There is an updated pull request by q66 against master on the void-packages repository

https://github.com/void-power/void-packages apl
https://github.com/void-linux/void-packages/pull/12661

apl: patch build on platforms with not enough ldbl range
Not every system has 80-bit or better extended range/precision long doubles, particularly ppc(64) uses either 128-bit long doubles implemented as a double pair (which has extended precision but not range) or just plain 64-bit that is the same as double (with musl which does not implement the 128-bit precision). On those systems, this patch pretty much reverts to the old code (i.e. before 1.8 update) which was not using precomputed tables.

@leahneukirchen 

A patch file from https://github.com/void-linux/void-packages/pull/12661.patch is attached