Squeak-dev June 1998

squeak-dev@lists.squeakfoundation.org

83 participants
130 discussions

14 new updates
by Dan Ingalls 15 Jun '98

15 Jun '98

Folks - We have just put 14 new updates out on the external servers. Here's a summary... 088sw_TabbedPalette.cs Puts a TabbedPalette in the standard parts bin. New TabbedPalette features: paste an entire book as a new tab, either: a. Before the currently selected tab, or... b. After all existing tabs remove an entire tab rename a tab recolor a tab add a new tab 089tk-ObjStore-FH.cs Update 089 fixes a bug I (Ted) introduced in Update 077. When I fixed the problem with writing long strings to the disk, I broke the ability to read any strings at all from older files. This is the same fix as before, but with the new string format under a new type ID. The old format can still be read. If you stored such a file between May 29th and today, here is how to bring it up to date: If you have the object in an image: Get Updates. Write the object out again. (throw away the old file) If you only have the file: Read the object into the image it came from (one that was Updated between May 29th and now). Get Updates. Write the object out again. (throw away the old file) 090jhmTweaksJune10.cs 091ProjectLabels-di.cs Introduces windowReqNewLabel:, and uses it to pass window label changes to the window's model for approval and possible further action. Used to pass project labels on to their changeSets in both MVC and Morphic. Introduces windowActiveOnFirstClick in Model, defaulting to false. This can be overridden by, eg, Projects to get first-click behavior. Organizes the methods in TranscriptStream. 092tk-CngSetList-FG.cs 093MutatingDoitFix-di.cs In the course of doits and fileIns, a class may be mutated. The mutation may invalidate methods that are in the process of being executed, if they are on the stack. This file redefines a couple of methods that are usually on the stack, so that most mutations, even of currently executing classes, will not cause a problem. 094mouseOverHalos.cs Allows a PasteUpMorph to have the property that a mouseover one of its submorphs should bring up a halo around that submorph automatically. 095replaceSubmorphFix.cs Fixes a bug reported by Juan Cires Martinez by factoring out part of Morph.replaceSubmorph:by: code to PasteUpMorph. Further, assures that the replacement morph and its submorphs get added to the step list if required 096tk-swiki-FI.cs 097jhmFileListFixesJun13.cs Fixes a FileList problem with file names contain parentheses characters. Fixes 'show image in a window' command in Morphic projects. Fixes a bug in SystemTracer reported by Tim Rowledge (thanks, Tim!!). 098jhmMinimalVM.cs Adds a C source file that demonstrates how little support code is needed to bootstrap a basic Squeak virtual machine. This may be useful as a guide to those porting Squeak to new platforms. 099HotList1.cs Fixes bug so Project windows again update their label with corres changeSet. Changes ChangeSorter changeSetList so it shows most recent first. Changes changes log menu so it shows most recent first. Adds menu item for opening a single change sorter. Makes windows able to be dragged on first click if in title bar. Reorganizes Project jumpTo so it presents a came-from list, the parent, then all projects. Makes Project enter menu come up centered on first item. Fixes problem with textured backgrounds in project link creation. 100ChangeSorterTweaks-di.cs A few tweaks to make the new ChangeSorter format work better... Update list after new, rename, remove. Present show... menu with most recent first, as in the list pane. 101PenFillHack-di.cs Defines Pen >> fill: fillBlock color: fillColor allowing a closed shape drawn by the fillBlock to be filled with the fillColor. For example, try... Display restoreAfter: [Pen new filberts: 4 side: 5]

2 1

Re: Morph Must Die!!! (a bug report)
by Chris Reuter 14 Jun '98

14 Jun '98

I Wrote: > I seem to have discovered a bug in Morphic. It seems that deleted > Morphs continue to hang around and bloat the image. I was able to > reproduce it in the following manner: [deletia] OK, it turns out I was wrong about the bug affecting all Morphs. Rigorous use of Smalltalk>>garbageCollect showed that only instances of the particular Morphic subclass I'm working on is affected. It turns out that this Morph accidentally creates a dependency that doesn't get broken when the Morph is closed (i.e. sent #delete). The stale Association in DependentsFields keeps the Morph alive. This Morph is a subclass of SystemWindow, created in the time-honoured tradition of creating a subclass and making it create the necessary submorphs. This one creates two submorphs--a custom widget and a PluggableTextMorph with a TranscriptStream as model. The parent window has no model. The dependency between the TranscriptStream and the PluggableTextMorph is what goes in DependentsFields--closing the window doesn't break it, so a reference to the PluggableTextMorph (and thus, the whole window) stays alive. The main window itself doesn't have a model. Is this a Squeak bug or did I do something wrong? Up until now, subclassing a Morph and adding submorphs didn't require any other code to work, which makes this behaviour incongruent with the rest of Morphic. It appears that there is a lot to Morphic that I still don't understand. --Chris

1 0

Re: squeak hogs memory on linux?
by Greg Lewin 13 Jun '98

13 Jun '98

> Service pack #3 >is the latest (and it's now over a year old <grumble>) and we've been >running >quite stable here for more or less ever. I find that Squeak is very >stable under >NT4 with service pack 3. HTH > Yes, I'm using #3 - NT Squeak is OK with me, it's the linux version that sprawls. I hear SP #4 is on the way RSN. -- Greg Lewin

1 0

Re: Some AOKI goodies for Squeak 2.0
by Brian Klump 13 Jun '98

13 Jun '98

This is what I was talking about a few days ago...I'm going to try to clean up the main wikki server this weekend and if I get the time make it so that there's a good way to find things...granted wikki doesn't really support searching thearchives and all, but at least a well ordered app list would be really nice. What Ithink would be ideal would be a site similar to freshmeat.net, only the Squeak one would have wikkiness...the ability for normal people to edit it. :) -brian ---Marcus Denker <marcus(a)ira.uka.de> wrote: > > On Sat, Jun 13, 1998 at 08:01:44AM +0900, NISHIHARA Satoshi wrote: > > Hello Squeakers > > > > I've just ported some AOKI goodies for Squeak 2.0 (or 1.31). > > including > .... > > Hmm... there are so many wonderful squeak-toys (like these), but > no central repository... > > I think we need some kind of "Squeak library Wiki", a wiki-like > website, were everybody can add his non-standard Squaek Classes. > With a nice Class-Browser, good documentation (online editable), > links to "file out" a ChangeSet, some way to submit patches.... > > > -- > Marcus Denker marcus(a)ira.uka.de fon@home:(0721)614235 @work:(0721)608-2749 > Deliver yesterday, code today, think tomorrow. > > _________________________________________________________ DO YOU YAHOO!? Get your free @yahoo.com address at http://mail.yahoo.com

1 0

implementation details questions
by Florin Mateoc 13 Jun '98

13 Jun '98

Please excuse me if some of these issues have already been discussed and settled. If this is the case, could someone please post at least a reply "by reference" Thank you in advance. 1. Why is Object>>~~ implemented: self == anObject ifTrue: [^ false] ifFalse: [^ true] while Object>>~= is implemented: ^self = anObject == false >From my measurements it looks like the second construct is slightly faster. (Note: in VisualAge they use ^(self == anObject) not and it is the fastest implementation (among the above three mentioned) for VisualAge, which suggests that #not has more potential for optimization. For Squeak this is the slowest implementation, #not not being optimized at all (?)) 2. Object>>= could call directly the primitive 110, like Object>>==, instead of calling Object>>==. The same is true for Object>>hash, it could call directly primitive 75, like Object>>identityHash. 3. I looked at the implementations of #= and found that in most places there is a parameter type check included. This is a more general issue (not only for #=) and I think that these checks should be performed only when necessary (and that implies checking outside the called method) - it should be up to the developer to decide if this is the case. Besides, some methods are not even internally consistent, for example: Interval>>equal uses a weaker test and checks if the species of the parameter is the same with self's species (Array), but then in the method body it asks for the parameter's increment (and #increment is only implemented by Interval). 4. In Object>>deepCopy the message #deepCopy is further sent to the variables. Shouldn't that message be #copy - this should be enough in most cases, it is faster and it does not break for recursive structures - especially since Squeak also provides a true deep copy (#veryDeepCopy) ? The same Object>>deepCopy method also starts with the following (useless, in my opinion) line: (class == Object) ifTrue: [^self]. That's all for now, Florin

1 0

Some AOKI goodies for Squeak 2.0
by nishis＠urban.ne.jp 13 Jun '98

13 Jun '98

Hello Squeakers I've just ported some AOKI goodies for Squeak 2.0 (or 1.31). including 1. Bit Field 2. Blue Button Enhancements 3. Browser Menu Enhancements 4. Color Spuit 5. Cursors 6. Dump 7. Encyclopedia 8. File Numbering 9. Gauge MVC 10. Histogram 11. HTML Report Writer 12. Lisp 13. Prolog (alpha minus version) 14. Source Code Saver 15. Transcript File also 2-Dimensional Voronoi and Delaunay Diagram Analog Clock and Digital Clock and others. They are available for 2.0 or 1.31. http://www.urban.ne.jp/home/nishis/SmalltalkAgents/goodies/index.html http://www.urban.ne.jp/home/nishis/SmalltalkAgents/goodies/Squeak1.31/index… regards. ------------------------------------------- 西原聡士 (NISHIHARA Satoshi) e-mail: nishis(a)urban.ne.jp : tcc00164(a)niftyserve.or.jp URL: http://www.urban.ne.jp/home/nishis/ -------------------------------------------

3 2

A few questions...
by Chris Reuter 12 Jun '98

12 Jun '98

Hi all, I have a few other minor questions I'm hoping to get answered. 1) In Squeak 2.0, the "format" menu option merely causes the pane to flash as though an error had occurred, even if I've just recompiled the method. Is this a bug or has some subtlety been added to format? 2) I notice that when I fileOut a project (from the Change Sorter), all versions of the methods seem to be written to the file. Is there a way to file out only the latest versions of the methods, classes and class comments? I tried "remove doIts" but that took away the class definitions too. 3) In order to get rid of The Morphs That Wouldn't Die (re: previous posting), I tried closing and deleting all things morphic and evaluating the following: Morph allSubInstances do: [ :x | x become: nil] This caused Squeak to hang. Did I do something wrong? Thanks in advance for any answers you may offer. --Chris

3 2

Questions about FloatArrays
by Luciano Esteban Notarfrancesco 12 Jun '98

12 Jun '98

I continue working on FloatArrays, and I have some implementation questions. Squeak's Floats are 64 bits C doubles in the PowerPC ordering. In some plataforms (Intel for instance) the doubles are word-swapped, i.e. the lower 32 bits are swapped with the upper. I implemented FloatArray as a variableWord subclass of ArrayedCollection. I want to see FloatArrays as 'double *' from translated primitives... so I need to keep them in the ordering of the plataform in which Squeak is running. To handle this, I do the following: 1) I keep a class variable in FloatArray (SwapWords) that says whether I need to swap words when storing and retriving Floats. 2) I added to the method SystemDictionary|processStartUpList a line doing FloatArray startUp 3) I implemented FloatArray class|startUp this way: ((self with: 1.0) * 2.0) first = 2.0 ifFalse: [SwapWords _ SwapWords not. self allInstancesDo: [ :each | each swap]]. 4) and I implemented FloatAray|swap as it's ovious. What do you think about this scheme. Is all this right? Is there another way to do this, more close to the Squeak spirit? Having access to 'double *' from translated primitives alows doing sound analysis, high-quality speech synthesis and speech recognition... that's way all this is too important to me. Thanks in advance, Luciano.-

3 4

Morph Must Die!!! (a bug report)
by Chris Reuter 12 Jun '98

12 Jun '98

Hi all, I seem to have discovered a bug in Morphic. It seems that deleted Morphs continue to hang around and bloat the image. I was able to reproduce it in the following manner: 1) Create a Morphic world. 2) Eval "FlasherMorph allInstances size". The number should be 0 (unless you've created one already--I hadn't and I'm assuming that you're using a vanilla image). 3) Open a new Morphic world. In the world, create a new FlasherMorph. Eval'ing (2) should return 1. 4) Delete the Flasher from the World. Eval (2) again. It should return 1 now, rather than 0. 5) Try closing the World and doing (2) again. This sometimes makes the instance go away. Also, sometimes the WorldMorph doesn't go away either. I did some investigation via a script that would search all instance variables of all objects for references to a particular object (very slow) and the only suspicious (to my inexpert eye) reference was from a number of MethodContexts. BTW, I'm running Squeak 2.0 (with a VM I built from source, IIRC) under Linux. The problem persisted after running the auto-update (which is way cool, BTW). --Chris

3 2

Re: is *.im a st80 image file?
by Douglas E. Hammond 12 Jun '98

12 Jun '98

Geox wrote: > I got a program to modify an "ancient" Smalltalk (st80) base software, > it runs under Windows3.1 with a command line "st80.exe xxxx.im". > I think that *.im is the image file(?), now I want to transfer it to > a newer Win95 base platform . How can I do this? help me please Hans-Martin replied: > It's probably an old ParcPlace Systems (now ObjectShare) Smalltalk-80. > [...] [Sorry. I can't lay hands on my old DOS directory listings just now.] If these look familiar: s8b: total 24 drwxr-xr-x 3 deh other 2048 Nov 24 1960 . drwxr-xr-x 28 deh other 16384 Jan 1 1970 .. -rwxr-xr-x 1 deh other 428 Feb 3 1986 install.bat -rwxr-xr-x 1 deh other 955 Jul 13 1987 install2.bat drwxr-xr-x 2 deh other 2048 Mar 8 1995 release s8b/release: total 2068 drwxr-xr-x 2 deh other 2048 Mar 8 1995 . drwxr-xr-x 3 deh other 2048 Nov 24 1960 .. -rwxr-xr-x 1 deh other 7682 Feb 4 1986 animatin.st -rwxr-xr-x 1 deh other 1027 Oct 16 1985 blueinsp.st -rwxr-xr-x 1 deh other 392 Feb 17 1987 char.st -rwxr-xr-x 1 deh other 4960 Oct 14 1985 clock.st -rwxr-xr-x 1 deh other 1298 Oct 30 1985 comander.st -rwxr-xr-x 1 deh other 2798 Jul 13 1987 dosfiles.st -rwxr-xr-x 1 deh other 1254 Sep 16 1986 dynmscrl.st -rwxr-xr-x 1 deh other 7301 Feb 4 1986 financia.st -rwxr-xr-x 1 deh other 2807 Oct 16 1985 findclas.st -rwxr-xr-x 1 deh other 5880 Feb 4 1986 hand.st -r-xr-xr-x 1 deh other 740 Jan 28 1987 msm.com -rwxr-xr-x 1 deh other 1741 Feb 4 1987 printout.st -rwxr-xr-x 1 deh other 3547 Oct 17 1985 project.st -rwxr-xr-x 1 deh other 4069 Oct 31 1985 protocol.st -rwxr-xr-x 1 deh other 1672 Jul 13 1987 readme -r-xr-xr-x 1 deh other 60006 Jul 13 1987 st80.exe -r-xr-xr-x 1 deh other 594944 Jul 13 1987 st80v2.im -rwxr-xr-x 1 deh other 1393824 Mar 8 1995 st80v2.st -rwxr-xr-x 1 deh other 1372 Feb 4 1986 toothpa.st then we're looking at a system only marginally different from "vanilla" Smalltalk-80 Version 2: 'From Smalltalk-80, version 2, of April 1, 1983 on 11 March 1986 at 10:28:33 pm'! This system was marketed by SoftSmarts, Inc. (now long defunct). It featured a 286 protected-mode interpreter and a hardware protection device (a.k.a. a "dongle"). I'm a bit surprised to learn that the interpreter ran on top of Win 3.1. I'd be very surprised if it ran on top of Win 95. Though it's possible to write an interpreter for an image of this vintage, or to simply hack on the Smalltalk sources file (st80v2.st), Geox might be better advised to consider a more actively supported system such as Squeak. - Doug

2 1

← Newer
1
...
4
5
6
7
8
9
10
...
13
Older →