchlan has quit IRC | 16:37 | |
| *** jward has quit IRC | 16:38 | |
| *** lchlan has joined #buildstream | 16:39 | |
| *** jward has joined #buildstream | 16:40 | |
| *** jward has joined #buildstream | 16:43 | |
| *** jward has quit IRC | 16:44 | |
| *** jward has joined #buildstream | 16:45 | |
| *** jward has quit IRC | 16:46 | |
| *** jward has joined #buildstream | 16:48 | |
| *** jward has quit IRC | 16:49 | |
| *** jward has joined #buildstream | 16:50 | |
| *** jward has quit IRC | 16:51 | |
| *** jward has joined #buildstream | 16:56 | |
| *** jward has quit IRC | 17:17 | |
| *** jward has joined #buildstream | 17:19 | |
| *** jward has quit IRC | 17:20 | |
| *** jward has joined #buildstream | 17:29 | |
| *** jward has quit IRC | 17:30 | |
| *** jward has joined #buildstream | 17:34 | |
| *** jward has quit IRC | 17:35 | |
| *** jward has joined #buildstream | 17:35 | |
| *** jward has joined #buildstream | 17:36 | |
| *** jward has quit IRC | 17:37 | |
| *** jward has joined #buildstream | 17:56 | |
| *** jward has quit IRC | 17:57 | |
| *** jward has joined #buildstream | 17:57 | |
| *** jward has quit IRC | 18:13 | |
| *** jward has joined #buildstream | 18:18 | |
| *** jward has quit IRC | 18:19 | |
| *** jward has joined #buildstream | 18:19 | |
| *** jward has quit IRC | 18:20 | |
| *** jward has joined #buildstream | 18:20 | |
| *** jward has quit IRC | 18:25 | |
| *** jward has joined #buildstream | 18:32 | |
| *** jward has quit IRC | 18:54 | |
| *** jward has joined #buildstream | 18:59 | |
| *** jward has quit IRC | 19:00 | |
| *** jward has joined #buildstream | 19:05 | |
| *** jward has quit IRC | 19:08 | |
| *** jward has joined #buildstream | 19:13 | |
| *** santi has quit IRC | 19:26 | |
| *** jward has quit IRC | 19:39 | |
| *** jward has joined #buildstream | 19:41 | |
| nanonyme | Hmm, I wonder if the running out of fd's with bst1 could also be related to refactoring in Python 3.8 | 19:41 |
| *** jward has quit IRC | 19:42 | |
| *** jward has joined #buildstream | 19:43 | |
| nanonyme | https://bugs.python.org/issue33695 | 19:44 |
| nanonyme | If Python ends up keeping fd's open longer than before, it might have resulted in further fd leakage | 19:44 |
| *** jward has quit IRC | 19:44 | |
| nanonyme | I wonder if there's any tools to debug fd leaks | 19:45 |
| *** jward has joined #buildstream | 19:46 | |
| *** jward has quit IRC | 19:47 | |
| *** jward has joined #buildstream | 19:48 | |
| *** jward has quit IRC | 19:49 | |
| *** jward has joined #buildstream | 19:51 | |
| *** jward has quit IRC | 19:52 | |
| *** jward has joined #buildstream | 19:53 | |
| *** jward has quit IRC | 19:54 | |
| *** jward has joined #buildstream | 19:54 | |
| *** jward has quit IRC | 19:55 | |
| *** jward has joined #buildstream | 19:55 | |
| *** jward has quit IRC | 19:56 | |
| *** jward has joined #buildstream | 19:56 | |
| *** jward has quit IRC | 19:57 | |
| *** jward has joined #buildstream | 19:57 | |
| *** jward has joined #buildstream | 19:58 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!