On 15 Nov., 20:03, Bill Janssen <bill.jans...@gmail.com> wrote:
> I'm building a new spkg for PIL 1.1.7.
>
> I'd like to build PIL, then run its test suite, then if that passes,
> install it.  Should I just do all of that in spkg-install, or should I
> run the test suite in spkg-check?  If I install it before I test it,
> that means that broken code could overwrite working code, so I'm
> inclined to do everything in spkg-install, rather than using spkg-
> check.  Or is there some rollback mechanism if spkg-check fails?
>
> Advice?

Unfortunately,

the current Sage (distribution) build system has no rollback
mechanism.

I advise to create a new directory "test" somewhere, and use a throw-
away/test installation of Sage there specifically for fiddling with a
new spkg. I always have half a dozen full Sage installations around.
PIL is/was especially unsuited for integration in Sage. On the one
hand, it installs quite some files e.g. under $SAGE_LOCAL/bin/ which
no Sage user needs (or even knows of, I guess), thus "polluting" the
tree without having cleanup mechansims at hand. On the other hand it
tries greedily to link against libraries it finds "somewhere" (by some
hand-made custom logic implemented in Python ...) on the system it
gets installed. Which is bad for Sage binary distributions (bdists)
made out of such an installation, if on the target system where the
bdist will eventually be installed, those libraries do not exist
(libtiff comes to my mind). IIRC, we had to introduce some extra
SAGE_BDIST (or so) environment variable especially for the PIL case,
in order not to have to uninstall each and every time those libraries
from the systems, where official Sage bdists are regularly generated.
Good luck for your new spkg! (Is there a trac ticket for it yet?)

Cheers,
Georg

-- 
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

Reply via email to