Squeak-dev January 2002

squeak-dev@lists.squeakfoundation.org

176 participants
612 discussions

[GOODIE] Date format arrays translation
by Leandro Caniglia 02 Jan '02

02 Jan '02

>From preamble: Change Set: DateFormat Date: 1 January 2002 Author: Leandro Caniglia If you cannot recall the format array used in Date printing methods, this goodie is for you. Print dates using messages as in: Date today printWithPicture: 'dd-mmm-yyyy' Date today printWithPicture: 'yy/m/d' Date today printWithPicture: 'month d yyyy' and so on. Want to know the meaning of printFormat: #(3 1 2 $. 2 1 2) ? use: DateFormat fromArray: #(3 1 2 $. 2 1 2) and get 'yyyy.dd.mmm' /Leandro

1 0

Re: Update on the Squeak-material on my site
by Maarten Maartensz 02 Jan '02

02 Jan '02

Hello. Doug Way wrote: >>Maarten Maartensz wrote: >> >> L.S. >> >> It may interest some to know that I have given Squeak its own directory on >> my site and have extended the html- versions of my user manual (in progress). >> >> The relevant direct link is: >> >> http://www.xs4all.nl/~maartens/squeak/squeak_index.htm >Thanks for the link; your diary about discovering Squeak is interesting. >A minor problem: The "Map", "Tour", etc., links from your Squeak page >seem to be pointing to invalid (local) addresses such as >file:///C:/webmm/map.htm. You may want to fix these. Thanks for the information. I'll look into it, and only add that while I DO make mistakes, FrontPage 2000 seems to have gone into a buggy mode again, for it *ought* to rewrite these links automatically from my harddisk to my servers harddisk: I paid for the program i.a. to have this facility. O, well ... I do want to fix this, indeed, and will soon. >> ... >> Having seen my own site recently on a local internet-cafe, that did not >> display part of my site as it should (probably due to a combination of >> factors including MS buggery): >> >> - my site should display in Arial 12 point bold >Looks fine to me on IE and Netscape. Thanks. It turns indeed out that the display at various places, even in nominally the same browser, may differ considerably (due to browser settings, computer settings, video settings, internet-cafe settings etc.). Not having an eternal life, I spend systematic effort only on trying to assure it does work on the kind of set-up I use. >Also, I don't mind your frequent use of coloring text (usually red) for emphasis, >but I might suggest that you use a different color such as blue (ideally the >default html-link blue color) rather than red for links, so that the links are easier >to distinguish from the regular emphasized text. Yes. However, what I found - the last 15 years of daily computing - is that it is difficult enough to work well withing one format (such as html), and nearly or totally impossible to satisfy from within such a format different conventions useful for other file-formats. So by and large what I try to do is having decent html that is recognisably the same over my site. (And long before knowing Squeak I settled on red+underlining for links.) >This confusion is a bigger problem if you're using Scamper in Squeak... >the links there are not underlined but are represented with colored text. This I haven't tried out, and indeed at the moment the differences in capacity between Scamper and standard html-browsers are too large to try to satisfy both on my site. >This makes me think that it might be a good idea to change Scamper >so that hyperlinks are underlined. Ignoring the fact that all other web >browsers do this, using underlines just seems like a better idea for >hyperlinks... using colored text basically prevents you from ever using >colored text for anything else. The underline seems to be a more significant >change in emphasis than a different color, which, for a hyperlink, is good, >IMHO. (Underline also seems more appropriate than bold, italic, strikethru, etc.) >Maybe I'll submit a changeset sometime if noone else does. :) That seems altogether a good idea to me. And maybe another thing you might try, which is convenient to have is that the format changes of a link when the mouse is on it. For example, in part of the files on my site the links - in IE at least - change to small capitals and change back to normal when the mouse is removed. This is helpful to see what is and isn't a link, but may not be easy to implement. In any case: Underlined hyperlinks would make Scamper's representation of links conform to what non-Squeak users expect. - Doug Way dway(a)riskmetrics.com Regards, Maarten. -------------------------------------------------------- Maarten Maartensz. Website in Amsterdam about philosophy, logic, M.E. (Myalgic Encephalomyelitis) and much more: http://www.xs4all.nl/~maartens/ more than 25 MB of stimulating and original ideas. ---------------------------------------------------------

1 0

Re: getting really weird: Player39 in PointerFinder
by Bob Arning 02 Jan '02

02 Jan '02

On Tue, 1 Jan 2002 14:19:45 -0800 Ross Boylan <RossBoylan(a)stanfordalumni.org> wrote: >I don't know how Etoy scripting ended up with one of my windows as an owner... Ross, Is there a method #script1 in Player39? What does it say? This may give a clue as to how things got started. Cheers, Bob

2 1

[ENH] Interpreter, ObjectMemory activate
by John M McIntosh 02 Jan '02

02 Jan '02

I'll throw this out for comment. I've been looking at why sends are slower than say 2.8. As part of this work I cleaned up a few methods to improve performance of message sends. So perhaps someone can review and let me know if this makes a difference? Of course letting me know if I've not broken anything would be cool too, mind usually at this level mistakes are very fatal... 'From Squeak3.2alpha of 1 November 2001 [latest update: #4599] on 1 January 2002 at 10:38:25 pm'! "Change Set: SlightlyFasterActivate-JMM Date: 1 January 2002 Author: johnmci(a)smalltalkconsulting.com A few minor changes to clean up the code for method activation. In some places too many if statements are executed, and in others we have a common sub expression that is evaulated as we stuff values into the context header. Most compilers mostly factor that out, but if we explicitly do it then we get much better assembler code for the for loops. This improves messages sends by about a percent or so"! !ObjectMemory methodsFor: 'allocation' stamp: 'JMM 12/29/2001 19:18'! allocate: byteSize headerSize: hdrSize h1: baseHeader h2: classOop h3: extendedSize doFill: doFill with: fillWord "Allocate a new object of the given size and number of header words. (Note: byteSize already includes space for the base header word.) Initialize the header fields of the new object and fill the remainder of the object with the given value." | newObj remappedClassOop end i | self inline: true. "remap classOop in case GC happens during allocation" hdrSize > 1 ifTrue: [ self pushRemappableOop: classOop ]. newObj _ self allocateChunk: byteSize + ((hdrSize - 1) * 4). hdrSize > 1 ifTrue: [remappedClassOop _ self popRemappableOop. hdrSize = 2 ifTrue: [self longAt: newObj put: (remappedClassOop bitOr: HeaderTypeClass). self longAt: newObj + 4 put: (baseHeader bitOr: HeaderTypeClass). newObj _ newObj + 4] ifFalse: [self longAt: newObj put: (extendedSize bitOr: HeaderTypeSizeAndClass). self longAt: newObj + 4 put: (remappedClassOop bitOr: HeaderTypeSizeAndClass). self longAt: newObj + 8 put: (baseHeader bitOr: HeaderTypeSizeAndClass). newObj _ newObj + 8]] ifFalse: [self longAt: newObj put: (baseHeader bitOr: HeaderTypeShort)]. "clear new object" doFill ifTrue: [end _ newObj + byteSize. i _ newObj + 4. [i < end] whileTrue: [self longAt: i put: fillWord. i _ i + 4]]. DoAssertionChecks ifTrue: [ self okayOop: newObj. self oopHasOkayClass: newObj. (self objectAfter: newObj) = freeBlock ifFalse: [ self error: 'allocate bug: did not set header of new oop correctly' ]. (self objectAfter: freeBlock) = endOfMemory ifFalse: [ self error: 'allocate bug: did not set header of freeBlock correctly' ]. ]. ^ newObj! ! !ObjectMemory methodsFor: 'allocation' stamp: 'JMM 12/29/2001 19:05'! allocateOrRecycleContext: needsLarge "Return a recycled context or a newly allocated one if none is available for recycling." | cntxt | needsLarge = 0 ifTrue: [freeContexts ~= NilContext ifTrue: [cntxt _ freeContexts. freeContexts _ self fetchPointer: 0 ofObject: cntxt. ^ cntxt]. cntxt _ self instantiateContext: (self splObj: ClassMethodContext) sizeInBytes: SmallContextSize] ifFalse: [freeLargeContexts ~= NilContext ifTrue: [cntxt _ freeLargeContexts. freeLargeContexts _ self fetchPointer: 0 ofObject: cntxt. ^ cntxt]. cntxt _ self instantiateContext: (self splObj: ClassMethodContext) sizeInBytes: LargeContextSize]. "Required init -- above does not fill w/nil. All others get written." self storePointerUnchecked: 4 "InitialIPIndex" ofObject: cntxt withValue: nilObj. ^ cntxt ! ! !ObjectMemory methodsFor: 'allocation' stamp: 'JMM 12/29/2001 17:43'! recycleContextIfPossible: cntxOop | header | "If possible, save the given context on a list of free contexts to be recycled." "Note: The context is not marked free, so it can be reused with minimal fuss. The recycled context lists are cleared at every garbage collect." self inline: true. "only recycle young contexts (which should be most of them)" cntxOop >= youngStart ifTrue: [header _ self baseHeader: cntxOop. (self isMethodContextHeader: header) ifTrue: ["It's a young context, alright." (header bitAnd: SizeMask) = SmallContextSize ifTrue: ["Recycle small contexts" self storePointerUnchecked: 0 ofObject: cntxOop withValue: freeContexts. freeContexts _ cntxOop] ifFalse: [(header bitAnd: SizeMask) = LargeContextSize ifTrue: ["Recycle large contexts" self storePointerUnchecked: 0 ofObject: cntxOop withValue: freeLargeContexts. freeLargeContexts _ cntxOop]]]] ! ! !Interpreter methodsFor: 'message sending' stamp: 'JMM 12/29/2001 15:39'! activateNewMethod | newContext methodHeader initialIP tempCount nilOop where argCount2 | methodHeader _ self headerOf: newMethod. newContext _ self allocateOrRecycleContext: (methodHeader bitAnd: LargeContextBit). initialIP _ ((LiteralStart + (self literalCountOfHeader: methodHeader)) * 4) + 1. tempCount _ (methodHeader >> 19) bitAnd: 16r3F. "Assume: newContext will be recorded as a root if necessary by the call to newActiveContext: below, so we can use unchecked stores." where _ (self cCoerce: newContext to: 'char *') + BaseHeaderSize. self longAt: where + (SenderIndex << 2) put: activeContext. self longAt: where + (InstructionPointerIndex << 2) put: (self integerObjectOf: initialIP). self longAt: where + (StackPointerIndex << 2) put: (self integerObjectOf: tempCount). self longAt: where + (MethodIndex << 2) put: newMethod. "Copy the reciever and arguments..." argCount2 _ argumentCount. 0 to: argCount2 do: [:i | self longAt: where + ((ReceiverIndex+i) << 2) put: (self stackValue: argCount2-i)]. "clear remaining temps to nil in case it has been recycled" nilOop _ nilObj. argCount2+1+ReceiverIndex to: tempCount+ReceiverIndex do: [:i | self longAt: where + (i << 2) put: nilOop]. self pop: argCount2 + 1. reclaimableContextCount _ reclaimableContextCount + 1. self newActiveContext: newContext.! ! !Interpreter methodsFor: 'message sending' stamp: 'JMM 12/29/2001 17:26'! internalActivateNewMethod | methodHeader newContext tempCount argCount2 needsLarge where | self inline: true. methodHeader _ self headerOf: newMethod. needsLarge _ methodHeader bitAnd: LargeContextBit. (needsLarge = 0 and: [freeContexts ~= NilContext]) ifTrue: [newContext _ freeContexts. freeContexts _ self fetchPointer: 0 ofObject: newContext] ifFalse: ["Slower call for large contexts or empty free list" self externalizeIPandSP. newContext _ self allocateOrRecycleContext: needsLarge. self internalizeIPandSP]. tempCount _ (methodHeader >> 19) bitAnd: 16r3F. "Assume: newContext will be recorded as a root if necessary by the call to newActiveContext: below, so we can use unchecked stores." where _ (self cCoerce: newContext to: 'char *') + BaseHeaderSize. self longAt: where + (SenderIndex << 2) put: activeContext. self longAt: where + (InstructionPointerIndex << 2) put: (self integerObjectOf: (((LiteralStart + (self literalCountOfHeader: methodHeader)) * 4) + 1)). self longAt: where + (StackPointerIndex << 2) put: (self integerObjectOf: tempCount). self longAt: where + (MethodIndex << 2) put: newMethod. "Copy the reciever and arguments..." argCount2 _ argumentCount. 0 to: argCount2 do: [:i | self longAt: where + ((ReceiverIndex+i) << 2) put: (self internalStackValue: argCount2-i)]. "clear remaining temps to nil in case it has been recycled" methodHeader _ nilObj. "methodHeader here used just as faster (register?) temp" argCount2+1+ReceiverIndex to: tempCount+ReceiverIndex do: [:i | self longAt: where + (i << 2) put: methodHeader]. self internalPop: argCount2 + 1. reclaimableContextCount _ reclaimableContextCount + 1. self internalNewActiveContext: newContext. ! ! !Interpreter methodsFor: 'control primitives' stamp: 'JMM 12/29/2001 16:22'! primitiveBlockCopy | context methodContext contextSize newContext initialIP where | context _ self stackValue: 1. (self isIntegerObject: (self fetchPointer: MethodIndex ofObject: context)) ifTrue: ["context is a block; get the context of its enclosing method" methodContext _ self fetchPointer: HomeIndex ofObject: context] ifFalse: [methodContext _ context]. contextSize _ self sizeBitsOf: methodContext. "in bytes, including header" context _ nil. "context is no longer needed and is not preserved across allocation" "remap methodContext in case GC happens during allocation" self pushRemappableOop: methodContext. newContext _ self instantiateContext: (self splObj: ClassBlockContext) sizeInBytes: contextSize. methodContext _ self popRemappableOop. initialIP _ self integerObjectOf: instructionPointer - method. "Was instructionPointer + 3, but now it's greater by methodOop + 4 (headerSize) and less by 1 due to preIncrement" "Assume: have just allocated a new context; it must be young. Thus, can use uncheck stores. See the comment in fetchContextRegisters." where _ (self cCoerce: newContext to: 'char *') + BaseHeaderSize. self longAt: where + (InitialIPIndex << 2) put: initialIP. self longAt: where + (InstructionPointerIndex << 2) put: initialIP. self longAt: where + (StackPointerIndex << 2) put: (self integerObjectOf: 0). self longAt: where + (BlockArgumentCountIndex << 2) put: (self stackValue: 0). self longAt: where + (HomeIndex << 2) put: methodContext. self longAt: where + (SenderIndex << 2) put: nilObj. self pop: 2 thenPush: newContext.! ! -- -- =========================================================================== John M. McIntosh <johnmci(a)smalltalkconsulting.com> 1-800-477-2659 Corporate Smalltalk Consulting Ltd. http://www.smalltalkconsulting.com ===========================================================================

1 0

[FIX] Mpeg3Plugin (fix to fix)
by John M McIntosh 02 Jan '02

02 Jan '02

If you squint you can see how some friendly feature removed an instance var at the last second... 'From Squeak3.2alpha of 1 November 2001 [latest update: #4599] on 1 January 2002 at 10:15:57 pm'! !Mpeg3Plugin methodsFor: 'primitives' stamp: 'JMM 1/1/2002 22:15'! primitiveMPEG3CheckSig: path | result sz storage | "int mpeg3_check_sig(char *path)" self var: #storage declareC: 'char storage[1024] '. self primitive: 'primitiveMPEG3CheckSig' parameters: #(String). sz _ interpreterProxy byteSizeOf: path cPtrAsOop. self cCode: 'sqFilenameFromStringOpen(storage, path, sz)'. self cCode: 'result = mpeg3_check_sig(storage)'. ^result asOop: Boolean ! ! !Mpeg3Plugin methodsFor: 'primitives' stamp: 'JMM 12/27/2001 15:40'! primitiveMPEG3Open: path | mpeg3Oop index sz storage | "mpeg3_t* mpeg3_open(char *path)" self var: #index declareC: 'mpeg3_t ** index'. self var: #storage declareC: 'char storage[1024]'. self primitive: 'primitiveMPEG3Open' parameters: #(String). sz _ interpreterProxy byteSizeOf: path cPtrAsOop. self cCode: 'sqFilenameFromStringOpen(storage, path, sz)'. mpeg3Oop _ interpreterProxy instantiateClass: interpreterProxy classByteArray indexableSize: 4. index _ self cCoerce: (interpreterProxy firstIndexableField: mpeg3Oop) to: 'mpeg3_t **'. self cCode: '*index = mpeg3_open(storage); makeFileEntry(*index)'. ^mpeg3Oop. ! ! -- -- =========================================================================== John M. McIntosh <johnmci(a)smalltalkconsulting.com> 1-800-477-2659 Corporate Smalltalk Consulting Ltd. http://www.smalltalkconsulting.com ===========================================================================

1 0

Update on the Squeak-material on my site
by Maarten Maartensz 02 Jan '02

02 Jan '02

L.S. It may interest some to know that I have given Squeak its own directory on my site and have extended the html- versions of my user manual (in progress). The relevant direct link is: http://www.xs4all.nl/~maartens/squeak/squeak_index.htm It contains almost 700 Kb of Squeak-related files on the moment, nearly all of which is in a transitional stage (and so is expected to be extended or rewritten in the none too far future, to end up in a Squeak Project). Having seen my own site recently on a local internet-cafe, that did not display part of my site as it should (probably due to a combination of factors including MS buggery): - my site should display in Arial 12 point bold - and is best read in 1024*768 resolution The letter I prefer is fairly large because of my none too good eyes and health (and the 1500 files I maintain on my site), and you can easily get rid of it by means of your browser if it displeases you. (Also, for some learned academics of acute logical awareness: If I talk about Squeak on my website the text really IS in html and - alas! - not in Squeak. It is true this is a logically subtle point, for some.) Finally, I am on the digest-form of the developers-list only, since I have no really useful code to contribute, and since I find it none too constructive to discuss philosophy etc. (including my own) on this developers list. I wish everyone on the list a happy, healthy and peaceful 2002! Regards, Maarten. -------------------------------------------------------- Maarten Maartensz. Website in Amsterdam about philosophy, logic, M.E. (Myalgic Encephalomyelitis) and much more: http://www.xs4all.nl/~maartens/ more than 25 MB of stimulating and original ideas. ---------------------------------------------------------

2 1

Block Closures coming soon
by Anthony Hannan 02 Jan '02

02 Jan '02

Happy New Year, A few weeks ago, I said I would be releasing my Block Closure work by the new year. Well... please allow me two more weeks. I've been making some changes, and I want to make sure its robust and documented before releasing it. Cheers, Anthony

4 3

Re: [BUGS] Looking for plugins (Linux)
by Lex Spoon 02 Jan '02

02 Jan '02

"Joshua 'Schwa' Gargus" <schwa(a)cc.gatech.edu> wrote: > On Fri, Dec 28, 2001 at 03:10:03PM -0800, Tim Rowledge wrote: > > > > > > Joshua 'Schwa' Gargus wrote: > > > > > > It seems to me that looking for a loadable module in the same > > > directory that the VM exists in is a very sensible behavior. For some > > > reason, the Linux VM does not do this. > > According to the code in > > platforms/unix/vm/sqUnixExternalPrims.c>ioLoadModule(), it looks in > > current dir, /etc/ld.so.cache, /usr/lib, /lib, SQ_LIBDIR, so it really > > should be working for you. > > I guess I haven't paid the necessary attention to SQ_LIBDIR. Is it set > via an argument to configure? It's set by the configure script, to the location that "make install" will install them for you if you only allow it. :) You can use --prefix to influence where the files will be installed; for example, "configure --prefix=$HOME" can be used to install into your own home directory. It's something like /usr/local/lib/squeak/3.1a-4137/ . An environment variable would seem pretty reasonable, too, but I don't think there is one. SQUEAK_PLUGIN_DIR might be a good name for it, to go along with the other environment variables. > > > I usually end up with the plugins in the same > > directory as the cwd. > > Really? You have your image, filed-out changesets, and other > miscellany all in the same directory with .so plugin files? I think > that I'd find that quite cluttered. I guess that you primarily do > this on your Acorn, and probably have a nice GUI file browser that > separates them based on their file type. Is this your Linux usage > pattern as well? I still haven't found a GUI for Linux that pisses me > off less than it helps me out. This is strange to me, too; you can simply use "make install" and everything is happy. Don't squelch your VM -- let it express itself in all the glorious directories that it wants to. -Lex

1 0

getting really weird: Player39 in PointerFinder
by Ross Boylan 01 Jan '02

01 Jan '02

In tracking down the cause of grossly excessive copies of some objects in my image, I've found that some Player classes are holding onto them. Can I just delete the classes? What is the safe way to get rid of these things? Extra credit to anyone who can explain to me how this could possibly have happened. I haven't used the tile scripting stuff much (Player is part of that, right?). I have, among other follies, published my project repeatedly, often abortively, and I have created copies by dragging things into ProjectViews. I also tried using the EventRecorder a bit to capture some activity. I have no reason to think any of these things would account for what I see, except that I can't think of anything else :) Here's what pointer finder shows me (by hand, since I don't know any other way to get the info here): #Player39 -> Player39 class scripts: IdentityDictionary #script1 -> UniclassScript currentScriptEditor: ScriptEditorMorph playerScripted: Player39 costume: TransformationMorph owner: SphereWindow "first sign of my stuff" myChildHolder: AlignmentMorph etc SphereWindow is a morph. Some of my classes mess around with the text or string morph editor--but only in subclasses. The scripts class instance variable (defined in superclass Player) seems magically connected to methods showing in the scripts protocol of the Player39 instance side. The scripts do little heading changes (which kind of fits with the EventRecorder theory). When I browse versions of the methods I get no info on creator or creation time. Player39 is in category Morphic-UserObjects

3 3

Re: [ANN][VM][SF] my Unix patches integrated into SF layout
by Joern Eyrich 01 Jan '02

01 Jan '02

> I would be perfectly happy for these to get folded into the communal > source forge site, if people want to go that route. Great! I'm all for integrating it. But it doesn't work for me (yet). - First, I expected to have to run the copied version of configure in the build directory as with the current SF tree; but ok, you documented that one should use the one in "platforms/unix/misc" - Then, I see that all the *.c files are compiled from the "platforms" hierarchy, and not in the copied hierarchy under the build dir; here it really doesn't make any sense to copy everything first Ok, but these are cosmetic/a matter of taste. - the compile of gnu-interp.c fails with a host of gnu-interp.c:5152: parse error before `:' error messages. the relevant source is CASE(0:) which is expanded as (sqGnu.h) #define CASE(N) case N: _##N: which is obviously problematic. It's generated from (interp.c) case 0: by (gnuify) (stage == 2) && /^ case / { print " CASE(" substr($NF,0,length($NF)) ")"; next; } so the result is what I would have expected. but as it has obviously worked for you, I wonder what I have done wrong/differently. Any ideas? i looked for inspiration in the current SF source, which has (gnuify) (stage == 2) && /^ case / { print " CASE(" $3 ")"; next; } here, I am completely at a loss how it could have worked before, because in my understanding the original line only has 2 fields... can anybody help? Thanks, Joern

1 0

Jump to page: