Squeak-dev October 2000

squeak-dev@lists.squeakfoundation.org

112 participants
424 discussions

RE: Porting Squeak to Bare Hardware
by Jay Carlson 31 Oct '00

31 Oct '00

Mark Guzdial [mailto:guzdial@cc.gatech.edu] writes: > Jim Roberts here at GaTech is attempting to port Squeak to the TINI > board. (See his page at http://coweb.cc.gatech.edu/csl/31 for lots > of links and additional info.) Cool page. > The TINI board is cheap (~$50) and can take up to 4M on the board > with reasonable FLASH memory. How much does it cost with all that memory attached? The obvious alternative to TINI for these applications is something like a uCsimm, but it's bigger and more expensive---8M RAM, 2M flash, $240 if you buy it built. That's not a particularly compelling price point, but welcome to the world of low volume systems. Hm, and then there's ARM's $200 eval kit---wait, I don't need *another* hobby architecture.... > The processor is a descendant of the > 8088 Not really. It's a decendent of the 8051, which is a much-propagated Intel family of 8-bit microcontrollers. > -- it's 32 bit internally, 16 bit data path. It's 8 bit internally, with various kludges to support expanded memory and a bolt-on-the-side math unit that you can push 16*16 multiplies and 32/16 divides into. (Actually, their expanded external addressing is a kludge on top of the 8051's external memory kludge....) It is one scary CPU for C code that thinks mostly in terms of 32-bit ints and void*s. > We've had three strategies that we've been exploring for getting > Squeak on there: > (1) Put Squeak into the 4M of memory and launch it via Java Native > Interface (JNI). The problem is that we can't find a C compiler to > work with this processor that supports JNI. And then fight with the garbage collector etc? Sounds like no fun. > (2) Rewrite the Squeak VM into something Java-ish that can run on top > of the native JVM. That seems like an enormous amount of work and > would run VERY slowly. All I have to say is: Two layers of interpreter on an 8-bit micro. :-) > (3) Trash the JVM and put a Squeak VM into the Flash memory. This > option sounds the most reasonable at this point, but we do this with > some trepidation -- we're essentially trashing the OS and porting to > bare hardware. For this platform, there is less to port than usual, > since there is no display, no mouse, no keyboard. But we also lose > even a filesystem once we trash the JVM. We also still have to > figure out someway of loading the image (maybe download it serially > via XModem, like the neat Helio port; or put it all in a C array like > Paul Fernhout's Squeakette (http://www.kurtz-fernhout.com/squeak)). I like the XModem approach (thanks :-), although TFTP shouldn't be too hard. In my poking around I found http://www.turbobit.com/tini.htm which is the start of a C-based loader/operating system/BIOS. Jay

1 0

Re: Apology
by Richard A. O'Keefe 31 Oct '00

31 Oct '00

Well, #toggleStartMonday or a preference retain optional ISO-8601 compliance. And with many French and German calendars, BTW. In German, Wednesday is called "Mittwoch". ISO 8601 handles both Monday-origin and Sunday-origin weeks. It's a while since I read it, so I don't recall whether it gives a preference to one or the other. toggleStartMonday is a really lousy interface. If I want to have weeks starting on Sunday, as they have for >2000 years, I have to do Week startMonday ifTrue: [Week toggleStartMonday]. I can well imagine people wanting to make sure Week will start with Monday. (It's basically ideological: "the only days that really matter are the work days".) I can only too well imagine people wanting to make sure Week will start with Sunday. But it is extremely difficult to invent a situation where someone wants to *change* which day of the week is counted as first without knowing or caring what it is right now. A better interface would be Week startsOn: #Monday Week startsOn: #Sunday But the fundamental design mistake is having the interpretation of weeks depend on a global mutable variable, as if Smalltalk had no threads and as if no two Smalltalk programmers could ever disagree about when weeks should start for the sake of their classes. As for what is the official start of the week, we might grant that it could depend on time, location, and faith. But Squeak 2.8 does not grant that. What is "official", anyway? The C standard rules that the first day of the week is Sunday and that it is numbered 0. That's "official". Some time ago, I had a good laugh at <http://zapatopi.net/metricweek.html>. The idea of tying a fundamental unit of time to the erratic rotation of one particular planet _is_ sufficiently farcical. (So does the idea of tying a fundamental unit of length to the circumference of the same planet, come to think of it.) Our ancestors got away with it on the rational grounds that it was the best clock they had. We don't have that excuse.

2 1

Re: Collecting Updates Again
by Henrik Gedenryd 31 Oct '00

31 Oct '00

Stefan Matthias Aust wrote: > Unfortunately, I haven't followed the discussions on the Squeak list for > the last three months or so and I'll probably be never able to catch up and > read more than 3000 pending emails. Therefore, I ask you all for some help. > > If you have posted a changeset to this list which hasn't been included in > the base image yet and you still consider it as an important contribution, > please send it to "squeak(a)3plus4.de" When I stepped in to collect updates from the summer, I went through everything submitted from Stefan's last set up to (and including) Sept 02. Therefore you should in principle not have to send him anything posted in that period. (I have also submitted a list of 2.9a updates to also release for a 2.8 update; this is presumably somewhere in the pipeline.) To all who want their contributions made into updates: Show appreciation for Stefan's _hard_ work and PLEASE PLEASE submit proper changesets, - ending with .cs not .st, and with an initials tag, like "Mychanges-hg.cs" (where hg are my initials) - with proper preambles, including full author name and a descriptive text that will go into the text listing of released updates that Dan sends out. - Use attachments that are saved to proper files, not text in the body of the mail. - Use Squeak's own .gz compression unless you _know_ that your mailer will deliver your attachments intact. All of this _must_ be done for all updates anyway, or they do not fit into the format required for the update release mechanism--and there is simply no excuse for making Stefan do this work for you!!! -- The hardest part of Stefan's work is the non-mechanizable job of judging whether a piece of code should be released or not. Therefore, the more descriptive your posting is about eg. rationale, image version compatibility, other prerequisites, etc., the better. Henrik

1 0

Question forkAt: Processor userSchedulingPriority
by Karl Ramberg 31 Oct '00

31 Oct '00

I made this method: progress [[foo getPercentage < 1] whileTrue: [Delay forMilliseconds: 300. progress value contents: ((foo getPercentage min: 1.0) max: 0.0)]] forkAt: Processor userSchedulingPriority Now I want to terminate it from another method before whileTrue: becomes false. How do I do that ? Karl

3 4

Help compiling plugins
by spair＠advantive.com 31 Oct '00

31 Oct '00

I've run into a snag in getting the MPEG plugin working on Windows. I've got the code compiling without MMX support into a DLL that looks good, however, Squeak doesn't seem to want to load it or call any of it's exported functions. I tried compiling TIPTestPlugin, and that isn't working either. But, when I re-compile DSAPlugin, no problems! I'm using VC++6 (with the latest SP installed)...and Squeak 2.8 (and 2.9a) I noticed that TIPTestPlugin uses the new primitive declaration syntax, while the DSAPlugin does not. Could there be a problem compiling these new style plugins using the MS tools? Can anyone else successfully compile TIPTestPlugin and call it? The only error indication is that when you try and call any of the primitives, they fall through and fail...debugging the process indicates that the DLL is not loaded. I know it must be something obvious, but I just can't see it. My next step is to try it with GCC. - Stephen

1 0

[FIX][ENH] SequenceableCollection>>from:to:put:
by Andres Valloud 31 Oct '00

31 Oct '00

Hi. Addresses a bug introduced in from:to:put:, update 2018. Andres.

1 0

[FIX] SequenceableCollection>>forceTo:paddingWith:
by Doug Way 31 Oct '00

31 Oct '00

>From the preamble: Fixes a bug in SequenceableCollection>>forceTo:paddingWith: introduced in update #2018. Executing ''asdf' forceTo: 3 paddingWith: $-' was causing an error, instead of being truncated. This new version still includes the optimizations from #2018, but with an extra check to handle this case. - Doug Way dway(a)riskmetrics.com

1 0

Re: [ENH][GOODIE] Process Browser "release candidate 1"
by Doug Way 31 Oct '00

31 Oct '00

Ned, here's one more non-critical tweak to my change which added the ID numbers to the process list. This tweak lines up the process names so that they're more readable again. - Doug Way dway(a)riskmetrics.com 'From Squeak2.9alpha of 13 June 2000 [latest update: #2774] on 30 October 2000 at 9:55:38 pm'! !Process methodsFor: 'printing' stamp: 'dew 10/30/2000 18:25'! browserPrintStringWith: anObject | stream | stream _ WriteStream on: (String new: 100). stream nextPut: $(. priority printOn: stream. self isSuspended ifTrue: [stream nextPut: $s]. stream nextPutAll: ') '. (self respondsTo: #processName) ifTrue: [stream nextPutAll: self processName] ifFalse: [stream nextPutAll: (self hash asString forceTo: 5 paddingWith: $ )]. stream space. stream nextPutAll: anObject asString. ^ stream contents! !

1 0

Re: Apology
by Bob Arning 31 Oct '00

31 Oct '00

On Tue, 31 Oct 2000 14:07:55 +1300 (NZDT) "Richard A. O'Keefe" <ok(a)atlas.otago.ac.nz> wrote: >I *did* find a bug while poking around (try inspecting >Week fromDate: (Date today), select self, and then choose Explore; >you'll get a MessageNotUnderstood) but didn't describe it. Well, could you describe it? I tried it in 2.8 and 2.9 and saw no MessageNotUnderstood. Cheers, Bob

1 0

ONCE AGAIN: 2.9a freezes! (WAS: Squeak becomes (almost) autistic???)
by jchludzinski＠worldkey.net 31 Oct '00

31 Oct '00

Bob, After much work (and some valuable assitance from Bob Arning) this weekend I was able to get my sim running without have Squeak become autistic (freezing up). This was accomplished by placing a Delay (1 second) between the #sendData: and #closeAndDestroy call on the socket. BUT when I try the exact same code on 2.9a - it freezes. Has anyone else seen this? Another engineer I work with has hung Squeak 2.9a by simply saving the image. ---John PS> I'm running on Win2K (a 1GHz Athlon + 256MB RAM + 30G HD box); so is the other engineer.

2 1

Jump to page: