23 Jan 2019 |
_hc | Bubu: what are you trying to fix? | 16:14:52 |
Bubu | The license shenanigans | 16:16:04 |
_hc | because you could also test against one of the ci buildservers | 16:15:37 |
_hc | they are a bit out of date, but would be fine for certain issues | 16:15:51 |
_hc | shouldn't including build-tools-28.0.3 in the base image fix that for now? | 16:16:33 |
_hc | that's in master | 16:16:43 |
Bubu | See s!610 for discussion | 16:17:06 |
[gibot] | [server] !610: Add the newly added Android license. - https://gitlab.com/fdroid/fdroidserver/merge_requests/610 | 16:17:03 |
_hc | otherwise, it is just a matter of including the right hash, unless the license handling has also changed | 16:17:04 |
Bubu | The issue has nothing to do with 28.0.3 | 16:17:34 |
Bubu | Apparently the license handling did indeed change | 16:17:47 |
Bubu | Including the hash doesn't fix it unfortunately | 16:18:04 |
_hc | I thought that app build trying to install build-tools 28.0.3 is what triggered it? | 16:18:35 |
_hc | that's what I saw | 16:18:39 |
Bubu | Makebuildserver can't currently install any packages as is | 16:19:45 |
_hc | ah, right, it uses `android update sdk`, I thought it unziped the tarballs itself | 16:21:13 |
_hc | I'd try to run this process in a simple gitlab-ci job | 16:21:43 |
_hc | it'll be a lot faster | 16:21:47 |
_hc | like unzip tools_*.zip | 16:22:55 |
_hc | then run the `android update sdk` steps | 16:23:08 |
_hc | see _fdroidserver_'s gitlab-ci.yml for examples | 16:24:09 |
_hc | I think the fedora job does that | 16:24:13 |
_hc | it could be that `android update sdk` is now borken, which would suck | 16:24:46 |
Bubu | Could we switch to sdkmanager? | 16:29:17 |
Bubu | That always looks a lot cleaner | 16:29:24 |
Bubu | I guess caching the downloads gets harder | 16:29:53 |
Bubu | Maybe we can have makebuildserver just maintain an sdkmanager repo or something | 16:30:23 |
cdesai | have you tried zfs with lz4 compression? | 16:42:29 |
cdesai | I use that on my desktop for android builds, and it doesn't affect the build times much, while taking up a lot less space | 16:42:56 |
cdesai | and the compression ratio is great, 1.5x on my source ssd and 2.73x on the output ssd | 16:44:17 |