Raymond Toy wrote:
> >>>>> "William" == William Brew <[EMAIL PROTECTED]> writes:
>
> William> I am running into a bug where ext::purify gets a segmentation violation
> William> depending on the amount
> William> of stuff that needs to be purified. In the real situation where this
> William> happens, I am saving a core image
> William> file. If you pass the :purify t arg, the running lisp dumps core. I
> William> have duplicated what
> William> I think is similar behavior in a simpler context -- I just cons up a
> William> bunch of stuff and call purify.
>
> William> I am using 19a (the 2003-12 release). The problem also occurs in 18e.
>
> Can you try the Apr 2004 snapshot? I ran your test on Solaris with
> the current CVS version, and purify works fine.
>
> Ray
Thanks for the quick reply.
I tried the experiment in the 2004_04 snapshot and purify works fine.
However, I tried building my application in the 2004_04 snapshot and there are some
serious problems.
- The compiler (or something related) seems to have slowed down dramatically.
The compile and load time for my application went from 47 minutes in the 2003_12
snapshot to 116 minutes in the 2004_04 snapshot.
- When I save a core file and try to restart, there are very varying results. The
most
common result is a div by zero error in something very very early (i.e.. before
the
herald is even printed). Trying to get a back trace fails. The behavior seems
to
be
related to the size of the
saved application -- I do partial loads of the application (with a save) and
depending
on where I draw the line, the new core either starts up fine or doesn't. There
seems
to be some indication that the problem happens when loading the site init file (I
traced a few of the functions that get run a startup and the div by 0, when it
happens, happens somewhere in the load of the site init file.)
Hope this helps.
-bill-
