Maemo disabled for armhf, armv7
For whatever reason, maemo related packages tend to fail to build with our setup when compiled to armhf and armv7. They don't fail always, but especially gtk+2.0-maemo is failing very often (and this one is required by all other maemo packages).
With the old binary repository scripts, it tries to build packages up to a hundred times, and only aborts after that. When there are new merge requests, I start the package building PC and leave it running over night, so this was not that much of an issue.
It seems like the issue has gotten worse, I've just built gtk+2.0-maemo after the APKBUILD was refactored in !730 (closed), and it took 19 (!!!) times before it built successfully for armv7. This is not acceptable, especially with the new infrastructure where we will start a job on builds.sr.ht for every package build.
Just recently we decided not to drop maemo (yet) in #373 (closed) as @NotKit stepped up to maintain it. So I've thought of a solution that makes sure we don't block the roll out of the new binary package repo and I've decided to disable building for armhf and armv7 for now. If aarch64 becomes as problematic, I'd disable building for that arch as well.
Note that this is not the fault of maemo (leste) folks, it must be related to our way of cross compiling packages with parts of it running in qemu. Sometimes it just hangs forever, and pmbootstrap then finally kills all running processes (example, example). I have a feeling that the linker is the problem, maybe running the linker natively fixes this (#233).
In order to make bugs like this more obvious, I'd like to get CI in pmaports.git to compile changed packages for all architectures that the package supports (just like Alpine has it nowadays, but use pmbootstrap and cross compiling in our case). The idea is, that we will really only merge changes to master where we know it builds, and if it does not, it should be network or other unrelated errors in almost all cases.
In any case, the cross compile setup seems to be working fine for pretty much everything else, and it doesn't seem like anybody is using maemo on postmarketOS at all. So it should be fine to disable it for these arches for now and move on.
Keeping this as an issue that we can close once we re-enable it for these arches.