Re: external dependencies

[Tristan Van Berkom <tvb gnome org>]

Elijah Newren wrote:
> Hi David,
[...]

   Sorry I didnt track this thread to the source and find out
what _exactly_ made people upset for what reason, but I do have strong
feelings about what I've read so far and will try to share them with you.

Somebody argued that if g-p-m can be compiled against version X of
libfoo.so, then it should not depend on version (X + 1); I wholeheartedly
agree that the tarball itself should only require the minimum requirement
because at that stage the software is in the hands of developers and
testers, its really not a problem in real life situations - distributions
should make sure to provide a "good" version of libfoo.so or when
packaging your module, adjusting the requirement at the package level
(not the tarball level)...

On the other hand, #ifdef littering of codebases IMNSHO dramatically
reduces the quality of those codebases (I will trade a functional codebase
that is a mess for a malfunctional code-base that makes sence any day of
the week - at least the latter can be improved whilst the former has
no hope).

So my question is - if all of that #ifdef litter can be dispensed with
in cvs HEAD then why wait ? remove all that cruft of the "current mess" and
work towards the future - once the #ifdef stuff is removed then it will
be impossible to compile against version X of libfoo.so - now that sounds
like a reasonable time to bump the requirement - no ?
its not like someone can argue that you shouldn't have bumped it then - since
the codebase was dramatically improved and now depends on a recent version
(hey guys, I refactored the code, its much better and now depends on libX.so v2.0,
so there ! end of story).

Furthermore, if people dont have the means to build/test cvs HEAD in gnome
then that's a problem, maybe a live development CD "a la knoppix" could solve
that kind of problem ?

Sure, compiling from cvs - even with innovations like jhbuild - is difficult
and cumbersome, but its currently the only way to get the tools we all need
to work on the platform of the future, we cant afford to deal with #ifdef
madness for the sake of people who dont have the means to compile cvs HEAD;
what if those people start contributing patches to #ifdef'd portions of
the code that are intended to be graveyard/obsolete code ? where are we going
with that ?

Cheers,
                     -Tristan