Re: 1.8p2 weirdness...

• Date: Thu, 21 Mar 1996 20:30:25 PST
• From:
  Jay Carlson <nop@nop.com>
• cc: Andrew Wilson <andrew@aaaaaaaa.demon.co.uk>, moo-cows@parc.xerox.com
• In-reply-to: Your message of "Thu, 21 Mar 96 12:56:57 PST." <96Mar21.135657pdt.19545@silk.parc.xerox.com>

> Andrew Wilson writes:
>>      http://www.cm.cf.ac.uk/User/Andrew.Wilson/JHCore.db-20Apr1995.db.new
>>  is a JHCore that grew from 4Mb to 12Mb process size (2Mb to 3Mb
>>  compressed), in the space of about 2 minutes, on its own.

> Here's what I was able to track down.  I ran this DB under 1.8.0p2
> with the force_input() patch applied.  It pegged my CPU and, indeed,
> kept on growing and growing.  After a while, I attached a debugger
> to it and found that it was in the following task: Here's the code
> it was running for #30:to_jtext (which, of course, might not be the
> current text of that verb, since this is a long-running BG
> task):

[elided]
> If you trace through this code, you'll see that, when `flat[1]' is "", this
> loop makes no progress in reducing the length of `flat', while it grows `new'
> every time around.  At the time I stopped the task, `new' was 99,835 elements
> and growing.

> This sufficiently explains the observed behavior that I won't be looking any
> further into this.

Note that this bug is not present in the 20Apr JHCore; it was
apparently introduced by Andy trying to understand the somewhat
inscrutable code.

Jay Carlson
nop@nop.com    nop@ccs.neu.edu    nop@kagoona.mitre.org

Flat text is just *never* what you want.   ---stephen p spackman

• Re: 1.8p2 weirdness...
• From: Pavel Curtis <pavel@parc.xerox.com>

• Prev: Re: 1.8p2 weirdness...
• Next: Re: 1.8p2 weirdness...
• Index: MOO-cows
• Thread: MOO-cows