| 22 Apr 2021 |
 _hc | [remote "upstream"]
url = https://gitlab.com/fdroid/fdroidserver.git
fetch = +refs/heads/*:refs/remotes/upstream/*
pushurl = git@gitlab.com:fdroid/fdroidserver.git
| 06:16:22 |
| _hc | I always do that, then I pull commits over unauthed HTTP over Tor, and only auth to gitlab to push | 06:16:54 |
| _hc | then only my pushes can be tracked, but they show up with my name/date anyway | 06:17:24 |
| _hc | 幸猫: some parts of the Android SDK still require Java8, I'm pretty sure that's your error. | 06:19:32 |
| _hc | I guess since the buildserver is using an old SDK tools.zip, even more so | 06:20:54 |
 @SylvieLorxu:matrix.org | _hc: FestplattenSchnitzel uniq whoelse https://meet.keanu.im/fdroid | 11:33:47 |
 cdesai | _hc: oh right I feel like you must have shown this to me before but I keep forgetting. thanks! | 11:39:20 |
 @freenode_jochensp:matrix.org | https://gitlab.com/fdroid/fdroiddata/-/issues/2371 | 11:55:03 |
| cdesai | https://android.googlesource.com/platform/build/+/refs/heads/master/target/product/security/ | 12:02:42 |
 @obfusk:matrix.org | _hc: I've been building my apps w/ openjdk 11 since the beginning. so I'm pretty sure the only thing that requires Java 8 is the old tools.zip used by the build server. I've modified my MR to install a newer version of tools.zip, which seems to confirm this (though it's not an actual solution at this point). | 12:12:21 |
| _hc | I think only the most recent gradle releases support java newer than 8 | 12:17:03 |
| _hc | also Gradle Android Plugin required Java 8 up until recently | 12:17:53 |
| _hc | Android Studio still ships with Java8 | 12:18:05 |
| @obfusk:matrix.org | my python-for-android apps don't use android studio. and the whole reason I'm installing openjdk 11 in the metadata yml is that if I don't, the build is not identical to my own, since java 8 generates different bytecode. | 12:27:44 |
| _hc | sure, I'm just saying FYI. we have to support Java8 for those reasons | 12:32:57 |
| @obfusk:matrix.org | ah, sure. I wasn't suggesting dropping java 8 support or anything. | 12:53:12 |
| @obfusk:matrix.org | I'm just hoping I don't have to switch my builds to a java version not supported by recent Debian versions just to get reproducible builds working. | 12:54:35 |
| @obfusk:matrix.org | so is there an acceptable way to use a newer version of tools.zip for *my apps* (that supports java 11)? | 13:19:40 |
| @obfusk:matrix.org | * so is there an acceptable way to use a newer version of tools.zip for *my apps* (that supports java 11)? | 13:20:15 |
| @freenode_jochensp:matrix.org | 幸猫: where do you want to use it? On the official build server? | 13:21:25 |
| _hc | 幸猫: sure, whatever you can do in a build entry is acceptible | 13:26:56 |
| _hc | maybe stretch-backports has java11? | 13:27:34 |
| @freenode_jochensp:matrix.org | there are a number of packages installing java11 already | 13:28:00 |
| @obfusk:matrix.org | _hc: my MR currently installs opendjk 11 from stretch backports and replaces android-sdk/tools/ with a newer version. that seems to work. (there are some other issues with the CI through). | 13:44:23 |
| @obfusk:matrix.org | jochensp: yes, I want the offical build server to build a reproducible APK. | 13:45:45 |
| @obfusk:matrix.org | I could also remove/ignore the installed android-sdk and just let buildozer figure it out, but jochensp suggested that would not be acceptable on the official build server IIUC. | 13:48:22 |
| @obfusk:matrix.org | (the whole reason I got that error is b/c I was installing openjdk 11 from stretch backports.) | 13:50:55 |
| @obfusk:matrix.org | the other issues with the CI for my MR are:
-
build paths are different than on the production build server so RB fails.
-
it's building a 32bit and a 64bit APK now and the lattee fails b/c of leftover state from the first build. I would assume the production build server doesn't have leftover state - please correct me if that's wrong - so that also seems to be a CI vs production difference.
| 15:44:33 |
| @obfusk:matrix.org | * the other issues with the CI for my MR are:
1) build paths are different than on the production build server so RB fails.
2) it's building a 32bit and a 64bit APK now and the latter fails b/c of leftover state from the first build. I would assume the production build server doesn't have leftover state - please correct me if that's wrong - so that also seems to be a CI vs production difference. | 15:44:58 |
|  @freenode_kempe:matrix.org left the room. | 15:51:13 |
