[Gpg4win-devel] Gpg4win build system failure (glib-genmarshal)

[Marcus Brinkmann]

At Wed, 4 Jul 2007 11:23:16 +0200,
Jan-Oliver Wagner wrote:
> 
> On Mittwoch, 4. Juli 2007, Jan-Oliver Wagner wrote:
> > On Dienstag, 3. Juli 2007, Marcus Brinkmann wrote:
> > > At Tue, 03 Jul 2007 20:23:42 +0200,
> > > 'Werner Koch' wrote:
> > > > 
> > > > On Tue,  3 Jul 2007 19:53, marcus.brinkmann at ruhr-uni-bochum.de said:
> > > > 
> > > > > To speed this up, we could consider to perform all
> > > > > unpack/patch/configure steps before all build/install steps.
> > > > 
> > > > That is not easy for the most of the software as there are a lot of
> > > > dependencies which are checked by the configure scripts.
> > > 
> > > D'oh, I complete forgot about that for a second.  My suggestion was
> > > totally bonkers, thanks for putting me straight.
> > > 
> > > Best we can do is to document the few extra packages required in the
> > > README.  As Jan is doing tests from a vanilla configuration (Debian
> > > with only build-essentials installed? or something similar), he is in
> > > the best position to collect such a list.
> > 
> > yes, I can do this. It is not 100% vanilla, so I might miss some,
> > but it will be a good start.

You might even miss some because of multiple dependencies on the same
package.  Ie, if A and B depends on X, and A is build first, we will
easily find out that A needs X, but that B needs X is shadowed by the
build of A.  Testing all possible configurations is just too much
work, but it is sufficient if we document what is required to build
the canonical binary installer, that means all packages grabbed by
download.sh.

> build process works now for me :-)
> 
> I've updated README and web site accordingly.

Thanks.

> BTW: required disk space increased from 200 to 500.

I don't know compared to what.  Compared to the very first versions,
we now build much more packages from source instead using binary
packages.

Thanks,
Marcus