Brion Vibber <[EMAIL PROTECTED]> writes:

> autogen.sh checks for intltool < 0.28 or > 0.31 due to a known problem
> with corrupt XML output, but if you have a bad version the script just
> says "the problem is harmless, you may continue the build" and
> automatically moves on.
> It appears however to actually produce a corrupt gimp-tips.xml if you
> have intltool 0.30 (which is the current version available in the fink
> package set for Mac OS X). Those tips containing <tt> get horribly
> munged... On startup, Gimp complains on the console:
>    Error while parsing '/opt/gimp-2.1/share/gimp/2.0/tips/gimp-tips.xml':
>    Error on line 48 char 29: ')' is not a valid character following the
>    characters '</'; ')' may not begin an element name
> As a result, only the first 4 tips actually get shown on a build
> compiled this way; the rest are discarded when the error halts XML
> parsing. Rebuilding it with intltool 0.32.1 gets things working right.
> I'll see about submitting an update to the fink package, but perhaps
> autogen.sh should be a little pickier to make sure bogus builds don't
> make it out the door.

I don't think so. If we would have let autogen.sh fail, people
couldn't have been able to build GIMP from CVS until a fixed version
of intltool was released (which wasn't the case for about a year).

What I don't understand here is why does fink use autogen.sh at all?
fink should use a released tarball and the released GIMP tarballs have
been created with a working version of intltool and thus include
working versions of the intltool scripts. You are not supposed to run
autogen.sh unless you have a very good reason to do so. Does fink have
such a reason? If so, why haven't we been told about it yet?

Gimp-developer mailing list

Reply via email to