On Sat, 22 Aug 2009, Raphael Geissert wrote:
> Package: sagemath
> Version: 3.0.5dfsg-4
>
> Hi,
>
> The binary package depends on many packages that I can't even imagine it
> actually needs them. Examples: mercurial, scons, gfortran?
>
> I remember seeing it pull in many many more packages when I tried it some
> months ago, but I don't see them anymore.
I think it previously erroneously recommended r-recommended rather than
r-base.
Sage's architecture is built around taking tying together a lot of
different C libraries; the Sage source distribution ships with the source
of some 75 or so other packages so that it is possible to install it from
source. So you should expect it to have a vast number of dependencies,
most of which are integrated so that they cannot be removed.
That said, the dependency list is something I've been meaning to clean up.
More of those packages are needed for some Sage functionality than you
might think (putting them in Suggests: might be fine if it is the case
that the problem is clear to the user what they need when an error occurs;
I'd need to check this). But given how huge the package's dependency list
must be anyway, it's not clear there's a lot of gain to be had here.
For example, Sage lets you write inline fortran in your scripts, so
removing e.g. gfortan will cause its internal test suite to fail, as well
as removing that functionality. And scons is required if you want to
download any of Sage's optional packages (it has a native interface for
installing certain parts of Sage that are not yet popular or good enough
to be in mainline).
-Tim Abbott
--
To UNSUBSCRIBE, email to [email protected]
with a subject of "unsubscribe". Trouble? Contact [email protected]
