Path: utzoo!attcan!uunet!cs.utexas.edu!tut.cis.ohio-state.edu!unmvax!deimos.cis.ksu.edu!rutgers!rochester!pt.cs.cmu.edu!vi.ri.cmu.edu!skh
From: skh@vi.ri.cmu.edu (Steve Handerson)
Newsgroups: comp.sys.apollo
Subject: Bug w/ malloc SR9.7.1 3500
Keywords: 3500 aegis malloc virtual paging size
Message-ID: <4980@pt.cs.cmu.edu>
Date: 12 May 89 18:35:30 GMT
Distribution: na
Organization: Carnegie-Mellon University, CS/RI
Lines: 36



I'm having trouble getting anything from Apollo on this,
maybe someone already knows the answer.
I also think that "response centers" should all be bulletin-board based,
anyway, rather than forcing every responder to learn the same things.

We have an application that builds linked lists of malloc'ed structures.
Lots of them.
On a 3500 running 9.7.1 (bsd4.2) with about 200 meg left free on a 330
 meg disk,we get up to about 96 meg and the program dies.
It dies inside sbrk inside malloc.
The messages is: 

unable to allocate due to lack of resources (process manager/r/w/storage manager)

The problem is not the unix resource limit, since that's 2 Gigabytes.

This is apparently some aegis limit inside the process manager.
It's also definitely not a disk limit, since we can run two of these 
pretty concurrently and they both bomb out around 96 meg.
For the same reason, it's not a problem with the "os paging file size"
on the disk, although I've tried setting it to its maximum (550 pages).

This could be a problem with software compatibility, since
we had a habit of using old release tapes for storage.

Mail me if possible, otherwise post.  
I'll summarize if anyone knows.

-- Steve
-- 
Steve Handerson              Carnegie-Mellon University Computer Science Dept.
ARPA: skh@spice.cs.cmu.edu   UUCP: ...!spice.cs.cmu.edu!skh (I think)
Without chemicals, life itself would be impossible.
  (Monsanto ad?)             4 lines?  I'll have to make it up in the body...
--