Squeak-dev September 2001

squeak-dev@lists.squeakfoundation.org

168 participants
408 discussions

[Modules] Updates (was: Re: [modules] documentation pages on the swiki)
by danielv＠netvision.net.il 28 Sep '01

28 Sep '01

>From the modules doc site: > !Repository > > I imagine 'updates.list' becoming structured, or even merged with the whole > > repository mechanism. > > I think we could just open "updates" folders in the repository for people to upload into, > then the "update collectors" will somehow process them, perhaps by merging them > into three/four-decimal releases. Someone noted Debian's system of Unstable (a slightly > better name for test pilot), Testing (IIRC, used by test pilots for some time with few > complaints), and Stable.> To spell it out (one way this could be done) - Updates.list can be replaced by a versioned "SqC Full" configuration living in some repository. The configuration would start by having one Module - "Squeak base". Creating an update means creating another version of the module. Releasing updates would mean creating another version of the configuration, that requires the latest module version. Then slewing functionality off the image would be pretty trivial - someone makes a new module called "Celeste", moves all the code to that. The Harvesters test this is a proper replacement for the code in "SqC base". If so, they make an "SqC base" version with out the code. The next version of the configuration "SqC full" will include two modules - "SqC base", and "Celeste". Voila. Making and maintaining a PDA image (or StableSqueak image, or SqueakNOS with MorphicWrappers and NK's Connectors...) is also trivial - make another configuration, and when things move to other modules, be selective about adding them to the configuration. After we make this change, the "Celeste" module has it's own maintainer, and the Harvesters just choose when to update the dependency of "SqC full" to point at the newest (or not) Celeste versions. It's up to module maintainers to integrate submitted changes into their modules. The average joe would only need to pick a configuration he likes, and update from it whenever it has a new version. Does this fit everyone? Dan? Harvesters? Henrik? Other module maintainers? If so - Everything described is supported by SCAN today. A few things are missing to be practical - * A mapping to Henrik's modules, so a SCANItem can be identified as a new version of an image module, and loaded into it. * Code to do selective download of versions of Items we don't yet have, and that use the existing dependencies to get the closure - move from a configuration to the list of modules required to support them. * A simple "update what I need from the server and install it" option. Daniel Henrik Gedenryd <Henrik.Gedenryd(a)lucs.lu.se> wrote: > I have now created a page on the swiki for building documentation about the > modules system. Currently it lists the specification that exists, plus > copies of mail discussions that may be helpful. [snip]

1 0

[modules] Repositories (aka: What SCAN does) (was: Re: [modules] documentation pages on the swiki)
by danielv＠netvision.net.il 28 Sep '01

28 Sep '01

I saw the site includes discussion about having local repositories, and also about using repositories for holding things other than modules, like documentation, and preferences files. I tend to agree, and wanted to note that SCAN does a lot of this. A few points - * SCAN includes a server and a client, working in MVC and Morphic. The Server is also web accessible for reading. * SCAN manages hierarchies of Containers, within which are Items. Items can be modules, for example, or they can be other files, eg, documentation, preferences. * SCAN allows you to define distributions, so if you say a file is "private" (your preferences) it stays on you disk. When you make a module's distribution is the "world", that means you'll be able to sign it and publish it. Having all these, I think we should definitely look at it (the living implementation, not just the SCAN and Collage page on the swiki) before we start cloning. http://squeak.heeg.de/ seems like a good place to start. A few things I miss in SCAN - * No concurrency, because no shared write access. If I can write to it, you can't. This is bad, because it's not what I'd like to have as a collaborative process. Though it's still better than what we have today - "SqC can write to it, you can't. Wait for the Harvesters response, right after they deal with all the other changes posted, ever, to everything." * No integration with other tools. No "get me this module's latest version" from the Browser. No "make a new version of that module with this changeset". This should probably be fixed by integrating SCAN, the new modules work, and the system tools, which is loads of work we have ahead of us. Daniel Henrik Gedenryd <Henrik.Gedenryd(a)lucs.lu.se> wrote: > I have now created a page on the swiki for building documentation about the > modules system. Currently it lists the specification that exists, plus > copies of mail discussions that may be helpful. > > http://minnow.cc.gatech.edu/squeak/2042&

1 0

Re: Storing and Retrieving Point
by raymondasselin＠sympatico.ca 28 Sep '01

28 Sep '01

John M McIntosh <johnmci(a)smalltalkconsulting.com> wrote: > >I'm having trouble storing and retrieving Points. What am I doing wrong? > > > >pt _ Point new. > >pt _ 25@125. > >dict _ Dictionary new. > >dict at: 'first' put: #('text' pt). > >readPt _ ((dict at: 'first') at: 2). "readPt is now a Symbol" > > > >Thx for any pointers. > >Ken > > Mmm I think some others missed just a little wee issue. I noticed > that you assign pt the value Point new, then assign it 25@125. > Sometimes I've seen this from folks that are use to declaring the > type of a variable. In Smalltalk the variable is just a reference to > an object, it has no typing information. > > Therefore all you really need to do is > pt _ 25@125 > snip.... John What you say is new to me -There is lots of examples in Smalltalk which suggest something like Morph new, or Foo new, so for me this was not linked to the use of declaring the type of a variable. -I know that when you try to define new for a Class you defined, Squeak answer a caution 'new is used in the existing class system......' I usually ignore it and go forward. Your remark is intriguing to me...did this means that it's not a good way to go to define a 'new' as a 'builder' Class method in Squeak ? I'm a little bit 'confus' because before your note I was pretty sure that pt_ Point new. was the way to go !!! Raymond

3 2

RE: Localization in code
by goran.hultgren＠bluefish.se 28 Sep '01

28 Sep '01

Hi guys! "Andreas Raab" <Andreas.Raab(a)gmx.de> wrote: > Richard, > > > I rather like having messages in separate files because > > I don't mind having them in separate files ... what I *do* mind is having > them in external files exclusively and no relation to the context they're > used in. > > > + You can give them to someone to translate who doesn't have to know > > about the programming language involved. > > True and not true. As I said, the loss of context is a big issue here. It > makes it extremely hard for that person to get an answer to the question "in > what context are you using this phrase?!" which was effectively the only > question we ever got from people translating stuff for us. [SNIP] I read Andreas vision for a Localization browser and I like it. I also agree that mucking about with symbolic names for localized strings is really ugly and a pain. But the stuff about context above gives me one more idea - the ability to translate strings DIRECTLY on the screen (so to speak) would be even better than editing the code. This would abide to the Morphic value "directness" (or whatever it's called) and that would REALLY keep all the context! For example, I would like to switch to "Svenska" (swedish), run the system and when I find a String in the UI that is in english (fallback on a default language when there is no swedish translation available) I would like to translate it "in place" by using the halo after that it would show the swedish translation. Then of course, you would also be able to use a tool that can show you places that still are not translated. regards, Göran

2 1

CRLF problem with Squeak3.1a-4332.zip on FTP server?
by Bernhard Pieber 28 Sep '01

28 Sep '01

Hi all, I just downloaded Squeak3.1a-4332.zip from ftp://st.cs.uiuc.edu/Smalltalk/Squeak/3.0/unstable-testPilot/. I extracted it using WinZip and started it. When I open a browser on class ArchiveViewer and select the method #extractAll I get a warning that there is a problem with my sources or changes file. Seems like the old CRLF problem. I checked my WinZip settings. I have disabled all automatic conversions. I tried extracting using the ArchiveViewer in an old 3.1a image. The problem is the same. I also downloaded and tried Squeak3.1a-4325.zip. It seems to me that the files on the FTP server are corrupted. Can anyone confirm that I am not the only one having this problem? (It would not be the first time. ;-) - Bernhard

3 3

GC root table overflows.
by Scott A Crosby 28 Sep '01

28 Sep '01

I've never had a root table overflow when I've used squeak, so this isn't important to me. (Though one of my queue's I was working on would have have faced this issue; my redesign has, among other things eliminated this.) But some of you may have them frequently, If so, then a change along one of these lines this makes them at least 20x cheaper. I think that instead of doing a full GC, do an incremental GC and tenure the results immediately. Looking over the code, this seems to be easy, if its OK to do an incremental GC from within: ObjectMemory>>noteAsRoot:headerLoc: Is it? If so, then having the above function trigger an incremental GC, then tenure should work. If not, then just tenuring all of the objects should also work. Here are the changes (Note that I've not tested these yet): -------- incrementalGC "Do a mark/sweep garbage collection of just the young object area of object memory (i.e., objects above youngStart), using the root table to identify objects containing pointers to young objects from the old object area." | survivorCount startTime | self inline: false. DoAssertionChecks ifTrue: [self reverseDisplayFrom: 8 to: 15]. DoAssertionChecks ifTrue: [self validateRoots]. self preGCAction: false. "incremental GC and compaction" startTime _ self ioMicroMSecs. self markPhase. survivorCount _ self sweepPhase. self incrementalCompaction. allocationCount _ 0. statIncrGCs _ statIncrGCs + 1. statIncrGCMSecs _ statIncrGCMSecs + (self ioMicroMSecs - startTime). rootTableCount >= RootTableSize | (survivorCount > tenuringThreshold) ifTrue: ["Either a root table overflow; do an incremental GC and immediately tenure" "Or, too many survivors, so move up the young space boundary; this limits the number of objects that must be processed on future incremental GC's" statTenures _ statTenures + 1. self clearRootsTable. youngStart _ freeBlock]. self postGCAction. DoAssertionChecks ifTrue: [self reverseDisplayFrom: 8 to: 15] -------------- noteAsRoot: oop headerLoc: headerLoc "Record that the given oop in the old object area points to an object in the young area. HeaderLoc is usually = oop, but may be an addr in a forwarding block." | header | self inline: true. header _ self longAt: headerLoc. (header bitAnd: RootBit) = 0 ifTrue: ["record oop as root only if not already recorded" rootTableCount < RootTableSize ifTrue: ["record root only if there is room in the roots table" rootTableCount _ rootTableCount + 1. rootTable at: rootTableCount put: oop. self longAt: headerLoc put: (header bitOr: RootBit)] ifFalse: [self incrementalGC]] -- No DVD movie will ever enter the public domain, nor will any CD. The last CD and the last DVD will have moldered away decades before they leave copyright. This is not encouraging the creation of knowledge in the public domain.

2 1

Morphic Rolodex now on BSS
by John Hinsley 28 Sep '01

28 Sep '01

The Morphic Rolodex (subject of a tutorial a while back) is now on BSS as Morphic Rolodex001.002.pr. This version is actually in advance of the tutorial which should be updated by the beginning of next week if I don't get snarled up trying to get some Java stuff working under Windows (before I re-write the lot in Squeak!). Please download it and read the note about needing to set the number of your serial port (easy enough) before trying it out. It's not finished yet: the next step is to get it to print the addresses to file and to incorporate Karl Ramberg's WorldTimeMorph before turning it into a distributable stand-alone. Cheers John -- Can't cope anymore? Desperate for help? Join the 12 step program for those who yearn to give up Microsoft: http://home.earthlink.net/~penguinrox/index.html

1 0

Re: More Profiling
by Hernan Wilkinson 28 Sep '01

28 Sep '01

Hi everybody, I'm really enyoing this thread but don't have enough time to participate actively, so I decided to share with you the work I have done for my Master Thesis. The thesis was about "method lookup". I started with an idea, ended with another.... anyway, common problems. But there is something that I did that can be usefull for this kind of "profiling" and research. Scott Crosby wrote: > >QUERY: Can I write my caching code into a seperate class, yet have the >C-translater still inline all of the methods? It'd help a lot; Interpreter >is pretty unwieldy to browse&alter. Well, I have wrote a "framework" that helps to change the method lookup algorithm and used it with many method lookup implementations. One thing that I concluded in my thesis for example, was that making the method cache bigger make the method lookup take longer (At least with Squeak I thought we could get better performance) I does not fullfill Scott requeriment because I did not converted the framework to C code, I just run it inside Squeak. The frameworks extends the interpreter in such a way that forwards the method lookup to another class. Like I said, it runs inside Squeak (I did not generate a new virtual machine) so time measureaments are affected by the real Squeak's virtual machine. Here is a short description of how it works and how to use it. 1) The class ConfigMLInterpreter (not the best name) subclasses InterpreterSimulatorLSB. This class redefines the method lookup methods and forwards the lookup to a MethodLookupStrategy object 2) MethodLookupStrategy has many implementations, one of them is DispatchTableSearchML (the basic lookup algorithm) and CachingStrategy 3) CachingStrategy has many implementations, one of them is GlobalCachingStrategy that is the copy of the current Squeak algorithm. It uses the same caching algorithm, etc. 4) You can define profilers for each MethodLookupStrategy and CachingStrategy. There are some implemented already that take time measurements. 5) There is a class StopWatch, that uses the real time clock to get better time measurement resolution. I implemented a bunch of primitives in C code that uses that clock. This primitives are only valid in a Windows operating system. 6) The idea is that with this framework you can run another Squeak image that has all the tests that you want to do. I used a WinCE image that has some tests, for example, compile a method, decompile a method, read a file, etc. 7) The image that contains the tests, comunicates with the "virtual machine" running on Squeak throught primitives that are implememented in ConfigMLInterpreter How to use it: The class ConfigMLInterpreter has a class message named #run: anImageFileName using: aMethodLookupStrategyClass to start running another Squeak image. It has some shortcats, for example #runUsingGLC that starts the running the image with the Squeak's algorithm. It the Squeak image that is beeing run inside the original Squeak, there is a ThesisBenchmarks class that has messages to run predifined test like #runAllTestsUsingDispatchTableSearchML, #runAllTestsUsingGLCWithDTS, etc. The test's result are printed in the original Squeak Transcript I implemented it with a Squeak 2.7 image. This is the only English documentation I have since the thesis is written in Spanish. Well I hope this helps. Here are the files: 1) Framework.zip: It contains the framework's code. Must be file it in the main image. 2) testClasses.zip: It contains the test's classes that have to be file in in the image that will be run inside the Squeak with the framework 3) Interp.zip: It has the changes I made to the file Interp.c to support the real time clock primitives. The changes are mark with the letters HAW I have UML diagrams that show how this thing works, and of course, an doc written in Spanish. Any question or suggestion will be welcome. Bye, Hernan. (See attached file: Framework.zip)(See attached file: TestClasses.zip)(See attached file: Interp.zip) -------------------------------------------------------------- Hernán A. Wilkinson Arquitectura Tecnologica - Arquitectura Corporativa Gerencia de Sistemas - Banco Galicia Tte. Gral. J. Perón 537, 2 "B" C1038AAK Tel.: (54- 11) 6329-4630 Fax: (54- 11) 4329-6568 e-mail: hernan.wilkinson(a)bancogalicia.com.ar =============================================================== La información contenida en este correo es para uso exclusivo de los destinatarios del mismo. Está prohibido a las personas o entidades que no sean los destinatarios de este correo, realizar cualquier tipo de modificación, copia o distribución del mismo. Si Ud. recibe este correo por error, tenga a bien notificar al emisor y eliminarlo.

2 1

Computerchannel.de: Squeak 3.0 tested
by Marcus Denker 28 Sep '01

28 Sep '01

Hi! Squeak was evaluated by some german computer-online-magazin, the article is available (in german): http://www.computerchannel.de/softwaretest/tools/squeak30/squeak30_1.phtml Of course, they didn´t understand it. And the review is awful, but it shows how "normal" users see squeak... Marcus

11 14

RE: Localization in code
by Richard A. O'Keefe 28 Sep '01

28 Sep '01

"Andreas Raab" <Andreas.Raab(a)gmx.de> wrote: > + Variant: the same image can provide messages in English > when run by me > or Maori, Chinese, whatever when run by a student who'd prefer that. > (Yeah, dream on...) The student wouldn't have to start a > different image or load a special change-set. What do you mean "dream on"?! I mean: the existence of a Maori translation AND a Squeak font that can display the characters it needs (ISO Latin 4 or ISO Latin 6 will do nicely thank you), Unicode support in "standard" Squeak. Do you realize that what you're describing is already practice in the eToy system?! ISO Latin 4 or ISO Latin 6 fonts are already practice in eToy? This I *gotta* see! How do I get them?

2 1

← Newer
1
...
5
6
7
8
9
10
11
...
41
Older →

Jump to page: