From: Godmar Back (gback@cs.utah.edu)
Date: Fri Dec 18 1998 - 10:35:48 EST
Forwarded message:
> From oliva@dcc.unicamp.br Fri Dec 18 08:28:51 1998
> Sender: oliva@araguaia.dcc.unicamp.br
> To: Godmar Back <gback@cs.utah.edu>
> Cc: kaffe-core@rufus.w3.org
> Subject: Re: [CVS] commit
> References: <199812181521.IAA01106@sal.cs.utah.edu>
> From: Alexandre Oliva <oliva@dcc.unicamp.br>
> Date: 18 Dec 1998 13:25:59 -0200
> In-Reply-To: Godmar Back's message of "Fri, 18 Dec 1998 08:21:29 -0700 (MST)"
> Message-ID: <or90g5qxo8.fsf@araguaia.dcc.unicamp.br>
> Lines: 22
> User-Agent: Gnus/5.070068 (Pterodactyl Gnus v0.68) XEmacs/20.4 (Emerald)
> Mime-Version: 1.0
>
> On Dec 18, 1998, Godmar Back <gback@cs.utah.edu> wrote:
>
> > Could you more be more specific about where calling MARK_OBJECT
> > introduces crashes?
>
> Sure. On sparc-solaris2.5, I'd get a SEGFAULT while
> markObjectDontChecking the interface list of class [java/lang/Thread.
> After I fixed it so as to avoid marking interface lists of arrays
> (because they're static data, not gc_alloced), I started to get
> problems in an array of methods; size was apparently miscomputed, so I
> fixed it to get the size of the class object of the first method. But
> then, I'd get the same problem with field, but field doesn't have a
> pointer to a class, so I gave up for a while (I've got a report due to
> 4 PM; I had only 12 hours left then, and all I had was the .bib I had
> prepared with 108 citations :-)
>
> But I'm almost finished with it.
>
So, are you saying what's currently checked in is broken?
Actually, one thing I just noticed and which I'm somewhat skeptical
about is that MARK_OBJECT changed the semantics. It's not just a
macro version of markObject, it's a macro version of
markObjectDontCheck. This is somewhat confusing.
Basically, using MARK_OBJECT instead of markObject will crash if
you attempt to mark static data (as you've observed for the interface
list of arrays.)
I agree that we want precise scanning; how about if we rename
MARK_OBJECT to MARK_OBJECT_PRECISE or some such?
- Godmar
This archive was generated by hypermail 2b29 : Sat Sep 23 2000 - 19:57:22 EDT