| *** hasebastian has joined #buildstream | 05:06 | |
| *** tristan has quit IRC | 05:15 | |
| *** hasebastian has quit IRC | 05:28 | |
| *** tristan has joined #buildstream | 06:28 | |
| *** ChanServ sets mode: +o tristan | 06:28 | |
| tristan | So links work perfectly now, but overrides are another tricky thing | 06:43 |
|---|---|---|
| tristan | In order to preserve the "duplicated" relationship of a project and it's subproject (or subsubproject), when the duplicated subproject in question is overridden, we need to translate real junction paths into original project paths (pre-override) | 06:46 |
| tristan | juergbi, Do you think it makes sense to do this ? | 06:46 |
| tristan | If I override a subsubproject which is either internal or duplicated in the subproject I junction, should the internal/duplicated relationship be preserved ? or should I be forced to re-duplicate it or re-mark it as internal ? | 06:47 |
| juergbi | otoh I'd say internal should definitely be preserved | 06:48 |
| tristan | Even if I override it ? | 06:48 |
| tristan | Hmmm | 06:49 |
| juergbi | yes, imo 'internal' is mostly between a project and its junction | 06:49 |
| tristan | If I override it with a local junction, it also technically becomes a junction in my project | 06:49 |
| juergbi | it means that those subproject elements are not exposed | 06:49 |
| juergbi | in such a case 'internal' should be mostly irrelevant as you're essentially manually deduplicating | 06:51 |
| tristan | Well, I'm telling the subproject to use a more up to date version of it's internal project or smth, maybe | 06:52 |
| juergbi | yes. ideally, if that toplevel junction is also used directly and the toplevel project has another junction to the same subproject, we would still get a conflict | 06:52 |
| juergbi | however, this may be tricky | 06:52 |
| tristan | No I don't think so | 06:54 |
| tristan | When you override a subproject there is still only one project | 06:54 |
| tristan | Only the search paths to reach it are changed, reversing that is indeed tricky | 06:55 |
| juergbi | the override itself is fine, of course | 06:55 |
| tristan | because there are an infinite number of overrides which can be traversed to reach an internal subproject | 06:55 |
| tristan | it may mean a try/search approach for each possibly way to reach a real path, finding every possible referrer to a project | 06:56 |
| tristan | s/possibly/possible | 06:56 |
| tristan | Hmmm | 06:57 |
| juergbi | sorry, have to go | 07:01 |
| tristan | yeah | 07:01 |
| tristan | I'll see if I can wrangle up some kind of path translation easier than that | 07:01 |
| tristan | without it being expensive | 07:01 |
| *** tristan has quit IRC | 07:05 | |
| *** hasebastian has joined #buildstream | 07:09 | |
| *** coldtom has quit IRC | 07:41 | |
| *** hasebastian has quit IRC | 07:43 | |
| *** hasebastian has joined #buildstream | 09:00 | |
| *** hasebastian has quit IRC | 09:14 | |
| *** coldtom has joined #buildstream | 10:32 | |
| *** coldtom has quit IRC | 10:47 | |
| *** coldtom has joined #buildstream | 10:47 | |
| *** tristan has joined #buildstream | 12:26 | |
| *** ChanServ sets mode: +o tristan | 12:26 | |
| tristan | juergbi, I've backported some cleanup/refactor to the top of my branch and made a separate MR: https://gitlab.com/BuildStream/buildstream/-/merge_requests/1964 | 13:42 |
| tristan | No hurry, I'll ping you or Ben tomorrow | 13:43 |
| tristan | if you happen to feel like checking it out... I'm rebasing everything on that, because having it in the middle of my branch would have been madness to review | 13:43 |
| tristan | gah | 14:27 |
| tristan | damn internal tests poking around at the ever changing insides :-S | 14:27 |
| tristan | how many do we have left to refactor out of there I wonder... | 14:27 |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!