Re: Ada policy (was: GCC 3.5 Status (2004-08-29))

[Laurent GUERBY <laurent at guerby dot net>]

On Mon, 2004-08-30 at 22:28, Richard Kenner wrote:
>     The scenario I want to avoid is that we first reach 100% ACATS pass on
>     the two targets (looks likely), then later a patch goes in that
>     introduces 20 ACATS regressions on those two targets and the patch is
>     not fixed or reverted following the usual rules for other components. 
> 
> Yes, but that's a *different* standard.  What you are talking about
> if whether the commit rules will require running ACATS to commit a
> patch.  

Sorry I wasn't clear on my intent. The commit rules don't require you to
test on all the platforms, but if someone report that you break one you
didn't test, it's still considered your fault and you should try to help
on the issue, and in some extreme cases the patch could be reverted (at
least that's my understanding).

This "platform" rule could also be applied to components, you don't
necessarily have to test Ada, but if someones points out
a breakage, you have to be helpful on the issue.

(BTW I didn't find on the web site a link to the 3.4 release criteria
page http://gcc.gnu.org/gcc-3.4/criteria.html, and didn't find one for
3.5 did I miss something?)

> The question raised was whether Ada should be part of the
> release criteria.
> 
> These are not the same.

May be but I must admit I fail to see the practical difference, if
you're not continuously looking in some way at regressions on one
platform or component for a full development cycle, it's unlikely that
you'll be able to reach any relase criteria involving this platform or
component at the end of the cycle (without potential delays).

The current rules differ on who should be "looking" at regressions
depending on component vs platform (the patch submitter or someone
else), but that's less important IMHO than having a consistent rule on
what to do when a regression is detected on a particular patch.

Laurent