October 2008 - Pharo-dev - lists.pharo.org

[Pharo-project] Pharo image size: 10.9MB
by Adrian Lienhard Oct. 12, 2008

Oct. 12, 2008

I've been looking into how the current 13.4MB a Pharo image uses are spent... I expected that at least half of it is used by classes (compiled methods). But this is not the case. Actually, only about 1/4th are classes. As it turned out, almost the same amount of memory is used by forms (actually bitmaps)! If you want to know which are the largest 10 forms try this: (Form allInstances asSortedCollection: [ :e1 :e2 | e1 bits size >= e2 bits size ]) from: 1 to: 10 do: [ :each | each asMorph openInWorld ] ...among which for example are 4 different color pickers. In total there are around 450 forms, most from EToys. I've manually traced down where the larger forms come from and have cleaned up those that were possible (many are only referenced by caches). The number of forms is at 320 now. In the process I produced the script below, which among other stuff removes all but the default font. Also, it actually removes the flaps from the system instead of only hiding them. Here a history of image sizes: Squeak 3.9: 15.6 MB Pharo 10091: 13.4MB (before cleanup) Pharo 10091: 10.9MB (after cleanup) Main conclusion: the system needs cleanup not only of the code (where the mess is more directly visible), but also of the object memory. Furthermore, I expect removing EToys will have an nice impact on the image size. Also removing Sound will release some memory because of sound samples. If anybody has suggestions to further reduce the image size (<10MB would be a nice target for the first release...), please let me know. Cheers, Adrian Preferences setDefaultFonts: #( (setSystemFontTo: 'Accuny' 10) (setListFontTo: 'Accuny' 10) (setMenuFontTo: 'Accuny' 10) (setFlapsFontTo: 'Accuny' 12) (setEToysFontTo: 'Accuny' 10) (setPaintBoxButtonFontTo: 'Accuny' 10) (setWindowTitleFontTo: 'Accuny' 15) (setBalloonHelpFontTo: 'Accuny' 10) (setCodeFontTo: 'Accuny' 10) (setButtonFontTo: 'Accuny' 8)). TextConstants associations do: [ :association | ((association value isKindOf: TextStyle) and: [ association key ~= #Accuny ] and: [ (TextStyle defaultFamilyNames includes: association key) not ]) ifTrue: [ TextConstants removeKey: association key ]]. TextConstants at: #DefaultFixedTextStyle put: (TextConstants at: #Accuny). (Flaps classPool at: #SharedFlapTabs) removeAllSuchThat: [ :each | (#(Tools Pharo) includes: each labelString) not ]. NaturalLanguageFormTranslator classPool at: #CachedTranslations put: nil. NaturalLanguageTranslator resetCaches. PartsBin clearThumbnailCache. PaintBoxMorph initialize. PaintBoxMorph classPool at: #ColorChart put: nil; at: #Prototype put: nil. Utilities classPool at: #ScrapsBook put: nil. Project allInstancesDo: [ :each | each setThumbnail: nil ]. InternalThreadNavigationMorph clearThumbnailCache. ScriptingSystem stripGraphicsForExternalRelease. ProjectHistory currentHistory initialize. Utilities zapUpdateDownloader. Smalltalk forgetDoIts. Behavior flushObsoleteSubclasses. CommandHistory allInstancesDo: #initialize. ScriptLoader new cleaningCS. RequiredSelectors initialize. ProvidedSelectors initialize. LocalSends initialize. SendCaches initializeAllInstances. Utilities cleanseOtherworldlySteppers. Smalltalk organization removeEmptyCategories. SystemBrowser removeObsolete. TheWorldMenu removeObsolete. AppRegistry removeObsolete. FileServices removeObsolete. SystemNavigation default obsoleteBehaviors. MCFileBasedRepository flushAllCaches. MCDefinition clearInstances. Undeclared removeUnreferencedKeys. Smalltalk flushClassNameCache. 3 timesRepeat: [ Smalltalk garbageCollect. Symbol compactSymbolTable ]. Set rehashAllSets. Smalltalk garbageCollect. ___________________ http://www.adrian-lienhard.ch/

4 10

[Pharo-project] SLICE-Mantis-KernelNumber-Fix-Mining-nice.1.mcz
by nicolas cellier Oct. 11, 2008

Oct. 11, 2008

I push here some more or less important fixes. These are limited to one or two method per fix and should not take very long to review. Cheers Nicolas This gathers several Tests and Fixes from Mantis for Kernel-Number ------------------------------------------------------------ http://bugs.squeak.org/view.php?id=6781 0.0 raisedTo: -1.0 unexpected result The patch correct Number >> #raisedTo: and uses [^self class one] so that Float >> #raisedTo; can be removed (99% identical code). ------------------------------------------------------------ http://bugs.squeak.org/view.php?id=6729 -Infinity closeTo: Infinity self deny: (Float infinity negated closeTo: Float infinity) ------------------------------------------------------------ http://bugs.squeak.org/view.php?id=6695 Float storeStringBase: does not store the radix self assert: (20.0 storeStringBase: 16) = '16r14.0' compare to this: (20 storeStringBase: 16) Note: in base 10, radix is not prepended (same as Integer). ------------------------------------------------------------ http://bugs.squeak.org/view.php?id=6697 Number>>storeOn:base: default implementation is wrong Default implementation does omit the radix prefix when it should not! Change it to subclassResponsibility, since every subclass will have a different one... ------------------------------------------------------------ http://bugs.squeak.org/view.php?id=6990 0.0 significandAsInteger is too big ------------------------------------------------------------ http://bugs.squeak.org/view.php?id=7114 SmallInteger minVal highBitOfMagnitude Try it, it should not raise an Error... ------------------------------------------------------------ http://bugs.squeak.org/view.php?id=5640 additional tests and fixes for Number>>printShowingDecimalPlaces: http://bugs.squeak.org/view.php?id=7028 printShowingDecimalPlaces: is broken in 3.10 Try (-0.9 printShowingDecimalPlaces: 1) for example... Where is the minus sign? Fix included in 3.10 is incorrect and not complete ------------------------------------------------------------

1 0

[Pharo-project] About trait mop
by StÃ©phane Ducasse Oct. 11, 2008

Oct. 11, 2008

Hi I have been working on a new API to access class, traits, selector for traits I wrote some tests in the PharoTaskForces/Traits-MOPTests package Now in essence we have - The selector of a compiled method should be its name. An aliased method should have the name of its alias name. - The class of a compiled method is the class that contains it. A compiled method cannot be shared. - The origin of a compiledMethod is its defining class or trait. - The reachHome of a compiled method is class or traits that you changed (and not override) when you want to change a method in the flat view. T1 c c1 T2 c c3 uses: T1 Class A foo uses: T2 testSelectorT11T2 "self debug: #testSelectorT11T2" "The selector of a compiled method should be its name. An aliased method should have the name of its alias name." self assert: (AUsingTrait2>>#foo) selector = #foo. self assert: (Trait1>>#c) selector = #c. self assert: (Trait2>>#c) selector = #c. self assert: (AUsingTrait2>>#c) selector = #c. self assert: (AUsingTrait2>>#c3) selector = #c2. self assert: (BUsingTrait1>>#c) selector = #c. testClassT11T2 "The class of a compiled method is the class that contains it. A compiled method cannot be shared." self assert: (AUsingTrait2>>#foo) methodClass = AUsingTrait2. self assert: (AUsingTrait2>>#c) methodClass = AUsingTrait2. self assert: (AUsingTrait1>>#c3) methodClass = AUsingTrait2. self assert: (AUsingTrait2>>#c1) methodClass = AUsingTrait2. self assert: (BUsingTrait1>>#c) methodClass = BUsingTrait1. self should: (Trait1>>#c1) methodClass = nil. "T1 or nil" self should: (Trait2>>#c2) methodClass = nil. "T2 or nil" self should: (Trait2>>#c1) methodClass = nil. "T2 or nil or T1" testReachHome "self debug: #testReachHome" " The reachHome is the method that you changed (and not override) when you want to change a method in the flat view." self assert: (AUsingTrait2>>#foo) reachHome = AUsingTrait2. self assert: (AUsingTrait2>>#c3) reachHome = Trait2. self assert: (AUsingTrait2>>#c1) reachHome = Trait1. self assert: (AUsingTrait2>>#c) reachHome = Trait2. self assert: (Trait11>>#c) reachHome = Trait11. self assert: (Trait2>>#c) reachHome = Trait2. self assert: (Trait2>>#c1) reachHome = Trait11. self assert: (AUsingTrait1>>#c2) reachHome = AUsingTrait1. testOrigin "self debug: #testClass" "The origin of a compiledMethod is its defining class or trait." self assert: (AUsingTrait2>>#foo) origin = AUsingTrait2. self assert: (AUsingTrait2>>#c3) origin = Trait2. self assert: (AUsingTrait2>>#c1) origin = Trait11. self assert: (AUsingTrait2>>#c) origin = Trait11. self assert: (BUsingTrait1>>#c) origin = Trait1. self assert: (Trait2>>#c3) origin = Trait2. self assert: (Trait11>>#c1) origin = Trait11. self assert: (Trait2>>#c1) origin = Trait11. "since it introduced the alias it is its origin" self assert: (AUsingTrait1>>#c2) origin = AUsingTrait1. testTraitT11T2 "it seems to me that we do not need trait: but we need origin and reachHome" "(A>>foo) trait -> nil (A>>c3) trait -> T2 (A>>c) trait -> T2 (A>>c1) trait -> T2 ***** since it is flat (T1>>c1) trait -> ???? T1 or nil (T2>>c1) trait (T2>>c2) trait -> ???? T2 or nil" self assert: (AUsingTrait2>>#foo) trait isNil. self assert: (AUsingTrait2>>#c3) trait = Trait2

4 7

[Pharo-project] #2
by StÃ©phane Ducasse Oct. 11, 2008

Oct. 11, 2008

Apparently #2 value is a number? but the parser parses it. Stef

3 9

[Pharo-project] Monticello 1.6 or 2.0?
by Janko MivÅ¡ek Oct. 10, 2008

Oct. 10, 2008

Hello, We have now three Monticellos: 1, 1.6 and 2.0. What is the Pharo plans here, to go to 1.6 then to 2.0? How to migrate old repositories like SqueakSource? When to migrate? How to persuade all Squeakers to upgrade MC in their images? That questions which are raising in my mind for a while... Janko -- Janko MivÅ¡ek AIDA/Web Smalltalk Web Application Server http://www.aidaweb.si

2 1

[Pharo-project] Installer refactor
by Matthew Fulmer Oct. 10, 2008

Oct. 10, 2008

I spent most of the day yesterday refactoring Installer. I split it up into 14 classes, so now it should be a bit easier to understand. I'll fix badly categorized methods and add some comments. -- Matthew Fulmer -- http://mtfulmer.wordpress.com/

3 7

[Pharo-project] problem with cursor
by StÃ©phane Ducasse Oct. 10, 2008

Oct. 10, 2008

I noticed that the cursor is quite slow at getting where I point. I often have to click two times so that it appears under my mouse. Does anybody else experiencing the same. I'm using OB + packageBrowser + for a moment I thought it would be related to the demo mode font When I click sometimes the cursor simply does not move to the point. Stef

2 2

[Pharo-project] [Fwd: gmane.comp.lang.smalltalk.sapphire.devel: Authorization required]
by nicolas cellier Oct. 10, 2008

Oct. 10, 2008

One more thing strange: when i post a reply through http://comments.gmane.org/gmane.comp.lang.smalltalk.pharo.devel/2360 then i get a confirmation for the old list name: gmane.comp.lang.smalltalk.sapphire.devel -------- Message original -------- From: - Fri Oct 10 17:46:41 2008 X-Account-Key: account4 X-UIDL: 48ef7f72-0004149f-32fff902 X-Mozilla-Status: 0003 X-Mozilla-Status2: 00000000 X-Mozilla-Keys: Received: from localhost (localhost [127.0.0.1]) by mailrecv01.ifrance.com (Postfix) with ESMTP id AAEDD1E8DD3 for <ncellier(a)ifrance.com>; Fri, 10 Oct 2008 17:30:02 +0200 (CEST) X-Spam-Score: 2 X-Spam-Level: ** X-Spam-Status: No, score=2 tagged_above=-999 required=6.2 tests=[FROM_LOCAL_HEX=2] Received: from mailrecv01.ifrance.com ([127.0.0.1]) by localhost (mailrecv03.ifrance.com [127.0.0.1]) (amavisd-new, port 10024) with LMTP id hkQcEZn16DU2 for <ncellier(a)ifrance.com>; Fri, 10 Oct 2008 17:30:02 +0200 (CEST) X-Greylist: from auto-whitelisted by SQLgrey-1.7.3 Received: from ciao.gmane.org (main.gmane.org [80.91.229.2]) by mailrecv01.ifrance.com (Postfix) with ESMTP id 412581E8DD2 for <ncellier(a)ifrance.com>; Fri, 10 Oct 2008 17:30:02 +0200 (CEST) Received: from list by ciao.gmane.org with local (Exim 4.43) id 1KoK4Q-0003Mv-S5 for ncellier(a)ifrance.com; Fri, 10 Oct 2008 15:38:26 +0000 From: Gmane Autoauthorizer <auth-5debf6d14fd23d9c22d27f04e5f7d097(a)auth.gmane.org> To: ncellier(a)ifrance.com References: <loom.20081010T152509-580(a)post.gmane.org> Subject: gmane.comp.lang.smalltalk.sapphire.devel: Authorization required Message-Id: <E1KoK4Q-0003Mv-S5(a)ciao.gmane.org> Date: Fri, 10 Oct 2008 15:38:26 +0000 You have sent a message to be posted on the gmane.comp.lang.smalltalk.sapphire.devel newsgroup. Before the message is posted on the newsgroup, you have to confirm that you exist. Just reply to this message, and the message will be posted. You have to respond within one week. -- Your friendly autoauthorizer at Gmane.org http://gmane.org/

1 0

[Pharo-project] [ANN] 10091
by nicolas cellier Oct. 10, 2008

Oct. 10, 2008

>>/ http://bugs.squeak.org/view.php?id=7180 +6455 +1603 +1602 />>/ />>/ Patches for Interval indexOf: and includes: />/ />/ I wonder about these fixes because of http://code.google.com/p/pharo/issues/detail?id=35 />/ />/ Did you reconsider the correctness of these fixes? They seem />/ complicated and after reading through the bug reports again />/ (especially http://bugs.squeak.org/view.php?id=7180) I concluded />/ that they may introduce more problems than they fix. /I know I tried to find my way there too. Now I trusted nicolas. Nicolas are you strong on these fixes? Stef >/ />/ />/ Adrian />/ _______________________________________________ />/ Pharo-project mailing list />/ Pharo-project at lists.gforge.inria.fr <http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project> />/ http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project /> Thank you Stef. Yes/, the fix is safe. I develop here after these 3 points A)- The code was wrong ///B)- Discussions were too long, but the fixes are very easy //C)- library could be enhanced //Adrian, hope I will convince you (also, read the code). /// Cheers Nicolas A) ///The code was wrong// /Prior this patch: ///- #includes: implementation were WRONG. - #indexOf: was missing, falling back to naive super slow implementation /// Try these two tests in any Squeak image of your choice (3.7 to: 3.10): // "There is no reason that the first is true while the second is false" //{(Float pi to: 4*Float pi by: Float pi) includes: Float pi*2 + 1.0e-11. (Float pi to: 4*Float pi by: Float pi) includes: Float pi*2 - 1.0e-11}. / /"There is no reason for this timeToRun to grow" (1 to: 7) collect: [:i | [(1 to: 10000000) indexOf: (10 raisedTo: i)] timeToRun] ///// ////B)- the fixes are very easy//// I simply implemented indexOf: with same algorithm as #includes: (corrected!). Then removed includes: because super includes: is implemented sending indexOf: Then I removed valueIncludes: because includes: was sole sender. You see, two methods removed, one fix added: no revolution / /C)- library could be enhanced /What I would prefer is: - implement a method with an argument so that user can choose comparison tolerance - expand this fuzzy inclusion test to any other collection Maybe something like: includes: aNumber moreOrLess: aTolerance. Then Interval includes: could eventually use a default tolerance. But the obvious bugs must first be corrected... /

3 2

[Pharo-project] Monticello: an idealist question
by Bill Schwab Oct. 10, 2008

Oct. 10, 2008

Colin, Among many useful things in Dolphin is an ability to assign multiple categories to any given method; Squeak and Pharo allow only one category, and it promptly gets chewed up by PackageInfo. Multiple categories would be a nice addition to Squeak and Pharo. How hard would it be to change Monticello to take its input not from PackageInfo and its (pardon me) abuse of category names, but from reifications of the categories? The new objects would form the basis for packaging system, and free method categories to do what they do best: aid in browsing and understanding code. In transition, a future version of Monticello could create the needed objects using the same rules as PackageInfo (or any time it is instructed to do so). Comments? Bill Wilhelm K. Schwab, Ph.D. University of Florida Department of Anesthesiology PO Box 100254 Gainesville, FL 32610-0254 Email: bschwab(a)anest.ufl.edu Tel: (352) 273-6785 FAX: (352) 392-7029

4 3