From:	cgf AT bbc DOT com (Christopher Faylor)
Subject:	RE: Heap split - cannot fork:-)
25 Mar 1998 15:23:28 -0800 :
Message-ID:	<199803251520.KAA00964.cygnus.cygwin32.developers@hardy.bbc.com>
To:	cygwin32-developers AT cygnus DOT com

Ok. Then the only thing that I can see that would be in the way of the heap is
some thread's stack.  A simple test shows that the signal handling thread
stack is apparently "in the line of fire" of the expanding heap but it doesn't
seem to the specific cause of this behavior.

Another "solution" would be to relocate the heap into higher memory, like
around 0x7000000.  I just did this and it worked for me.  Of course, there
might be other stuff up in that region.  And, there it *will* hit the shared
section eventually.

I guess the correct solution is to copy a split heap correctly on fork.
It shouldn't be *that* hard.

cgf

>From: Sergey Okhapkin <sos AT prospect DOT com DOT ru>
>Date: Wed, 25 Mar 1998 16:58:14 +0300
>
>Christopher Faylor wrote:
>>I give up.  Is it because the shared_info shared memory segment gets in
>>the way?
>
>No :-) Default heap (without my patch) starts from ~0x24XXXXX, but
>shared section starts from 0xa000000 - the gap is big enough.

- Raw text -