Re: Ada cross-compiler (Windows) to Android

[jrmarino <dragonlace.cla@marino.st>]

On Jul 15, 12:06 pm, Ludovic Brenta <ludo...@ludovic-brenta.org>
wrote:
>
> Could you explain how your repository is organized?  I would think
> you'd cloned the upstream git repository[1] and created a branch of
> your own where you maintain a stack of patches using quilt[2] or
> StGit[3]?  This technique allows you to make your patches available to
> anyone who wants them, even before you submit them to the FSF.  (The
> Debian patches are similarly maintained in a monotone repository,
> using quilt.)
>
> [1]http://gcc.gnu.org/git/
> [2]http://savannah.nongnu.org/projects/quilt
> [3]http://git.or.cz/course/stgit.html
>
> --
> Ludovic Brenta.
> The enabler generates a wide-spectrum, agile, bottom line, while
> siloed tactics enable the enablers.

Sure, Ludovic.
I need to first make a distinction between gcc 4.6 series and 4.7
series.
For the 4.6 series, I basically dumped gcc/ada (and subdirs) and the
two Ada testsuites along with some key files (e.g. config.gcc),
maintaining the same hierarchy.  When it's time to build a tarball, I
overlay these files on top of a stock gcc.  (for c++ testsuite and a
few key files, I also apply a set of patches).

That technique basically forced me to use "meld" a lot and manually
maintain a list of files that differs.  It is painful to do.

So solve that, and to prepare proper patches for import into gcc, I
created a new branch for gcc 4.7 that only contains patches.  To
produce 4.7 tarball, I just apply the patches to a copy of the stock
gcc.  That way I can just send the patch to gcc and remove it from the
repository after it gets accepted.  That was the plan anyway, and the
idea is that over time the number of patches would decrease to a very
low number.

I took a quick glance at that quilt site which I had never heard of
before, but it wasn't immediately obvious to me how it is supposed to
work.