>>>> "SY" == Steve Youngs
<youngs(a)xemacs.org> writes:
SY> |--==> "OP" == Ovidiu Predescu <ovidiu(a)xemacs.org> writes:
OP> I've just checked-in the repository the 1.0 version of the XSLT-process
package
OP> for XEmacs. I've setup a makefile that builds only the elisp files, without
OP> trying to compile the Java classes.
SY> OK, great. I'll take a look and see if I can't build and release it.
OP> This minor mode depends on beanshell.el from JDE. However the JDE version is
OP> very old (2.1.9) and the functionality required from beanshell.el by
SY> Paul's been madly committing to CVS the last couple of days. He's
SY> updating to 2.2.7. The version of beanshell.el in CVS currently is
SY> 1.3. Is that the one you need.
OP> One issue I found while writing the makefile was that in the REQUIRES variable
OP> you not only have to put the package you depend on, but also all the packages
OP> that package depends on. Am I missing something? If this is the case I believe
OP> we have a problem, because changing one dependency in a base package requires
OP> updates on all the depending packages.
SY> I'm definitely open to suggestions for improvements.
To keep things sane, it makes sense to have an unstable branch of the
packages. Developers could check out the unstable branch using
cvs checkout -f -r unstable xemacs-packages
so that at any one time, only a few packages could have an unstable
branch.
This could also allow changes of questionable stability into the
packages without breaking users of stable xemacs. And commits would
not have to be delayed because of code freeze. When commits to
unstable have been deemed stable, they can be merged onto the trunk,
one package at a time.
The hard part of this is committer education. There are far more
xemacs packages committers than core committers, at least potentially.