Again error "class-definition not found"

[Chris Muller]

This stack trace does indeed look like Magma is attempting to recover
from a true crash.  There are simple test cases cover this type of
recovery, but the fact yours was looking for class 146 when class 145
was your highest-known class suggests the crash occured at the moment
it was writing a commit involving a new class definition, which occurs
only the first time you commit an instance of a new type to the
repository.  The test cases do NOT cover this rare scenario.

For recovery, you should be able to simply delete the commitPackages
file in the db directory.  This is the last 5-seconds of commits it
was writing, so you will lose those, but the rest of the db should be
fine.

BUT, before you do that, there may well be a Magma limitation here
which I would like to study more.  Would it be possible to zip up your
db directory so I can look at it in the debugger?  It would really
make things a lot easier.  I promise I won't look at anything but the
recovery process and will delete it immediately after.

Thanks..
  Chris

PS - I've been working on a new Magma release and have completely
fixed the compress issue you found.  You have been really helping
lately plumbing the depths of Magma and finding some true issues,
thanks for that.


On 4/13/07, Hilaire Fernandes <hilaire2006 at laposte.net> wrote:
> After a Squeak crash I have this error which pop up when I restart a
> Magma session. This time the repository was local. See attached log.
>
> I did change some object interface before the crash (ie only method
> changes). Is this error message common during the developing phase?
> I don't know how to recover from it.
>
>
> Hilaire
>
> _______________________________________________
> Magma mailing list
> Magma at lists.squeakfoundation.org
> http://lists.squeakfoundation.org/mailman/listinfo/magma
>
>
>