From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 13707 invoked by alias); 1 Jul 2002 16:19:57 -0000 Mailing-List: contact gcc-help@gcc.gnu.org; run by ezmlm Precedence: bulk List-Archive: List-Post: List-Help: Sender: gcc-owner@gcc.gnu.org Received: (qmail 13700 invoked from network); 1 Jul 2002 16:19:54 -0000 Received: from unknown (HELO gandalf.codesourcery.com) (66.60.148.227) by sources.redhat.com with SMTP; 1 Jul 2002 16:19:54 -0000 Received: from gandalf.codesourcery.com (localhost.localdomain [127.0.0.1]) by gandalf.codesourcery.com (8.11.6/8.11.6) with ESMTP id g61GErN30820; Mon, 1 Jul 2002 09:14:54 -0700 Date: Mon, 01 Jul 2002 09:19:00 -0000 From: Mark Mitchell To: Daniel Jacobowitz , Steven Bosscher cc: "gcc@gcc.gnu.org" Subject: Re: Test suite for new front end: DejaGNU or QMtest Message-ID: <46000000.1025540093@gandalf.codesourcery.com> In-Reply-To: <20020701011916.GA12193@nevyn.them.org> References: <20020701011916.GA12193@nevyn.them.org> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii; format=flowed Content-Transfer-Encoding: 7bit Content-Disposition: inline X-SW-Source: 2002-07/txt/msg00012.txt.bz2 --On Sunday, June 30, 2002 09:19:16 PM -0400 Daniel Jacobowitz wrote: > On Sun, Jun 30, 2002 at 11:54:08PM +0200, Steven Bosscher wrote: >> Hello, >> >> Now that G95 generates some code, we'll soon need to be able to run test >> cases. Eventually we want to integrate G95 in GCC, so I guess we should >> write test cases for the testing framework that GCC will use in the >> future. >> >> GCC uses DejaGNU now, but I read in some post that g++ will use QMtest >> in the (near?) future, and that other parts of GCC will use it, too. So >> should a new front-end test suite use DejaGNU or QMtest? > > As far as I know there is no intention to replace the DejaGNU testsuite > in the short- to medium-term - only long-term if QMtest proves viable. > So for the moment I recommend continuing with DejaGNU, since that > framework exists now, and since converting over seems to be a > relatively simple process. As the prime proponent of QMTest, I agree with most of what Daniel says. The state that we are in with QMTest is that most people seem moderately pleased with the tool -- modulo the fact that all the right test classes for supporting complicated cross environments are not in place. We don't know yet whether CodeSourcery (or someone else) will do the work required to implement that support. DejaGNU is certainly the GCC standard at this point. On the other hand, conversion is not entirely trivial; it takes work to write test classes that can understand the DejaGNU code. Things will be simpler if you modify the existing DejaGNU code as little as possible; the more your tests look like the C or C++ tests the simpler the conversion will be. (And, the more consistent the tests will be internally, even if we never convert!) So, I would suggest that you write the tests using the "dg" style of DejaGNU test, as found for the tests in the "g++.dg" subdirectory. -- Mark Mitchell mark@codesourcery.com CodeSourcery, LLC http://www.codesourcery.com