[Gnash] Re: gnash testing

[Alex Dupre]

Jens Petersen <address@hidden> writes:

> >   I like pkg-config too, and used it for a bunch of tests. Most of my 
> > first bug reports on Gnash were all from people without pkg-config, so I 
> > rewrote all the configure tests to not need it.

If they hadn't pkg-config, sdl-config, etc. it was their dependency problem.
Like they installed the XXX libray because the configure said so they could
install pkg-config, too.
 
> Hmmm, pkgconfig doesn't seem an unreasonable dependency to me, 
> considering that libxml2, gtk2, cairo, pango, glib2, and gtkglext all 
> support it.

I completely agree. pkg-config could also not be a required dependency, but if
exists it should be used.
 
> >   In fact, I even dropped all the compile or link tests too, and stuck 
> > to simple directory & file searching. The compile tests are also often 
> > too complex for true portability. I figured it would be better if the 
> > configure finished, so people can see what they need to install rather 
> > than just having configure abort because somebodies machine isn't setup 
> > quite right.
> 
> I dunno, personally I would rather "configure" failed than "make". :) 
>   IMHO not using pkgconfig will lead to unnecessary configure 
> maintenance work.

Personally I prefer a configure that doesn't fail if I have all the
dependencies, rather that one that fails because it doesn't use the correct way
to detect them, or that doesn't fail but silently disable half of the features I
selected because it didn't find the libs.

--
Alex Dupre