problem(s) with "make test" and Sage-4.3.1
Georg S. Weber
I just noticed the following. In the Sage-4.3.1 source tarball, there
are both the packages "lapack-20071123.p0.spkg" as well as
"lapack-20071123.p1.spkg". If you build Sage from source, no problem
(yet), under /spkg/installed/ there will be only the newer "p1" spkg.
But if you "sage -bdist ..." a binary Sage distribution, then, under /
spkg/standard/, for both the "p0" and the "p1" spkg, a placeholder
file will be created (that shouldn't happen).
Worse, if I take this binary distribution, install Sage (elsewhere)
from that one, and issue "make test", then happily the "p0" lapack
spkg will not be found as installed (it isn't, "p1" is installed ---
but the "p0" shouldn't be looked for, which is bad), and then Sage
begins to download the "p0" one from the web and begins to build it
(deleting the files from the "p1" spkg first, according to the logs).
That should not happen at all!
I hardly got accustomed to the fact that "make test" in the above
situation will, at the beginning, output quite some crap, e.g. about
"circular dependencies dropped", and then rebuild just all of the
documentation files (taking a long time), for nothing --- these were
already provided. But trying to download older packages than the ones
provided, and installing these older ones over the newer ones is a bit
too ugly for my taste.
Did anyone experience the same problem(s)? We should create some
tickets, then.
Cheers,
Georg