On 28 Mrz., 17:08, leif <not.rea...@online.de> wrote: > On 28 Mrz., 12:31, "dahl.joac...@gmail.com" <dahl.joac...@gmail.com> > wrote: > > > matplotlib is only used for some of the examples with CVXOPT; there is no > > reason to introduce such a dependency. If it's a matter of automated > > tests, then it would be much better to comment out the > > plotting in the few examples. > > I'd also probably do /that/, although I don't have the dependency > graph at hand right now... > > Simply making CVXOPT depend on matplotlib is certainly easier, but > IMHO not really appropriate since the former *builds* without the > latter. > > Since we're meanwhile /generating/ Sage's Makefile, I'd simply add > dependencies depending on whether SAGE_CHECK=yes or not, which doesn't > require patching CVXOPT's tests.
FWIW, you don't even have to /generate/ the Makefile (or deps) for that purpose; just make CVXOPT depend on $(MPL_for_testsuite), and set that variable depending on the value of SAGE_CHECK. We already have odd (or confusing) dependencies like that of ATLAS on Sage's Python package; similarly, ATLAS should depend on $ (PYTHON_for_build_script) or the like, for now just to make the reason explicit. -leif > Anyway, upstream should make the test suite not fail in case no > matplotlib installation is available. (These tests should simply get > skipped then, perhaps along with a warning.) > > -leif -- To post to this group, send an email to sage-devel@googlegroups.com To unsubscribe from this group, send an email to sage-devel+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URL: http://www.sagemath.org