New comment by tornaria on void-packages repository

https://github.com/void-linux/void-packages/pull/34030#issuecomment-1004350340

Comment:
#### Rebased to master with 3 new commits:
 - Main change: it now uses  ~ 100 python packages from system using patch from https://trac.sagemath.org/ticket/29665
 - patch to fix a long time test in `matrix2.pyx` that was causing a timeout on i686 (see https://trac.sagemath.org/ticket/33091)
 - include sphinx update (#34789)

#### Doctest status:
 - occasional 1 doctest failed in `src/sage/schemes/toric/sheaf/klyachko.py` (https://trac.sagemath.org/ticket/32773)
 - occasional 1 doctest failed in `src/sage/schemes/cyclic_covers/cycliccover_finite_field.py` (https://trac.sagemath.org/ticket/30419)
 -  for i686: 2 or 3 doctests failed in `src/sage/interfaces/maxima.py`, I have to investigate since I had not seen this before.

The first two are not very common, but the last one is happening almost always (IIRC only since the switch to system site-packages, may be some package version incompatibility --- perhaps `ptyprocess` which I patched it to use the void package at version 0.7.0 but stock sage accepts only 0.5.1).

#### Sources:
 - The current template is still using the whole huge tarball for sage which includes all upstream sources, git, etc. For the next version I'll switch to the clean sources of sagemath + just the required packages. At this point we only need 30 out of 200 that are included in the huge tarball.

It would be really convenient if we could split gap to a separate package since that's by far the largest package remaining (~500M source compressed) and it's quite useful independent from sage (will need https://trac.sagemath.org/ticket/29644 and https://trac.sagemath.org/ticket/31761).