[elk] Re: found bug related to GC crash

[martin rumori <lists@rumori.de>]

On Mon, Aug 09, 2004 at 11:40:32AM +0200, Alexander Lechner wrote:
> On Saturday 07 August 2004 11:50, martin rumori wrote:
> > fixed nor the unfixed version.  was that bug.scm for elk-3.0 or
> > elk-3.99.6?  AFAIK some GC bugs were fixed between 3.0 and -99.6.
> >
> Sorry, I've just re-verified this. I have elk 3.99.6, running on SuSE 8.2,
> kernel 2.4.20-64GB-SMP. 'bug.scm' still crashes with 

bad news.  same thing here.

72508
[Garbage collecting... 
Panic: Visit: object not in prev space at 0x406b5d08 ('pair') 5893 5895 (dumping core).

have had a closer look at this, i noticed that my fix does very
unlikely fix your crashtest.  the crash fixed by my patch just occurs
when doing something like (require 'record) or so before, i. e. when
new types are defined.  can't therefore be related to your
"empty-world" test (without even running toplevel).  should have
noticed that before.  i am just wondering why that never was dicovered
before, as with the stopncopy gc a "broken heart" warnin occurs
immediately after creating an object with the new type.  maybe on
other platforms ints were rounded rather than truncated... :-)

btw. i am now trying to crash elk with stopncopy -- that seems to be a
safe bet.  for the purpose of "foo", our non-realtime sound synthesis
system, fully sufficient.

o.k., let's see what future brings, more luck for the next time...

martin
-- 
This is the elk mailing-list, see http://sam.zoy.org/projects/elk/
Trouble unsubscribing? Please contact <sam@zoy.org>
List archives are at http://sam.zoy.org/projects/elk/threads.html