Squeak-dev February 2001

squeak-dev@lists.squeakfoundation.org

144 participants
725 discussions

Re: Shrinking an image
by Yoshiki Ohshima 26 Feb '01

26 Feb '01

Hello, > Thanks for the response. This is a Squeak2.7 image. I don't have a > CommandHistory class and I can't find the resetAllHistory method.... It is likely that there are many things in the trash can. Try to "empty" it. If it doesn't change the situation, try Smalltalk printSpaceAnalysis and view the resulting "STspace.text" file. If gives, well, the space analysis and sometimes helpful to figure out what is wrong. -- Yoshiki

1 0

Modular Squeak, Server Environments
by ajh18＠cornell.edu 26 Feb '01

26 Feb '01

Hello Squeakers, Below are some suggested enhancements for a modular Squeak universe. Make it so you can access published objects in other images (or super swikis) via variables in your own Smalltalk environment. Globals such as BobsSuperSwiki, UIUCArchive, etc, could be included in your image that would trigger a connection with a server image upon access. For example, to download and enter the FunWithMorphic project from BobsSuperSwiki you could execute "BobsSuperSwiki FunWithMorphic enter". Com, Org, Edu, Fr, Uk, etc, would also be globals in your image. Doing "Org Squeak" would access the image listening at squeak.org:8020 and return a proxy to its "Org Squeak" environment. The same environment path is used in the remote image, so originals and downloaded copies can live in the same environment, and variable references to it and from it can remain the same. When a published object is downloaded, the variable that refered to its remote proxy will now point to the dowloaded copy. To make this work, a global like "BobsSuperSwiki" would contain a special object called a ServerEnvironment which will first lookup a variable in its local dictionary and the local dictionaries of it parent environments, and if not found, connect to its remote server and ask it for the variable. When a variable is downloaded it gets added to the local dictionary. "Org" and "Org Squeak" would also be ServerEnvironments. When an object is uploaded/downloaded its entire export segment is copied. For adequate decoupling, the leaf objects of export segments are byteArrays, wordArrays, SmallIntegers, and/or DiskProxies. DiskProxies are just references to environmental objects (objects that are part of the user's system and should be referenced via the Smaltalk environment). Adding servers to the Smalltalk environment allows published objects to reference each other indirectly through environment variables. What if a downloaded segment references a Smalltalk variable (like AnthonysSuperSwiki) that does not exist in the client image? Environment references (DiskProxies) should contain a default value to install in case the environment variable does not exist. For AnthonysSuperSwiki, the default value would be a ServerEnvironment containing the network address of the AnthonysSuperSwiki server. Constraints can be added to environment references (DiskProxies) that make sure the referencing object is compatible with the environment value and converted if necessary. This is a generalization of what SmartRefStream does with classes and the instances that refer to them. Instead of variables like BobsSuperSwiki and AnthonysSuperSwiki being added to the top Smalltalk environment, they should be added to a sub-environment like Swikis. I like keeping the top level for general categories like Tools, Network, VMConstruction, etc., like when the image is partitioned into Environments by system category. These system environments could be ServerEnvironments that pointed to Squeak code servers. This way you can download only the classes you need, or want to override. What do you guys think? We need a consensus to make this work. Thanks, Anthony

1 0

Re: Shrinking an image
by Karl Ramberg 26 Feb '01

26 Feb '01

I ran into this too, not so long ago. I think you have infinite undo turned on. Check that in the preferences. Then zap the undo history: CommandHistory resetAllHistory Karl Larry Kellogg wrote: > > Hello, > One of my images has exploded to over 100meg and I was wondering if there > was any way to shrink it. Could somebody send me some good code snippets to > fix this problem? I think the base image was around 40-50 meg before I ran > my program a number of times. I don't know what is holding on to so much > storage... > > Regards, > > Larry

1 0

Re: Usenet client for Squeak
by Bob Arning 26 Feb '01

26 Feb '01

There are a few snippets on the class side of SimpleClientSocket. Cheers, BOb On Mon, 26 Feb 2001 13:04:44 -0500 "Stephan B. Wessels" <stephan.wessels(a)sdrc.com> wrote: >It does not exist in Squeak yet. However, you should be able to begin with the >foundation in Celeste since I think the messages are in the same format in NNTP. or >if not, they should be very similar. > > - Steve > >"Strugatsky, Alex" wrote: > >> Hi Squeakers, >> >> Does anyone work on Usenet client for Squeak or on NNTP protocol implementation?

1 0

RE: [Win32][VM] CreateThread() failures
by Raab, Andreas 26 Feb '01

26 Feb '01

Andreas, > The CreateThread() calls in the Win32 VM sources use a > dwStackSize of 0. According to the appended CreateThread() > documentation-fragments this seems to be pretty expensive. It's not too expensive (usually the Squeak stack is rather small). > This might explain the "CreateThread() failed (8) -- Not > enough storage is available to process this command." problem. Nope. It doesn't. I tried it and it didn't change a thing. However, reading the documentation more carefully points out that: > The number of threads a process can create is limited by the available > virtual memory. By default, every thread has one megabyte of > stack space. Therefore, you can create at most 2028 threads. "By default" reads as in "by default of your linker". If you do a quickview on Squeak.exe you'll find that the GCC linker doesn't give one megabyte as stack size. It gives "by default" 32MB (!!!) of stack size. Attempting to allocate this for each and every thread leads to a tiny bit of a problem as one might guess ;-) Unfortunately, there seems to be a bit more going on. When I reduced the stack size to approx. 128k I could open about 150 sockets at the same time (which accounts for roughly 300 threads). A lot better. But when I reached that limit, I got those weird messages that I've seen every now and then giving me error messages in the screen update (oops...) I'll have to look at this but thanks for pointing me to the information. Cheers, - Andreas

2 1

RE: [QUESTION?] Scope (visibility) of variables declared in block s?
by Mike Shields 26 Feb '01

26 Feb '01

> mumble > > self do: [ :i | ]. > i inspect > > results in > > mumble > > self do: [ :i | ]. > out of scope ->i inspect > But... Isn't the latter case "correct"? I can't imagine being able to use the first case shown.... It's certainly nonstandard, as far as other Smalltalk implementations are concerned, right? Maybe I'm missing something here?

1 0

[UNIX] patch page updated
by Lex Spoon 26 Feb '01

26 Feb '01

I updated my Unix patches page for 3.0pre2. Along the way, I updated my sound driver to support volume querying and control. http://www.cc.gatech.edu/~lex/squeak/ -Lex

1 0

[ENH][VM][UNIX]generalized async IO
by Lex Spoon 26 Feb '01

26 Feb '01

I've patched the Unix VM so that any part of the VM -- including plugins -- may now register file descriptors to be checked with select() whenever ioRelinquishProcessor() is called. I also updated the XWindows, sound, and sockets modules to use this general mechanism instead of their own private one. The immediate benefit is that sound events, too, will wake up ioRelinquishProcessor(). A smaller benefit is that several hacks such as soundPollFunction() and the extraFd argument to socketPollFunction() are removed. The longer-term benefit is that plugins don't need to modify the core VM as often. (To really do this right, there should also be a way to register a general polling function. I didn't get around to this, because it turned out not to be necessary for anything in the core VM.) The code for all this is at the bottom of: http://www.cc.gatech.edu/~lex/squeak/ It turned out to be pretty simple, since Ian already wrote the core mechanism for sqUnixSocket.c. Cheers, -Lex

3 3

Re: [QUESTION?] Scope (visibility) of variables declared in blocks?
by Bob Arning 26 Feb '01

26 Feb '01

Lex, On Sun, 25 Feb 2001 12:40:05 -0500 "Lex Spoon" <lex(a)cc.gatech.edu> wrote: >Yes, but the fix got undone. At least in 3.0, "[ :i | ] value: 3. i" >returns 3. The problem is that there are now *two* places temps are >looked up, and one of them has the code wrong. > >I sent out a patch a little while ago to re-instate the scope >checking.... and then forgot about it. (Okay, okay, so maybe a bugs and >fixes database would be a good thing after all). No idea on whether >it's in 3.1. Let's see.... The attached is untested, but it did the >trick once upon a time. It should at least give the right idea. The changes you originally sent and the one you attached to your latest email are in both 3.0 and 3.1. The "fix got undone" part was to allow out-of-scope references to be inspected. Otherwise the debugger lost some of its utility. In both 3.0 and 3.1, actually trying to compile a method mumble self do: [ :i | ]. i inspect results in mumble self do: [ :i | ]. out of scope ->i inspect Cheers, Bob

2 1

[ENH] Name the new HandMe PreferenceButton
by Lantz Rowland 26 Feb '01

26 Feb '01

Modifies the new 'hand me a button' message to set the external name for the Preference button, instead of the name 'Alignment'. outerButton setNameTo: aString , ' Button' . Lantz -- Lantz Rowland <lantz(a)aabysgallery.com> PgpKey: 0x67E5DFA5 Think Different! Think YottaHertz! zL - Lantz's Scriptible Web <http://Lantz.EditThisPage.com/> ----------------------------------------------------------------------

1 0

← Newer
1
...
4
5
6
7
8
9
10
...
73
Older →

Jump to page: