| 16 Nov 2018 |
 _hc | plus, checkupdates is the scariest piece, security wise | 16:41:32 |
 Tovok7 | so you say, don't optimize, but rewrite? | 16:41:53 |
| _hc | so it can have a dedicated VM that gets thrown away between runs | 16:41:54 |
| _hc | neither, I say use checkupdates where it is the only way | 16:42:22 |
| _hc | and add a new, streamined Tags mode handler in fdroid build | 16:42:41 |
| _hc | that does not parse versionCode/versionName | 16:42:54 |
| _hc | i.e. put all trust in the tag | 16:43:05 |
| Tovok7 | so mixing adding new build entries and building them? | 16:44:06 |
| _hc | so checkupdates would ignore metadata files where the update mode is Tags and the last Builds: entry lacks versionName/versionCode | 16:44:10 |
 Marcus (old) | _hc: I was thinking about just replacing the checkupdates=tags mode we currently have | 16:49:16 |
| Marcus (old) | the new one would I think be fully backward compatible with the old one. | 16:49:29 |
| Marcus (old) | but at that point I wasn't thinking about mixing it with fdroid build | 16:50:02 |
| Marcus (old) | I guess I instictively prefer when build is just building whatever is in the metadata at that point. | 16:50:51 |
| Marcus (old) | instead of trying to build new things | 16:50:58 |
| Tovok7 | yep, because here be dragons ;) | 16:51:10 |
| Marcus (old) | but with s#387 it kind of gets muddied a bit anyway | 16:51:28 |
| Tovok7 | builds runs a long time and doesn't update the data repo between builds, right? | 16:51:29 |
![@freenode_[gibot]:matrix.org](./avatar/%5Bgibot%5D) [gibot] | [server] #387: allow the version of a build to be figured out at… - https://gitlab.com/fdroid/fdroidserver/issues/387 | 16:51:30 |
| Marcus (old) | In reply to @freenode_Tovok7:matrix.org
builds runs a long time and doesn't update the data repo between builds, right? at that point it really should | 16:51:52 |
| Tovok7 | even then you might run into concurrency issues, they just become less likely | 16:52:29 |
| _hc | there is the ideal design, then there is what we can do to improve things with the least effort... | 16:54:09 |
| _hc | I'm proposing a nice step, but yes, still leaves things imperfect | 16:54:21 |
| _hc | ideally each build would be added to the index once it is done, then the sign/publish/deploy would just take whatever is there when it runs | 16:55:15 |
| _hc | and builds would be running constantly | 16:55:21 |
| Tovok7 | all I am saying is: watch out for concurrency issues | 16:57:00 |
| Tovok7 | This gets easier when you have just one external (non-gitlab) writer that adds stuff to the repo | 16:58:30 |
|  etue joined the room. | 16:59:27 |
| etue left the room. | 17:00:28 |
| etue joined the room. | 17:00:29 |
| _hc | yeah I think checkupdates would still be the only thing generating update commits | 17:00:55 |
