[Debian-science-sagemath] Final touches

[Tobias Hansen]

On 2/19/19 7:48 AM, Ximin Luo wrote:
> Tobias Hansen:
>> [..]
>>
>> It seems that everything is coming together. Another round of testing with the updated jupyter-notebook and matplotlib would be welcome.
>>
> One thing I noticed is that we hard-depend on a bunch of doc packages. The biggest one is singular-doc with an installed size of 264 MB.
>
> Is this really necessary? Ofc we must install them at build-time to run the tests, but I wonder if we can get away with doing Recommends: singular-doc for users. We don't even hard-depend on sagemath-doc-*, for example.

We definitely still hard-depend on too much stuff. We have been very conservative (or lazy) about this, see all the dependencies I removed in 8.6-4. There might still be dependencies that are not needed at all, for example a bunch of python packages (I don't know which ones) don't need a direct dependency. Do we need runtime dependencies on cython and sphinx? Other dependencies should be recommends to allow people to remove them if they want. Apart from the doc packages one could do this for example with the gap libraries, sagenb, r-cran-lattice etc. If people install sage with --no-install-recommends that would break certain functionality but then it's their own fault... It's hard to draw the line what breakage is acceptable for ignoring recommends.

Help on deciding what to change would be appreciated.

>> Two known issues that would be nice to fix are the known issues with the documentation with the sphinx 1.8 patch:
>> https://trac.sagemath.org/ticket/26451
>>
>> and the broken user management in sagenb (#888822):
>> https://github.com/sagemath/sagenb/issues/432
>>
> Didn't look at these yet but I just wanted to remind that the deadline for fixing these is actually March 02 since it now takes 10 days for the Testing migration process.

I'd like to point out that 8.6-5 didn't segfault anymore on mips64el for some reason, so now the only reason for the FTBFS on mipsel and mips64el are the timeouts in jmol.

Best,

Tobias