Re: [RFC] maxima: add a subpkg maxima-ecl needed for sagemath

[tornaria <tornaria@users.noreply.github.com>]

[-- Attachment #1: Type: text/plain, Size: 1617 bytes --]

New comment by tornaria on void-packages repository

https://github.com/void-linux/void-packages/pull/34273#issuecomment-981080349

Comment:
A few more comments:

a. this package is not marked nocross, but there's no way to build on cross since sbcl, clisp, emacs are nocross. OTOH emacs is not really required: I tried having emacs as a build option (disabled on cross) and using ecl to build maxima cross aarch64. However, this results in x86_64 binaries rather than aarch64 binaries. Ofc if we use target ecl it doesn't run and if I use host ecl it builds for host. We'd need a cross-ecl which we don't have, so we should flag the package as nocross, right?

b. there is a lline `vopt_conflict clisp sbcl`, but I don't think there's any conflict on those. In fact, I compiled maxima with the three backends (sbcl, clisp, ecl) and it builds and runs ok. You can run `maxima -l clisp` to use the clisp backend, etc. Note that when you compile with multiple backends sbcl is the default, that's why the default I set (sbcl + ecl) doesn't affect current users.

Note that sbcl is much more efficient by far. Running the testsuite with `echo 'run_testsuite();' | maxima` (and same with -l ecl, -l clisp) takes me:
 - sbcl: real	1m35.744s
 - ecl: real	8m54.797s
 - clisp: real	18m14.480s

This justifies the default choice of compiling with sbcl in the main package (maxima) and moving the ecl backend into a split subpkg which is used for sagemath (ofc it would be nice if sagemath was able to use maxima compiled with sbcl, but at this time it can't). I'm not sure what advantage clisp brings to the table.