Re: [PATCH] Dejagnu testsuite for GNAT

[Roger Sayle <roger at eyesopen dot com>]

Hi Janis, Eric and Mark,

Re: http://gcc.gnu.org/ml/gcc-patches/2006-04/msg00455.html

On Wed, 12 Apr 2006, Eric Botcazou wrote:
> 2006-04-10   James A. Morrison  <phython@gcc.gnu.org>
>             Eric Botcazou  <ebotcazou@adacore.com>
>
>	PR ada/18692
> ada/
>       * Make-lang.in: Add check-gnat to lang_checks.
>       Rename existing check-gnat into check-acats.
>
>testsuite/
>       * lib/gnat.exp: New file.
>       * lib/gnat-dg.exp: Likewise.
>       * gnat.dg: New directory.
>       * gnat.dg/dg.exp: New driver.
>	* gnat.dg/specs: New directory.
>	* gnat.dg/specs/specs.exp: New driver.
>	* gnat.dg/style: New directory.
>	* gnat.dg/style/style.exp: New driver.
>
>	* libgloss.exp (find_gnatmake): Rewrite.


It looks like the efforts to get a non-ACATs testsuite for Ada have
stalled again.  It looks like Eric most recent revision addresses all
of Janis' earlier comments, including logic to test for a system
find_gnatmake should a future release of dejagnu provide this
functionality, and fall back to a GCC implementation for older, i.e.
current, deja-gnu releases.

I suspect that the innocent problem is that Janis is waiting to see
if the Ada folks have any objections, and the Ada folks are waiting
to see if the testsuite and dejangnu folks have any objections.  Both
side are probably waiting to see if the release manager is OK with
this (not actually a regression) infrastructure improvement.


One reason I ask, and am pinging Eric's latest patch, is that
we're beginning to accumulate a backlog of Ada/GCC patches and
fixes in the review queue all waiting upon the gnat testsuite.
For example, http://gcc.gnu.org/ml/gcc-patches/2006-06/msg00616.html
contains three middle-end patches all with Ada testcases.  Rather
than request the patch submitters attempt to produce an equivalent
C testcase where possible, my belief was that a non-ACATs testsuite
was/is "imminent".  Likewise, just today Eric has posted another
patch along with an potential Ada test case.
http://gcc.gnu.org/ml/gcc-patches/2006-06/msg00955.html


Hopefully, Mark will agree that this type of patch is suitable for
stage3.  Ada isn't currently built/tested by default, so it shouldn't
affect most people/platforms.  There may also be regression
fixes being blocked by lack of infrastructure, and indeed without
testsuite support its often impossible to confirm (i) what is a
regression or (ii) that it is resolved (on all platforms) by a fix.
For other front-ends, maintainers may often block/reject a regression
fix for lack of a new testcase for the testsuite.


I suspect that everyone is already in agreement, we just need
everyone to know that everyone else is happy.  Perhaps by a show
of hands, or standing and shouting "non placet" or "objection,
your honour".  A classic "New Mexican" stand-off:
http://en.wikipedia.org/wiki/Mexican_standoff  :-)


Ping?   Pretty please.

Thanks in advance,

Roger
--