Squeak-dev December 2000

squeak-dev@lists.squeakfoundation.org

96 participants
272 discussions

Re: Help me!
by Bert Freudenberg 19 Dec '00

19 Dec '00

On Tue, 19 Dec 2000, Sven Delaney wrote: > System Description: > Linux Red Hat 5.2.2 (Apollo) > Netscape 4.6 > > Problem Description: > Cannot get Squeak! > > Ref: > http://www.squeak.org/ Try these rpms ftp://st.cs.uiuc.edu/pub/Smalltalk/Squeak/2.8/Squeak-2.8-3.i386.rpm ftp://st.cs.uiuc.edu/pub/Smalltalk/Squeak/2.8/Squeak-image-2.8-3.noarch.rpm If this doesn't work for you, please give us more details why you cannot get Squeak. -- Bert

1 0

Re: Squeak Plug-in
by Jochen F. Rick 19 Dec '00

19 Dec '00

> It's implemented already, but not wired yet. Try What does "not wired yet" mean? I integrated this with our code and it works fairly well, though having the browser go there instead of waiting for download to the plugin would be better. Excellent work!!! Peace and Luck! Je77

2 1

The Puzzle Page
by Dan Ingalls 19 Dec '00

19 Dec '00

Folks - Here are a couple of problems I've noticed, and which I would love to clean up right now. I enjoy such projects in much the same way that I enjoy the puzzles in the Sunday paper. However, right now I have other projects going on that I don't want to slight. It occurred to me that some other eager Squeakers out there might equally enjoy the challenge, and the jobs might get done sooner if I gave them away. So, here they are (these all happen to be morphic)... Have you noticed the new preference, debugShowDamage? It flashes every part of the screen that gets rewritten during morphic's display cycle. Turning this preference on will be very useful for these projects. 1. Open, eg, a browser (with pop-out scrollbars). Move your cursor in and out of a list pane, and notice that the contents do not get rewritten (no damage flash). Now move your cursor in and out of a text pane. Notice that the entire text contents gets rewritten. Although the region of the selection does get redisplayed, it should not bve necessary to repaint the entire text area. Fix this. 2. Open any two browsers. Click first in one, then the other. Note that the entire window gets repainted not once but twice. Fix this. 3. Open a browser. Collapse it. Expand it. Note that the entire merge of the before and after rectangles gets repainted. Also, if you slow it down, it looks like even more happens when collapsing a system browser. Fix this. [A word about morphic damage management: The damage manager will merge intersecting rectangles. If you know that an entire region needs to be repainted, you can report that at the beginning with self changed, or self invalidRect: someRect, and this will frequently cause a more effective grouping of damage rectangles. You may find some code that forces a redraw to flush the damage set. The invalidRect: method is a better approach, and you may wish to change the old code]. 4. Open one morphic project within another. Enter the sub project. Then return to the previous project. Notice that on each transition the project entry code painstakingly paints the entire screen for the new project, and then repaints it once the project is entered. Fix this. Check it with both settings of the projectZoom preference. 5. Open a MonthMorph, and notice its behavior when you change its extent using the morph menu resize command. Notice that it repaints the entire region of the screen from screen topLeft to the bottomRight of the MonthMorph. Fix this. Now use the halo resize control. Notice how much slower this is. Fix this. Now open a PDAMorph. Resize it. Make it so the monthMorph fits its pane properly. Now one more thing (but it may be fixed by the time you get to this): when you first open a PDA (with pop-out scrollbars), move the cursor in and out of one of the panes on the left. Notice that it invalidates the entire upper left quadrant of the screen. Now resize the category pane, using the little yellow pane boundary handle. Notice that entering and leaving the left panes no longer invalidates the outer screen. Make it so this is true from the beginning. People sometimes comment that morphic seems unduly slow. It's now wonder, given some of these problems. If we make it through the puzzle page this month, I think things will start to feel a good bit faster. It's probably worth announcing to the list your intentions to solve any of these so we can work together (teamwork can be fun and effective here), or at least avoid duplication of effort. Have fun! - Dan

4 4

Re: [FIX] Halo Invalidation Tweak
by Bob Arning 18 Dec '00

18 Dec '00

On Mon, 18 Dec 2000 12:49:44 -0500 Doug Way <dway(a)riskmetrics.com> wrote: >Bob Arning wrote: >> >> The big damage rectangle is the HaloMorph itself (the thing that the handles are submorphs of). It could be argued that, since this guy is transparent, he doesn't need to report damage, but then you would need to be sure to report the damage for all the individual parts (handles and name). Which would be faster would depend a lot on what was on the screen. > >Ah, that makes sense. Probably it mostly just depends on the size of the morph... for large morphs it would be faster to report damage for the individual parts, but for small morphs the opposite. I suppose you could have a threshold size with which it decides between the two options. > >Related to that, at some point it might be interesting to put a fancier algorithm in DamageRecorder>>recordInvalidRect: which determines whether damage rectangles should be combined together based on how large they are, how far apart they are, etc. Hi Doug, It's probably even more complicated than that. A single, large rectangle may end up causing more morphs to redraw, while many, smaller rectangles may end up causing one morph to draw many times. Depends on what's beneath the damage rectangles. Pondering improved redraw logic is fun, but I haven't found the perfect solution yet. ;-) Cheers, Bob

1 0

Re: Advice on squeaklets needed
by Michael Rueger 18 Dec '00

18 Dec '00

Bert Freudenberg wrote: > Older one? It's a vanilla 2.8 image (#2348) minus VM construction minus > FFI plus the browser primitives/code loader minus sources plus variable > names, amounting to a 2.5M download (5.5 M expanded), which is not too bad > IMO. It starts up in Morphic and displays a little icon while loading. Oops, sorry, I misinterpreted the comment on the page. Blame it on jetlag ;-) Michael -- "To improve is to change, to be perfect is to change often." Winston Churchill +------------------------------------------------------------+ | Michael Rueger m.rueger(a)acm.org ++1 (310) 937 7196 | +------------------------------------------------------------+

1 0

Re: [FIX] Halo Invalidation Tweak
by Bob Arning 18 Dec '00

18 Dec '00

On Mon, 18 Dec 2000 01:56:39 -0500 Doug Way <dway(a)riskmetrics.com> wrote: >After playing with showing the damage rectangles, I was thinking that >maybe this was because the various halos (and images?) added up to more >than 15 rectangles, so it was just combining them into one rectangle. >However, I changed the line in DamageRecorder>>recordInvalidRect: to >"invalidRects size >= 500", and it still redraws the big rectangle, so >it must be invalidating it somewhere else, but I haven't dug into it. I >guess I'd like to add this one to the "Puzzle Page" if someone wants to >look into it. :) Hi Doug, The big damage rectangle is the HaloMorph itself (the thing that the handles are submorphs of). It could be argued that, since this guy is transparent, he doesn't need to report damage, but then you would need to be sure to report the damage for all the individual parts (handles and name). Which would be faster would depend a lot on what was on the screen. Cheers, Bob

1 0

2.9 speed
by Stephane Ducasse 18 Dec '00

18 Dec '00

Hi all I looked at the last verison 2.9a3125 and it seems to me slower than 2.8. Is it subjective? Is it just an "eye related" symptoms or really a speed issues? Does anybody else got the same feeling? Stef

2 1

Re: Command line arguments?
by Bob Arning 18 Dec '00

18 Dec '00

On Sun, 17 Dec 2000 17:10:43 -0600 (CST) Aaron J Reichow <reic0024(a)d.umn.edu> wrote: >Is there any way Squeak can get a hold of arguments passed to it on the >command line? Like C's argv? Aaron, I suspect that this is quite platform- and VM-dependent, but see the comment in SystemDictionary>>getSystemAttribute:. Cheers, Bob

2 1

[updates] A couple more...
by Dan Ingalls 18 Dec '00

18 Dec '00

... of possible interest for shrinking. - D ------------------------- 3126betterInterpolate-raa -- Bob Arning -- 17 December 2000 - allow use of B3D on crunched sketches in the storyboard" 3127shrinkFix-raa -- Bob Arning -- 17 December 2000 Some impeovements to the #majorShrink process - faster rehashing of MethodDictionaries - pruning some class variables that kept old stuff around - better reduction of the (weak) SymbolTable - elimination of a direct reference to WorldState in AutoStart that caused a walkback when saving - reimplementation of #discardMIDI to avoid a walkback Smalltalk majorShrink produced a 888,336-byte image in 2.9a at change set 3114

1 0

Headless Unix 2.8 VM?
by Aaron J Reichow 18 Dec '00

18 Dec '00

I am trying to get up a little SqueakScript image and VM for a little project of mine. I am working with a 2.8 image made on my Mac on a Linux machine. It's got David Lewis' SqueakScript changeset and OSProcess 1.4 filed in, as well as selective things executed from #majorShrink. I tried building a headless VM from the CVS sources from Sourceforge. the BUILD text file points to "make headless" being sometihng useful, but headless isn't defined as a target in any of the Makefiles (even after the creation of the proper platform directory and running configure). With no headless target, I tried adding a "#define HEADLESS" to sqXWindows.h (because it's included by sqUnixMain.c) and to sqCopyImage.h (because it's included by sqXWindows.c and sqCopyImage.c), in an attempt to get it to compile. When I look at sqXWindow.c at the supposedly error-containing lines, it seems it's using variables which already do exist, and are included earlier in sqXWindow.c. The error I get while compiling is so: root@nubrain:~/svm/i586-pc-linux-gnu# make gcc -g -O3 -fomit-frame-pointer -Wall -I. -I../src -I../src/CoreVM -DSQVERSION="\"2.8a\"" -DHAVE_CONFIG_H -DUNIX -c ../src/sqXWindow.c ../src/sqXWindow.c: In function `setLazy': ../src/sqXWindow.c:670: `sleepWhenUnmapped' undeclared (first use in this function) ../src/sqXWindow.c:670: (Each undeclared identifier is reported only once ../src/sqXWindow.c:670: for each function it appears in.) ../src/sqXWindow.c: In function `ioLoadNextEvent': ../src/sqXWindow.c:1491: `eventBufGet' undeclared (first use in this function) ../src/sqXWindow.c:1491: `eventBufPut' undeclared (first use in this function) ../src/sqXWindow.c:1508: `eventBuf' undeclared (first use in this function) ../src/sqXWindow.c:1514: `EVENTQ_SIZE' undeclared (first use in this function) ../src/sqXWindow.c: In function `ioGetKeystroke': ../src/sqXWindow.c:1621: warning: unused variable `keystate' make: *** [sqXWindow.o] Error 1 Anyone have any ideas? I'd also be fine with just recieving a pre-compiled headless 2.8 VM. :) Aaron Aaron Reichow :: "In essence, Smalltalk is a programming Twin Ports ACM Vice President :: language focused on human beings rather http://www.d.umn.edu/~reic0024/ :: than the computer." - Alan Knight

1 0

← Newer
1
...
6
7
8
9
10
11
12
...
28
Older →

Jump to page: