Re: Why not gnat Ada in gcc?

[Geoff Keating <geoffk at geoffk dot org>]

Jeffrey A Law <law@cygnus.com> writes:

>   In message <20001102025351.6992434DAF@nile.gnat.com>you write:

>   > Our objective at Ada Core Technologies in working to get the GNAT sources
>   > integrated with the main gcc tree is precisely to make it possible for
>   > more development to be done in the open. I am not quite sure what Jeff 
>   > is arguing for here, are you arguing that you would prefer we NOT make
>   > this effort, if so, that seems a bit peculiar.
> I'm arguing that ACT should contribute work incrementally to the GNU Ada
> compiler just like everyone else, ie no massive merges from the ACT internal
> tree.  All patches are sent to gcc-patches along with justification for
> the changes and that in general ACT employees play by the same rules as
> everyone else involved in GCC development.
> 
> In short that GNU Ada development fall under the same guidelines as the 
> rest of the compiler with no company or individual in a privileged
> position in regards to the development process.

Yes, this is the important thing.

It's not really the business of people outside ACT to say how they
arrange their work.  If ACT wants to wait a year and then merge in
dozens of features in the same week, they can try to do that.  (I
think that this approach won't work, but hey, maybe the ACT people
know something I don't.)

However, I do believe that ACT should follow the same rules as other
developers.  In particular, there's a general policy:

   If you have several unrelated changes, you should check them in
   with separate cvs commit commands.

that each patch, as submitted to gcc-patches and as committed to the
tree, should not include multiple unrelated features or bug fixes.
This makes it possible for those reviewing patches, trying to isolate
a single bug fix, trying to isolate a single bug introduction,
regression testing, or trying to use the CVS history to do this
effectively.

-- 
- Geoffrey Keating <geoffk@geoffk.org>