22 Apr 2021 |
_hc | the build tool should be fixed to not include build paths, like Gradle Android Plugin and other tools. The production buildserver is hard reset on each build | 16:03:51 |
Fay (she/her) | I might be able to get clang/gcc to not include build paths, but python for android ships compiled .pyc files which contain build paths. that's not something I can just fix. | 16:13:49 |
Fay (she/her) | as long as I build my app locally in /home/vagrant/build/$appid with ndk in ~/android-ndk/r22b etc. (like the build server) it produces identical APKs. but the CI doesn't use those paths. | 16:13:56 |
linsui | In reply to @obfusk:matrix.org as long as I build my app locally in /home/vagrant/build/$appid with ndk in ~/android-ndk/r22b etc. (like the build server) it produces identical APKs. but the CI doesn't use those paths. The CI **should** use the same path. If not, it should be fixed. | 16:19:23 |
_hc | 幸猫: at the very least, you should fix a bug report about it, if there isn't one already. Python should also be able to do RB | 16:24:42 |
Fay (she/her) | linsui: CI build has /builds/obfusk/fdroiddata/build/dev.obfusk.sokobang/... instead of /home/vagrant/build/dev.obfusk.sokobang/... and /home/vagrant/android-ndk/android-ndk-r22b/... instead of /home/vagrant/android-ndk/r22b/... . | 16:28:50 |
Fay (she/her) | _hc: Python does RB as long as the build path is identical. Though usrmerge is also fun since e.g. /bin/install (or /usr/bin/install ) shows up in some the .pyc files as well. | 16:31:20 |
Fay (she/her) | I'll see whether I can find out how to fix the CI. unless someone else wants to. | 16:32:44 |
Fay (she/her) | and whether there's a bug report for Python RB. | 16:33:12 |
| kempe joined the room. | 16:33:30 |
Fay (she/her) | but I think the build should be reproducible on the production server. and using the same paths for CI is probably a lot less work than fixing Python. | 16:34:41 |
Fay (she/her) | <bug report for Python RB> https://bugs.python.org/issue29708#msg391616 | 17:03:14 |
Fay (she/her) | <I'll see whether I can find out how to fix the CI> WIP MR: https://gitlab.com/fdroid/fdroiddata/-/merge_requests/8843 | 17:42:28 |
izzy | Sylvia: _hc you wantet to raise your thumbs up on admin issues 218 and 219 when I created them (promotion for fossdd and FestplattenSchnitzel to developer) – here's the requested ping :) | 19:29:40 |
Sylvia | Done | 19:30:18 |
izzy | thanks! | 19:30:36 |
izzy | As I just pinged you two developer-to-be: issuebot now also reports results from my library scanner. Please disregard those for now (unless they really report findings): something is not working correctly with the CI integration, I'm still working on it. Will let you know when it's ready. | 19:32:29 |
izzy | _hc: looks like the apktool on stretch is too old. With the latest (2.5.0) I cannot reproduce that crash: https://gitlab.com/fdroid/fdroiddata/-/jobs/1204648221#L535 | 21:04:41 |
izzy | Is it possible to install it from backports? | 21:05:17 |
izzy | (in the issuebot module) | 21:05:28 |
izzy | stretch has 2.2.1 (that's from 10/2016), buster has 2.3.4 (from 9/2018); current would be 2.5.0 from 12/2020. | 21:08:34 |
_hc | izzy: try directly downloading the apktool binaries from the author | 21:13:58 |
_hc | in your module | 21:14:08 |
_hc | seems easiet | 21:14:19 |
izzy | Thought so, already started implementing that. BUT: dependencies? | 21:16:10 |
izzy | Maybe doing both (apt + direct download, so apt takes care for dependencies)? | 21:17:08 |
_hc | you could try changing your apt line by removing apktool and adding libsmali-java/testing apktool/testing | 21:18:11 |
izzy | Which version would that be? | 21:19:25 |
izzy | sid or bullseye? | 21:20:02 |
_hc | testing==bullseye for the next 6 months | 21:20:30 |