Re: Ada policy

[Robert Dewar <dewar at gnat dot com>]

Joseph S. Myers wrote:
> On Mon, 30 Aug 2004, Robert Dewar wrote:
>
>
> > 1. We submit fixes for bugs for which we have only proprietary
> > test cases, meaning that overall the quality of the Ada front
> > end will be higher, since for one thing, it will be far closer
> > to our in house tree for GNAT Pro (not quite indentical, since
> > there are things in the GNAT Pro tree that cannot go into the
> > FSF version for various reasons, related to FSF/GCC requirements,
> > not our requirements).

> Bugs shown by proprietary testcases should indeed be fixed.  But each 
> individual patch submission that can't include a testcase should include a 
> statement of why (e.g. that the test is proprietary and a synthetic one 
> not readily producable), and such bug fixes need more detailed 
> explanations of what the problem was and why and how the patch addressed 
> it, if not obvious, than bug fixes with included testcases, in case the 
> patch causes problems and someone trying to address those problems needs 
> to understand why the patch was needed in the first place.

In practice that is probably more of an issue with back end fixes
than front end fixes. Nearly all the front end fixes are clear, and
in any case the patch always includes appropriate comments. I am not
saying that following the above protocol would not be desirable, just
that in practice, we don't have the bandwidth to do much more than
the synchronization we are doing now. I do agree that general gcc
patches require appropriate documentation etc.
>