Bug#870698: [Debian-science-sagemath] Possible fix for fplll/fpylll issue with default strategy path

[Ximin Luo]

Ximin Luo:
> Julien Puydt:
>> Hi,
>>
>> I just pushed to fplll's Debian git repository a tentative 5.1.0-3 which
>> would fix the recently reported issue with fpylll : I rewrote the patch
>> for the default strategies path changes.
>>
>> I'm a bit at loss on how to check if that really fixes anything : I
>> tried to build fpylll (success), but then trying to "import fpylll" in
>> python gave:
>> RuntimeError: You must get the file local/bin/sage-maxima.lisp
>>
>> so I can't really test seriously...
>>
> 
> I can reproduce this, it works if you uninstall sage (but of course we have to fix it properly).
> 
> The underlying issue is caused by the fact that Sage sets things like SAGE_LOCAL outside of python itself, so that one gets e.g.:
> 
> $ python -c 'from sage.env import DOT_SAGE, SAGE_SCRIPTS_DIR; print(SAGE_SCRIPTS_DIR)'
> $SAGE_LOCAL/bin
> $ sage -c 'from sage.env import DOT_SAGE, SAGE_SCRIPTS_DIR; print(SAGE_SCRIPTS_DIR)'
> /usr/share/sagemath/bin
> 
> I would guess that this also affects upstream Sage, but perhaps the issue is hidden somehow (Debian uses a different "sage-env" shell script).
> 

So for example this works, with sage installed:

$ SAGE_LOCAL=/usr SAGE_SCRIPTS_DIR=/usr/share/sagemath/bin python -c "import fpylll"
# no errors, exit code 0

Now the question is, does Sage always expect that itself should be run via the "sage" program, with particular envvars set? 

If yes, then fpylll should run the "sage" version of itself after detecting the presence of these envvars - a simple "try: import sage; except ImportError" is not enough.

If no (i.e. "import sage" etc should work in a plain python program), then sage should put its own "sage-env" shell script logic, into the "sage.env" python module instead, so that it gets run properly when things like "from sage.env import *" are done by third-party libs/programs. (And we in Debian would patch this logic for Debian paths.)

In both cases, this should be documented somewhere upstream in the Sage source code / documentation.

François & Jeroen (and any other Sage devs here), what do you think? Shall I file an upstream trac ticket about this?

(In any case, I think the fplll/fpylll bug is fixed, and we can upload + close #870698).

X

-- 
GPG: ed25519/56034877E1F87C35
GPG: rsa4096/1318EFAC5FBBDBCE
https://github.com/infinity0/pubkeys.git