On 08-10-2009 09:07:45 +0200, Markus Duft wrote:
> since some time, libffi is removed from python. either of one things
> would have been cool:

it would've been cool if this wasn't actually, since 80% of arches got
broken with this "change" IMO, which was of course totally undocumented
and hidden away in zillions of whitespace changes and other pleasant
things like wrapping comments to 192 chars instead of 72 or 80.

> a) inserting the rm -fr _after_ patching, so that patches don't break.
>    i don't even ask you to check whether patches still apply ...

the rm -fr comes from gentoo-x86, and indeed I don't have time any more
to check, even though I better do for python, but I blindly relied on
gx86 that internal libffi isn't necessary... well, it at least revealed
that python's beautiful build system hardcodes paths and places all
over, so it's totally broken on Darwin/OSX at least for the moment
still.  Since there is "support" code for Solaris too, I expect it to be
broken too.

> b) revbumping, so that i at least see the problem immediately, not
>    at the _next_ revbump, which is prolly unrelated.

what do you think *I* feel, when I find out that if I recompile python
because I neec ucs2 that it is totally broken because libffi is
gone/misdetected/overridden by a "smart" setup.py script??!?!?

> at first i was really puzzeled as to why my older prefixes all break on
> python updates (as opposed to more current ones, which have the python
> from before the rm -fr was inserted).

same here, get the idea why python-3.1 is masked in prefix (as opposed
to gx86) now?  I'm *very* unhappy about what's happening to python,
its wrapper and the mentality around it, and I can assure you, you are
not the only one that nearly explodes of irritation...  I've stessed my
opinion strongly towards the maintainer, but I can't do more than that,
unfortunately.  There has been way to much broken already with this
experimental crap.


-- 
Fabian Groffen
Gentoo on a different level