From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <eclipse-return-112-listarch-eclipse=sources.redhat.com@sources.redhat.com>
Received: (qmail 6582 invoked by alias); 24 Feb 2004 17:25:37 -0000
Mailing-List: contact eclipse-help@sources.redhat.com; run by ezmlm
Precedence: bulk
List-Subscribe: <mailto:eclipse-subscribe@sources.redhat.com>
List-Archive: <http://sources.redhat.com/ml/eclipse/>
List-Post: <mailto:eclipse@sources.redhat.com>
List-Help: <mailto:eclipse-help@sources.redhat.com>, <http://sources.redhat.com/lists.html#faqs>
Sender: eclipse-owner@sources.redhat.com
Received: (qmail 6572 invoked from network); 24 Feb 2004 17:25:36 -0000
To: Paul Nasrat <pauln@truemesh.com>
Cc: eclipse@sources.redhat.com
Subject: Re: Successfull build on rawhide
References: <20040222095428.GC21947@lichen.truemesh.com>
	<20040222121358.GD21947@lichen.truemesh.com>
	<20040222223733.GF21947@lichen.truemesh.com>
	<20040223114134.GC9822@raq465.uk2net.com>
From: Tom Tromey <tromey@redhat.com>
Reply-To: tromey@redhat.com
X-Attribution:  Tom
X-Zippy:  Yow!  Are we wet yet?
Date: Tue, 24 Feb 2004 17:25:00 -0000
In-Reply-To: <20040223114134.GC9822@raq465.uk2net.com>
Message-ID: <87y8qs5rhu.fsf@fleche.redhat.com>
User-Agent: Gnus/5.09 (Gnus v5.9.0) Emacs/21.3.50
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
X-SW-Source: 2004-q1/txt/msg00022.txt.bz2

>>>>> "Paul" == Paul Nasrat <pauln@truemesh.com> writes:

Paul> JarWriter  fails (see eclipse-jar.out) 

Paul> IIRC both these failures doenot terminate build.

In most places we've changed the eclipse build to actually fail on
error.  Occasionally we've missed one.  Hmm, maybe we made this
change after the snapshot you're using.

Paul>      [exec] org/eclipse/jdt/ui/jarpackager/JarWriter.java:112: error: might transfer control into subroutine

This is a bug in gcj's bytecode verifier.  For gcc-ssa we just
commented out the error message.  That's not really a patch we could
put in the upstream gcj, though.

It is safe to disable compilation of any .jar file, however.  Just
edit gcjBuild and tell it not to build this one.  The only
ramification is that execution of this code will be a bit slower.

Too bad you missed FOSDEM, we talked extensively about our plans for
gcj and how they will affect this stuff :-).  For the verifier my plan
is to put the libgcj verifier into gcj as well, as it is built on a
more hackable (and less buggy) foundation.

Tom