Building sources twice

[Toshio Kuratomi]

We have a few packages that need to build themselves from their sources
twice.  For instance, vim builds three times (a minimal version for /bin/vi,
and two versions with more dependencies for /usr/bin/vim and
/usr/bin/gvim).  Working on the python3 Guidelines, it looks like we'll have
some more with packages that build both python2 and python3 modules from
source that undergoes an automated transformation as part of its build.

So there's approaches to doing this:

1) Copy the source tree and build in both places
2) Build once with the python2 interpreter, copy the results away, and then
build a second time with python3.

vim takes a C version of option #2 (build with one set of configure options,
copy the results; make clean; and build with another set of options) but I'm
not sure there's a reason for that.  It uses less disk space as we don't
have to duplicate the source tree.  However, in software that might pollute
its source tree when it builds (maybe substituting file paths directly into
a source file, for instance), this could break.

Are there other reasons than I see for doing #2?

-Toshio
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 198 bytes
Desc: not available
Url : http://lists.fedoraproject.org/pipermail/devel/attachments/20100118/57f2a10e/attachment.bin