Re: Fwd: [Win32] just another VM update (3.1 alpha build 1) - Squeak-dev

28 Feb 2001

      Hi,
    This VM is up on the 3.1_alpha directory on the uiuc ftp
site.
cheers
bruce
Bruce O'Neel bruce_oneel@yahoo.com wrote:
...
--- "Raab, Andreas" Andreas.Raab@disney.com wrote:
...
From: "Raab, Andreas" Andreas.Raab@disney.com
To: "'squeak@cs.uiuc.edu'" squeak@cs.uiuc.edu
Subject: [Win32] just another VM update (3.1 alpha
build 1)
Date: Wed, 28 Feb 2001 01:42:10 -0800
Reply-to: squeak@cs.uiuc.edu
Folks,
There's good news and there's bad news. The bad news
is that you'll probably
have to download a new VM. The good news is that you
probably *want* to ;-)
Some serious improvements and fixes have been made
in this latest VM,
including:

dynamic growth of ObjectMemory:

Yeah! Finally. No more -memory:, no more space low
warnings unless you *are*
out of memory. Note that this is why I consider it a
3.1alpha1 VM; the
changes do work with older images but it's really a
3.1 feature. If you're
updated through the last series of changes that Dan
posted for 3.1 then you
should get this VM and do a "space left" - could be
interesting ;-)
[BTW, if you're interested in what is going on under
the hood, then go into
the VM preferences and turn on "show allocation
activity" - this will print
out when the OM is actually growing and shrinking]
IMPORTANT: If you're planning to use this VM with
3.0 images you should get
the change set 3723SizeHint-ar from the 3.1 update
stream. Otherwise it
might be that the project publishing mechanism tries
to allocate huge (and I
mean: HUUUUGGGGGEEEEE) amounts of memory. It'll work
without the CS but it's
a gigantic waste of resources.

fixes for the networking code:

The new VM limits the stack size to 128k and should
allow for at least 200
sockets to be open concurrently. I also fixed the
code so that it fails the
primitive cleanly when the threads cannot be
created. I'd be interested in
seeing how many sockets can be open on NT in
parallel (I've only tried on
'98). Here's how you can find out:
| index tmp s |
   Socket initializeNetworkIfFail:[^0].
   index _ 0.
   tmp _ WriteStream on: #().
   [s _ Socket new.
   s isValid] whileTrue:[
   	index _ index + 1.
   	tmp nextPut: s].
   tmp contents do:[:sock| sock destroy].
   ^index
[BTW, I'm seriously considering to switch once again
to a different
implementation of the socket prims, using only a
single thread and event
notification which should basically put up the limit
of socket creation to
virtual infinity]

fixes for the display code:

Some of you have seen those weird error messages
saying "SetDIBitsToDevice()
failed". While running the socket tests I found that
the error reported was
just nonsense (invalid argument where it should say
out of resources). This
has been fixed as well (cost you a factor of five in
speed if you need it
but still better than not seeing anything on screen
;-)

MSVC fixes:

Some places that were incorrectly relying on certain
MingW features have
been fixed. I've included the Makefile that Phiho
Hoang posted (thanks!) so
that an MSVC build should work out of the box again.
Please try.
Okay, that's it. The stuff is in the usual places:
http://isgwww.cs.uni-magdeburg.de/~raab/squeak/SqueakExe.zip
...
[VM only]
http://isgwww.cs.uni-magdeburg.de/~raab/squeak/Squeak.zip
...
[VM+image]
http://isgwww.cs.uni-magdeburg.de/~raab/squeak/src/Squeak3.1Alpha1Src.zip
...
[VM sources]
Enjoy,

Andreas

=====
Bruce O'Neel - beoneel@acm.org http://homepage.iprolink.ch/~bioneel
The corporate culture is concerned less with Occam's razor than his
aftershave lotion. -- R. L. Peskin peskin@caip.rutgers.edu

Do You Yahoo!?
Get email at your own domain with Yahoo! Mail. 
http://personal.mail.yahoo.com/