| *** tristan has joined #buildstream | 06:36 | |
| *** ChanServ sets mode: +o tristan | 06:36 | |
| tristan | After updating the documentation, I think it makes sense to remove the `junction` portion of a `dependency` dictionary | 06:45 |
|---|---|---|
| juergbi | tristan: in principle, I agree. however, that would be a format break from 1.x that isn't strictly necessary, so I'm not sure whether it's worth it | 07:13 |
| juergbi | we could also deprecate but still accept it | 07:14 |
| juergbi | tristan: regarding duplicates of the top-level project, we could have something like `.` referring to the current project. or we could consider it a special case that if the duplicate list contains the project name of the current project, it's implicitly in the duplicate list | 07:28 |
| juergbi | (I expect 'internal' to be usable for e.g. bootstrapping self-junctions, however, at least for completeness 'duplicates' should indeed support duplicates of the top-level project) | 07:29 |
| tristan | Yeah internal is usable for some cases but cannot assume it always is I think | 08:15 |
| tristan | I was also considering "." | 08:16 |
| tristan | Right now, while rebasing overrides on top of full paths (!1956), I'm thinking this copy_subprojects() approach in tests/format/junctions.py is making things very convoluted | 08:17 |
| tristan | reuse of these projects across multiple tests fuzzes things up and it's hard to tell what's being tested, modifications to one test can have unintended side effects in other tests | 08:17 |
| tristan | juergbi, Any objections to my tearing it all to the ground and making separate recursive statically committed directories for different test cases ? | 08:18 |
| tristan | Similar to what I have in tests/format/link for instance | 08:18 |
| juergbi | no objections | 08:19 |
| juergbi | explicit test cases are definitely better | 08:19 |
| tristan | Also, !1956 is totally ready for review I think (it's the weekend, but if you feel like it :)) | 08:19 |
| tristan | Cool, I ran into a wall with full path tests conflicting with override tests so I'll just try to untangle the whole thing | 08:20 |
| tristan | Maybe I should rewind the branch though, I guess I should do the untangling after full path support and before the junction overrides commit (these changes are starting to stack pretty high) | 08:22 |
| *** tristan has quit IRC | 09:04 | |
| *** tristan has joined #buildstream | 13:25 | |
| *** ChanServ sets mode: +o tristan | 13:25 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!