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/
squeak-dev@lists.squeakfoundation.org