This is the mail archive of the
java@gcc.gnu.org
mailing list for the Java project.
Re: Miranda methods again (Was: compatibility with jikes ?)
- From: Tom Tromey <tromey at redhat dot com>
- To: Mark Wielaard <mark at klomp dot org>
- Cc: Michael Koch <konqueror at gmx dot de>, java at gcc dot gnu dot org, classpath at gnu dot org
- Date: 01 Mar 2003 10:24:51 -0700
- Subject: Re: Miranda methods again (Was: compatibility with jikes ?)
- References: <200302271335.37040.konqueror@gmx.de><87vfz51yj1.fsf@fleche.redhat.com><1046517060.19764.124.camel@elsschot>
- Reply-to: tromey at redhat dot com
>>>>> "Mark" == Mark Wielaard <mark at klomp dot org> writes:
Mark> I vote that we add the Miranda methods explicitly to the
Mark> abstract classes in Classpath that need them.
Don't we only need this to work around gcj bugs? Maybe this should
only happen in the gcj tree.
In either case, let's mark the methods so we know to remove them
later.
Mark> The reason is that "modern" compilers like jikes (when used with
Mark> -target 1.4) no longer generate them and buggy compilers like
Mark> gcj -C never generated them. But not having those methods for
Mark> the standard classes now makes the serialisation different
Mark> (depends on Declared Methods) and breaks "buggy" VMs (that is
Mark> almost all VMs since till very recently they all depended on the
Mark> compiler generating these Miranda methods).
I think we have to solve serialization problems differently -- by
setting the UID explicitly and having tests. That seems like the only
approach that will work with all compilers.
Tom