21 Feb 2021 |
izzy | :shrug: Guess in this case we've got to live with it. There are more important issues to be solved. Thanks for looking it up, SylvieLorxu! | 22:40:43 |
Sylvia | Well, it would be worth considering: should we really force it to be the exact same as fdroid rewritemeta would output or should we just accept any valid YAML and maybe just run fdroid rewritemeta as a bot like the checkupdate bot to make it "prettier" in other commits? https://gitlab.com/fdroid/fdroiddata/-/blob/master/.gitlab-ci.yml#L43-50 | 22:41:23 |
| marzzzello_ changed their display name from marzzzello to marzzzello_. | 23:03:57 |
22 Feb 2021 |
| jochensp1 left the room. | 02:54:19 |
jochensp | Thomas: Medilog 2.0.1 was published :) | 07:08:05 |
Thomas | ...and...drum roll...it's working! | 07:10:00 |
Thomas | A big thank you to everyone who stepped out of his way to turn this around a lot lot faster than usual!👍️ 😅 | 07:11:27 |
jochensp | (actually no one stepped out of the way, it was part of the usual cycle now) | 07:12:24 |
jochensp | Thomas: maybe you want to add a nightly build to detect problems before the release, next time: https://f-droid.org/docs/Publishing_Nightly_Builds/ | 07:34:49 |
Thomas | Ah, I wasn't aware of this, I will definitely change my roll out strategy going forward, there are lessons to be learned. | 07:35:47 |
Thomas | Just posted this on GitLab: https://gitlab.com/toz12/medilog/-/issues/110#note_513980090 | 07:36:55 |
Sylvia | I wonder where mondstern icons are, that was built but I still don't see the update on F-Droid | 07:53:35 |
_hc | mondstern: Sylvia like jochensp said, Ciaran cannot do it all. He says he works hours a week on F-Droid. Running servers that securely build thousands of apps is a big chunk of work. There are constant edge cases breaking things and troubleshooting takes time. | 08:22:55 |
_hc | there is a complete ansible automation for setting up the whole stack, and it can be run in any VM that supports nested KVM. So it definitely will run on Fedora | 08:24:44 |
_hc | we also have VMs for core contribtors to work in, so there are many ways to have access to a full buildserver | 08:25:24 |
Sylvia | And we want to help but if nobody helps us to understand what is going wrong we can't do much. All I know is that the APKs are built according to wiki logs but they somehow aren't all included in the index. And I have no clue how that could happen and when I ask it is always "fixed now" but it happens again anyway | 08:26:02 |
_hc | there is no contributor who knows it all, so fixing core issues like with the buildserver means digging in, reading crazy code, and running test setups. | 08:27:56 |
_hc | I answer the questions I can | 08:28:06 |
_hc | Ciaran hasn't worked on the code in years, so he doesn't know it. He runs the core build servers, so only he knows the details there. | 08:28:34 |
Sylvia | So you also do not know why an app that has been built according to the wiki logs does not show up in the next index update? | 08:30:38 |
Sylvia | Or why an app gets built multiple times on different days according to wiki logs? | 08:31:15 |
_hc | right, I don't know, its a large automated system, it could be so many things | 08:31:42 |
_hc | I generally fix issues when I know the cause or at least document them in an issue | 08:31:58 |
Sylvia | So, how can we figure out what went wrong to cause this? Fixing issues without knowing the cause is extremely difficult if not impossible | 08:33:23 |
_hc | by running test setups of the buildserver | 08:33:39 |
_hc | by improving the logging in the existing setup, like some recent additions to the status JSON | 08:34:01 |
Sylvia | But how will we run a test setup that can reliably reproduce it if nobody knows the exact steps that Ciaran takes to sign and everything? | 08:34:40 |
_hc | also jochensp just worked through a method to post the unsigned APKs that should be deployed son | 08:34:43 |
_hc | Sylvia: we know the exact steps, they are all in fdroidserver | 08:35:06 |
jochensp | and we see them in the logs | 08:35:30 |