Squeak-dev February 1999

squeak-dev@lists.squeakfoundation.org

148 participants
356 discussions

RE: Squeak in IEEE Software
by agree＠carltonfields.com 01 Feb '99

01 Feb '99

Mark writes: >I wonder how much of his argument is related to architecture. If you have >a good underlying architecture that supports decoupled development efforts, >can you go further with unpaid-but-enthusiastic programmers? From what I >know of the literature, the data about errors per KLOC (thousands of lines >of code) is amazingly stable across projects and languages. But I don't >know how many of the studies have looked at truly O-O projects like Squeak >and Python. I think it has to do with a host of factors. Tools for programmers, including alternative operating systems, and utilities users of such systems might use are candidates for Cathedral development. What about accounting programs for law offices? One of the key virtues of OSS is that, if the program isn't behaving as you want it to be behaving, you can change it because you have the source. But what if you couldn't use the source code because you simply aren't technical, or if the costs of hiring programmers to learn the source and change it are large when compared to the benefits of the change? Suddenly, the presence of free source code changes nothing. While for key, widely used, applications, a consulting business of pre-trained experts in the application do and have prospered, who is going to develop a consulting practice for maintaining and making changes to more obscure applications? As the prices go up for such OSS consulting, traditional business models become more attractive. Now, mind you, OSS makes economically feasible things that were not before. Suppose there was a great, general purpose Word Processing program (closed source) named Letters. In the marketplace, users ask for features and publishers trade off costs and benefits in determining what the next upgrade will include. Their bottom line is usually to maximize profits by maximizing sales, which may mean that features highly valued by small numbers of customers will be given less weight than other more generally desired requests (for one-price/one-version business models). So these features, though highly desired in a small market, may never get built? Why not? Presumably an enhanced product would sell at a nice premium. Answer: Often, the marginal benefits of such software features do not justify building the entire framework of a Letters program, the cost of which can only be amortized by selling the general framework to a large number of people. But with OSS Letters, a consultant can be brought in to make the marginal changes at relatively low cost. Because Letters is a general purpose program, a practice of consulting can be built around supporting it, and thus the cost of learning Letters' code can be shared among all clients seeking Letters support. Suddenly, a new kind of software becomes economically possible. (Of course, here's where Mark's architecture arguments come in). At any rate, my key point is that OSS Cathedral development benefits may be limited to broad, general purpose programs, or programs for which the user community has a natural built-in technical capability. I think that scripting languages may in time go a long way to change this, effectively broadening the community of persons who can adapt adaptable software. (Again, I'm implicitly buying into Mark's architecture issues). But fundamentally, I think the issue of OSS' chances in the future is economic and not technical in nature, although of course technical issues will affect the economics -- and that these economic forces must be understood first and foremost to even begin to guess where OSS is going. For my part, the economic arguments are simply gratuitous -- the true virtue of OSS is its support of the hacker ethic -- information wants to be free. Now, as a stodgy IP attorney, its strange to see those words leaving my keyboard, but not so strange when you considered my roots. I still disagree strongly with gurus such as RMS who argue that the hacker ethic should be policy, for a host of reasons, economic and liberty-based. However, in MY OWN COMMUNITY, things like Squeak improve my life BECAUSE they are OSS, because its just plain fun to play with. Its the joy of hacking that supports things like Squeak, not the economics of its merits. Let's not lose track of the fact that some products, like Bind, are not labors of love for the community as a whole, despite their intensively important utility. Squeak is a joy for the people who use Squeak, and it is that joy that will drive its success under an OSS model. Were Squeak closed like its friends, my Xerox books might still be packed in a box in my attic and I might still be looking for something fun to hack. So, there is an economic component, yes. But lets not forget the community, fun, hacker ethic component as well. The latter, more than the former, drives some of these movements. I doubt we will find such interest and joy in hacking the law firm accounting program.

1 0

Possible bug in Form?
by Kevin Fisher 01 Feb '99

01 Feb '99

Hi all: I'm just toying around with some of the system morphs. I've been trying to evaluate the following snippet in a workspace: | syswin button frm | frm _ Form fromBMPFileNamed: 'close.bmp'. syswin _ SystemWindow labelled: 'test'. button _ ThreePhaseButtonMorph new. button offImage: frm. button extent: 16@16. syswin addMorph: button. syswin openInWorld. .....where close.bmp is a little 16x16 BMP file. However, when I evaluate this, I get a LOT of debugger popup windows which seem to be complaining about ColorForm>>colormapIfNeededForDepth. As far as I can tell it seems that the message pixelValueForDepth is being sent to 'nil' in: 1 to: colors size do: [:i | newMap at: i put: ((colors at: i) pixelValueForDepth: destDepth)]. When I inspect 'colors' it contains many values, but has a large section at the end filled with 'nil'...looks like the 'nil's shouldn't be there. Upon further inspection it seems that the 'nil's are coming right from the Form object, specifically in what gets read by fromBMPFileNamed. So, inspection of Form>>bmpColorsFrom:count:depth: revealed: | maxLevel colors b g r | colorCount = 0 ifTrue: [ "this BMP file does not have a color map" "default monochrome color map" depth = 1 ifTrue: [^ Array with: Color white with: Color black]. "default gray-scale color map" maxLevel _ (2 raisedTo: depth) - 1. ^ (0 to: maxLevel) collect: [:level | Color gray: (level asFloat / maxLevel)]]. colors _ Array new: (2 raisedTo: depth). 1 to: colorCount do: [:i | b _ aBinaryStream next. g _ aBinaryStream next. r _ aBinaryStream next. aBinaryStream skip: 1. colors at: i put: (Color r: r g: g b: b range: 255)]. ^ colors How I read this is that a generic color map is created based upon the depth. The colors are translated from the BMP file into the map using the colorCount as an index. This leaves the remainder of the map unassigned (set to nil). So, my question to someone who understands BMP's in the scheme of Squeak is...should the unfilled sections of the colour map be set to 'nil' (meaning the pixelValueForDepth method has the bugs) or, should the unfilled sections be set to black or some other generic colour (meaning the bmpColorsFrom method is the culprit)?

1 0

Re: Squeak 2.3 HP-UX port
by Markus Kohler 01 Feb '99

01 Feb '99

Hi guys, It seems I got it ! Unfortunately I have to leave right now, so I couldn't test it, but at least the image starts up. Look at http://www.stat.umn.edu/~luke/xls/projects/dlbasics/dlbasics.html There's a dlopen library for HP-UX. Markus -- Markus Kohler mailto:markus_kohler@hp.com

2 1

Re: Squeak 2.3 HP-UX port
by Bruce O'Neel 01 Feb '99

01 Feb '99

Hi, It's been a little while, but, I'm pretty sure HPUX does support shared libs. From the Tcl configure file: HP-UX-*.08.*|HP-UX-*.09.*|HP-UX-*.10.*) AC_CHECK_LIB(dld, shl_load, tcl_ok=yes, tcl_ok=no) if test "$tcl_ok" = yes; then SHLIB_CFLAGS="+z" SHLIB_LD="ld -b" SHLIB_LD_LIBS="" SHLIB_SUFFIX=".sl" DL_OBJS="tclLoadShl.o" DL_LIBS="-ldld" LD_FLAGS="-Wl,-E" LD_SEARCH_FLAGS='-Wl,+s,+b,${LIB_RUNTIME_DIR}:.' fi ;; Now, what does that all mean? Generating shared library code requires +z on the cc line, you generate shared libs with the -b option to ld, shared libraries are called blah.sl, you need to add -ldld to the ld command line to get the shared library library, and you also need to add -Wl,-E, probably to the cc command line, which, (guess), passes -E to the linker. If you go get the Tcl 8.0.3 distribution you can then find out what's in tclLoadShl.c, which does seem to be system dependent. Just for reference solaris uses tclLoadDl.c. On the system I do have access to uname -a responds HP-UX sewanee B.10.10 A 9000/725 and a man shl_load starts responding with: shl_load(3X) shl_load(3X) NAME shl_load(), shl_definesym(), shl_findsym(), shl_gethandle(), shl_getsymbols(), shl_unload(), shl_get(), shl_gethandle_r(), shl_get_r() - explicit load of shared libraries SYNOPSIS #include <dl.h> shl_t shl_load(const char *path, int flags, long address); If nothing else you can build the Tcl distribution and watch what it does. I hope this helps. I never became an expert on HP/UX :-( cheers bruce Markus Kohler <markus_kohler(a)hp.com> wrote: > > Hi all,, > I'm wondering whether someone has already ported > Squeak to HP-UX. > I just started, but after getting the Makefile correct > I came across the problem that HP-UX doesn't support > dlopen and those other shared library calls. > > Any shared library experts out there ? > > > > Markus > -- > Markus Kohler mailto:markus_kohler@hp.com > > == Bruce O'Neel - beoneel(a)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(a)caip.rutgers.edu> _________________________________________________________ DO YOU YAHOO!? Get your free @yahoo.com address at http://mail.yahoo.com

1 0

CFP: ECOOP'99 Multi-User Object-Oriented Environments Workshop
by Ian Piumarta 01 Feb '99

01 Feb '99

This might interest some of you. More details at: http://www-lifia.info.unlp.edu.ar/MOOEworkshop/ Ian > Multi-User Object-Oriented Environments Workshop > > In association with the > 13th European Conference on Object-Oriented Programming (ECOOP'99) > (http://ecoop99.di.fc.ul.pt/) > > Lisbon, Portugal, 14-18 June 1999 > > > MOOs are object-oriented environments where multiple users > synchronously and asynchronously interact in the context of virtual > worlds. MOO environments constitute a challenging context for > object-orientation theory and practice due to their unique > requirements for adaptability, simplicity of use, dynamic changes, > functional and non-functional domains to be considered, > domain-specific languages, software architectures, object-oriented > frameworks, coordination and distribution. > > The goal of this workshop is to bring together MOO users and > researchers to discuss the technical requirements of Multi-User > Object-Oriented Environments on object-oriented languages and > frameworks. [...]

2 1

Re: Help - MW.zip entry garbled
by nishis＠urban.ne.jp 01 Feb '99

01 Feb '99

try http://www.sugarweb.com/Miniatures/files/Morphic/MorphicWrappers.zip I can extract it by Mac stuffit expander, fine. At 08:41 02/01/1999 -0600, William Barnett-Lewis wrote: >Hello, > >Each time I have attempted to d/l MW.zip from the math morphs site, I >get a scrambled copy. I am on a Mac stuffit expander normally handles >.zip files just fine; but, alas, not this one. Is it available elsewhere >or as a .sit file? > >TIA, > >William ------------------------------------------- 西原聡士 (NISHIHARA Satoshi) e-mail: nishis(a)urban.ne.jp URL: http://www.urban.ne.jp/home/nishis/ -------------------------------------------

2 1

Re: X Bitmap support for Scamper
by Lex Spoon 01 Feb '99

01 Feb '99

Cool!! Here's a change to hook your code in for file: and ftp: accesses. Lex 'From Squeak 2.3 of January 14, 1999 on 1 February 1999 at 8:57:33 am'! !MIMEDocument class methodsFor: 'content-types' stamp: 'ls 2/1/1999 08:55'! guessTypeFromName: url "guesses a content type from the url" | extension | extension _ url asString. (extension includes: $.) ifFalse: [ ^self defaultContentType]. extension _ (extension findTokens: '.') last asLowercase. extension = 'html' ifTrue: [ ^'text/html' ]. extension = 'htm' ifTrue:[ ^'text/html' ]. extension = 'txt' ifTrue: [ ^'text/plain' ]. extension = 'gif' ifTrue: [ ^'image/gif' ]. extension = 'jpg' ifTrue: [ ^'image/jpeg' ]. extension = 'jpeg' ifTrue: [ ^'image/jpeg' ]. extension = 'xbm' ifTrue: [ ^'image/x-xbitmap' ]. extension = 'mid' ifTrue: [ ^'audio/midi' ]. ^self defaultContentType! ! William Cattey <wdc(a)mit.edu> wrote: > This is a multi-part message in MIME format. > --------------BE7B82566A9F4394CD619441 > Content-Type: text/plain; charset=us-ascii > Content-Transfer-Encoding: 7bit > > I hope I'm not gonna embarass myself too much publicly circulating my > first real Smalltalk subclass. > > This subclass teaches ImageReadWriter to read X bitmaps. > This would be useful, for example, if you used Scamper to look at > an MIT home directory through our school-wide web.mit.edu. > For Example: http://web.mit.edu/wdc/www/. > > The MIT web server pushes back a bunch of pretty file icons > that are in X bitmap format that Scamper, until now, just > complained about. > > Here is the class definition. I'd welcome any off-list replies > telling me style things I should do differently, or code > improvements. > > Special thanks to Paul Boutin who wrote the X Bitmap Parser > in C for the Andrew Toolkit when he worked for me ~5 years ago. > I used his most excellent code as the basis for this work, > and to show me how to do X Bitmap parsing. > > -wdc

1 0

Help - MW.zip entry garbled
by William Barnett-Lewis 01 Feb '99

01 Feb '99

Hello, Each time I have attempted to d/l MW.zip from the math morphs site, I get a scrambled copy. I am on a Mac stuffit expander normally handles ..zip files just fine; but, alas, not this one. Is it available elsewhere or as a .sit file? TIA, William

1 0

Postscript Canvas / Morphic Drawing
by Marcel Weiher 01 Feb '99

01 Feb '99

Squeakers, while working on a Canvas that produces Postscript code for printing (and could serve as a model for native drawing interfaces as well as a reference for the Balloon engine), I've come accross a couple of related problems that I think I need some feedback on before tearing into the code. TransformMorph The TransformMorph currently creates a FormCanvas, draws into that and then warp-blits that onto its own canvas. (I now understand why PlayWithMe 16bit is a tad slow...). For a PS compatible canvas, this can be handled by simply drawing the TransformMorph's submorphs with the transformation matrix appended to the graphics state. The obvious way of implementing this is to use the #canvasForSubmorphs: mechanism to let each Canvas decide what type of sub-canvas to return for drawing with rotate/scale transformations in place. A PostscriptCanvas simply returns a TransformedPostscriptCanvas (or maybe something with a shorter name :-) and a FormCanvas returns a TransFormCanvas, thus moving the warping code out of TransformMorph. However, there is a problem with that approach: the sub-canvas doesn't know when drawing starts and more importantly, when it's done. For the TransformedPostscriptCanvas, this is the time to restore the graphics state to its non-transformed state, for the TransFormCanvas it is the time to actually warp-blit the form onto the main canvas. Solution 1: bracket drawing on a sub-canvas with #beginDraw/#endDraw Solution 2: tell the sub-canvas to draw its contents onto the main canvas. Solution 3: ping-pong the drawing of submorphs between the canvas and the morph, so the canvas can do any bracketing it needs before/after calling the morph back to actually draw the submorphs. I currently favor solution 3, but have run into the problem that there are 2 drawing messages to forward, #fullDrawOn: and #fullDrawOn:without:andStopThere:, with lots of duplicate code. It seems to me that #fullDrawOn:without:andStopThere: is the more general method, and all implementations of #fullDrawOn: could be replaced with a single one that simply calls the more general one with nil values for the exclusion/stop morphs. If that is the case, I could just ping-pong one message instead of two. Text Text is currently drawn using the old character-scanning code called line-by-line from within NewParagraph's drawOn: method. It seems to be that what NewParagraph should do is send the individual lines to its canvas. Are the #text:at:font:color etc. methods intended to be the single-line text primitives to be called by the paragraphs drawing code, or is the intension for them to also take multi-line arguments? For now, I introduced a ping-pong method in Canvas: displayLine: line of: paragraph offset: offset leftInRun: leftInRun using: displayScanner ^ displayScanner displayLine: line offset: offset leftInRun: leftInRun This at least allows the PostscriptCanvas to get at the individual lines. Marcel

1 0

BUG: 2 buglets and a Nino PDA
by "Gran" Hultgren 01 Feb '99

01 Feb '99

Hi all! A tiny buglet found: "gives nil:" (RWBinaryOrTextStream on: 'abc') ascii next "but this gives $a:" (RWBinaryOrTextStream on: 'abc' from: 0 to: 3) ascii next Another buglet - if you remove ascii from above, the second version will not work - isBinary does not get initialized since it relies on a #reset to be sent to the stream, which does not occur when using #on:from:to: A silly fix to the first buglet would be to add (this little readLimit fellow sure does getting set a lot!): ReadWriteStream>>on: aCollection super on: aCollection. readLimit _ aCollection size (well, my image did not crash after this... :-) But I must say that the ReadWriteStream's hierarchy makes me a little bit confused. And if you start looking at instvar refs on readLimit it gets quite messy, or perhaps I am not quite getting it. I have not looked into other Smalltalks Stream hierarchies, but I assume that ReadWriteStream dates back a while - anyone like to comment on the rationale behind the hierarchy? Another little funny thing: Got the 2.2 VM running with the mini image on a Philips Nino - very cool! It just worked right from the start. Impressed. Now I just have to figure out how to press that Alt-key... Have not found out how to do that under PalmPC-WinCE. Does anyone know? Since my Nino has 8Mb of RAM and a flash disk with 20Mb I will try some fatter images to see what happens. Morphic maybe? Hehe.... slooooow, I suspect - but would be nice to see if it could work. "123 factorial" timed in at 1665 ms instead of 31 ms on my 233MhzPentium laptop... 52 times slower. Ok. But still - the UI stuff feels quite ok! It is a little bit slow but selecting text, typing etc. works fine. Hmmm, 52 times? Sounds a lot. Note: I had to browse the Time class and add the #millisecondsToRun: method since the debugger jumped up in my face, but hey, it felt so good to just hack in a method and then get it working ON A NINO. And the classic UI feels quite appropriate for such a small screen - just a little bit annoying with all the panes resizing all the time... :-) While browsing around the Stream hierarchy I came upon a few identical methods (bytecode identical also) in different classes, which made me wonder if perhaps there is a need for a tool for "discovering" stuff like that? It probably means that there is some redundancy going on that could be fixed. Or perhaps such a tool already exists? Well, thanks for listening and my regards to the Squeak team. The coming Squeaky months (Dan Ingalls list of stuff to wait for) sound truly interesting - as always! regards, Gran PS. What is the "correct" procedure for bug reports and fix proposals? Just drop it here in the mailing list with a nice header like BUG:etc.etc? DS === Gran Hultgren | goran.hultgren(a)inspire.se | OO Consulting Inspire | Hem: +46 8 54022950 | http://www.inspire.se Karlavgen 18 | Mobil: +46 708 813936 | icq#:6136722 114 31 Stockholm | "All my opinions are MINE, MINE, MINE!!!" _________________________________________________________ DO YOU YAHOO!? Get your free @yahoo.com address at http://mail.yahoo.com

1 0

Jump to page: