[UPDATES] 52 for Squeak3.7alpha, advance to 3.7beta

stéphane ducasse ducasse at iam.unibe.ch
Thu Apr 1 08:15:49 UTC 2004


thanks for your time doug I love to see all the fix and enh. This is 
always a so good feeling.

Stef

On 1 avr. 04, at 09:30, dway at mailcan.com wrote:

>
>
> Here is the last batch of updates for 3.7alpha, plus an update to
> advance to 3.7beta.  Quite a bit of stuff here!  Now that we're in 
> beta,
> further updates for 3.7 will just be fixes (and possibly things like
> preferences tweaks, etc).  Beta is scheduled to last 6 weeks.  The 
> alpha
> stream for the next version will open then, or possibly sooner, but not
> right now.
>
> - Doug
>
>
> ------------------------------
>
> 5817SmallTypos2
>
> 5818CompilerFromSM-md -- Marcus Denker -- 8 March 2004
> This cs loads the current Compiler* from SqueakMap3
>
> 5819MsecClockValueOptmz-nk -- Ned Konz -- 8 March 2004
> Optimize Time>>millisecondClockValue by saving one call.
> Doesn't sound like much, but makes a difference since we call it so
> much. Also keeps millisecondClockValue from appearing in profiler
> output.
>
> 5820AssociationHashTest -- Marcus Denker -- 8 March 2004
> This is just associationEqualTest-brp.1.cs with
> a change to the class name: The methods are added
> to the existing AssociationTest.
> This just refactors the Equality test and adds a test for #hash"
>
> 5821devideIsZero -- Marcus Denker -- 7 March 2004
> Just a small fix from the Units package to
> use isZero in Float>>/ and SmallInteger>>/"
>
> 5822divideIsZero-fbs
>
> 5823OrganizeTheAllCategory-nk -- Ned Konz -- 11 March 2004
> This is offered as a better way to fix the problem with filing in the
> '--all--' category from the File Contents Browser.
> It makes the ClassOrganizer actually return all the selectors when 
> given
> the name of the allCategory. It also changes the definition of
> allCategory to answer a literal Symbol, rather than calling #asSymbol 
> on
> a String.
>
> 5824BlankMenuIconsPref-nk -- Ned Konz -- 10 March 2004
> I couldn't avoid the recent discussion of menu icons, so I am
> contributing this little ENH.
> This will line up all the text in menus containing icons by using blank
> icons for those items without icons.
> I believe that this leads to an easier to read menu.
> v2: handled cases where submenus have icons but parent menus don't.
> v3: removed preference. Made checkboxes line up too.
> Preferences removePreference: #useBlankMenuIcons
>
> 5825loadProjectFix-mir -- Michael Rueger -- 9 March 2004
> Fixes a problem with loading projects through the world menu."
>
> 5826PseudoClassPrettyPrint-nk -- Ned Konz -- 9 March 2004
> As reported by David Shaffer,
> If color pretty printing are enabled, if a method being viewed in the
> FileContentsBrowser includes an i-var reference which the PseudoClass
> can't find in it's scope table, you get the attached stack trace.  I'm
> not exactly sure how PseudoClass is supposed to work so I don't know if
> this is considered a bug (*) in that class but I think that the browser
> should non-pretty print the text if parsing fails.
> This CS adds instVarNames to PseudoClass to avoid this bug.
>
> 5827testChronologyHours12-brp -- Brent Pinkney -- 12 March 2004
> Fix test cases to match the implementation of DateAndTime>>#hour12.
> The original implementation has been fixed by Avi. This just makes the
> tests compliant"
>
> 5828chronologyClassComnts-brp -- Brent Pinkney -- 12 March 2004
> Add missing class comments. Self approved as per instructions on
> squeak-dev"
>
> 5829chronlgyTimeStampNow-brp -- Brent Pinkney -- 12 March 2004
> Deprecated class TimeStamp has second precision.
> After the introduction of the ANSI Chronology classes, TimeStamp is
> implemented as a subclass of DateAndTime - inadvertantly inheriting
> nano-second precision.
> This changest ensure that TimeStamp class>>#now and TimeStamp
> class>>current clear out the nanosecond field.
> Note. This means that (TimeStamp now = TimeStamp now) could be true. 
> Fix
> TimeStampTest>>#testComparing to accomodate this."
>
> 5830DeclareClass-rr -- rr -- 4 March 2004
> Adds an option when a global variable is not found,
> which is to define a new class named like it, instead
> of merely defining a global variable. Modifies a method
> and adds one.
> Example :
> NewClass new inspect.
> NewClass removeFromSystem.
>
> 5831asCommTest-ls -- Lex Spoon -- 28 October 2003
> md: V2
> 	 moved to StringTest.
> A test case for #asSmalltalkComment, a method in the browseWithComments
> enhancment."
>
> 5832DeclareInstVarFix-rr -- Romain Robbes -- 6 March 2004
> v2: fixed to not uncomment the DeclareClass changes. -- md
>        please file in DeclareClass-rr before.
> Fixes the bug in the declaration of inst vars. Previously declaring an
> instance variables
> when an unknown variable was encountered could result in the wrong iv 
> to
> be referenced if there were instance variables in some of the class'
> superclasses.
> This was visible when the method's bytecode were decompiled. 
> Recompiling
> the method would solve the problem.
> This changes only adds a line to correctly compute the index, by taking
> into account
> ivs in the superclasses, and uncomments the menu entry as it works
> correctly now.
> No test is included as this is quite hard to automate. Try it in a 
> class
> with a superclass
> which contains iv, and check the result in the decompiled view of the
> method. "
>
> 5833SimpleButtonLabeling-sge -- Steve Elkins -- 11 March 2004
> It's easy to demonstrate the 2 bugs this ChangeSet corrects.  Open the
> Squeak flap, make a duplicate of the save button, drop it on the
> desktop, and choose 'change label' from its menu.  The prompt string is
> missing a word; after entering a label, font and emphasis information
> from the previous one is lost."
>
> 5834AccunyInstall -- Ned Konz -- 15 March 2004
> Installs the text style 'Accuny' from a compressed MIME encoding in the
> postscript, originally edited by John McIntosh.
> March 15: do a better job tracking down loose and derivative fonts.
>
> 5835GenieStubs-nk -- Ned Konz -- 11 March 2004
> Version:			4
> This package supplies replacements for a few methods that will let
> Genie-aware applications run (without Genie) without change in an image
> with Genie removed (i.e. in post-5764 images).
> I recommend that this package be added to the update stream.
>
> 5836SystemDictRenameClasses-rr -- Romain Robbes -- 11 March 2004
> The event triggered by the SystemChangeNotifier had a nil class
> category, as it was fetching it after the class was renamed. This fix
> stores the category in a temp before the renaming.
>
> 5837ShortQuantitiesEndian -- Yoshiki Ohshima -- 17 March 2004
> Later fixes by:	Ned Konz
> Restoring the endianness of 16-bit quantity array object are wrong from
> a place to another.  This fixes the ones I noticed.
> With this patch, you can:
> 	
> 	* save and restore the Squeaky Mouse.
> 	* save and restore TTFontDescriptions across the platforms.
> 7 Mar (Ned Konz): fixes to make ShortIntegerArray and ShortPointArray
> work too:
> * moved restoreEndianness up to ShortIntegerArray
> * added bytesPerElement to ShortIntegerArray
> 17 March (Ned Konz):
> 	more fixes to ShortIntegerArray hierarchy
>
> 5838BitmapStreamTests -- Ned Konz -- 17 March 2004
> v5 (17 March): added a couple more tests.
> v4 (16 March): added tests with literals to verify absolute endianness
> in addition to round-tripping.
> v3 (7 March 2004): added MatrixTransform2x3
> v2 (5 July 2003): added WordArray and several kinds of storage:
> * ImageSegment on disk
> * ReferenceStream on RWBinaryOrTextStream
> * ReferenceStream on file
> * SmartRefStream on RWBinaryOrTextStream
> * SmartRefStream on file
> This is an incomplete test suite for storing and reading various word-
> and short-word subclasses of ArrayedCollection.
> It demonstrates some problems with filing in of certain kinds of 
> arrayed
> objects.
> In 3.6b-5331, I get 8 passed/6 failed/6 errors.
> I ran into problems when trying to read back the SqueakLogo flash
> character morph, after I'd done a 'save morph to disk' from its debug
> menu.
> The words within the ShortPointArrays and ShortRunArrays were reversed.
>
> 5839accurateDateTimeNow-avi -- Avi Bryant -- 21 February 2004
> This fixes DateAndTime class>>now to ignore the millisecond clock and
> simply return an accurate value with 1 second precision.  It also 
> clears
> out the class variables and #startUp/#initialize methods that were 
> being
> used to attempt to get (inaccurate) millisecond precision."
>
> 5840KCP185RespondsToCleans -- stephane ducasse -- 30 January 2004
> remove some unnecessary respondsTo invocations.
> (merged PNGReadWriter>>nextImage with 5810 -dew)
>
> 5841SMTPClientBracket-fbs -- Frank Shearer -- 23 March 2004
> Allows sender email addresses of the form
> 	""Someone"" <someone at somewhere>
> or
> 	<someone at somewhere>
> as well as
> 	someone at somewhere
> Upon further inspection (in RFC 2821), MAIL FROM takes a reverse-path.
> So MailAddressParser Does The Right Thing already. Whee. Herewith
> attached
> then a changeset with my previously-mentioned implementation, plus a
> suitably amended test case.  -- Frank
> (Updated changeset comment. -dew)
>
> 5842KCP200BehaviorSubcls -- stéphane ducasse -- 14 March 2004
> Behavior>>subclasses was not defined. Therefore a lot of the methods
> of Behavior were defined in Behavior but could only be run on
> ClassDescription instances. Breaking the principle that a class should
> not rely on its subclasses. This fix implements a slow implementation 
> of
> subclasses that can be executed on Behavior instances. Tests provided"
>
> 5843TextStyleName -- sd -- 21 February 2004
> add knowStyle without default information"
>
> 5844ANSICompatibility -- Alain Fischer -- 24 November 2002
> I modified this changeset by removing the DateAndTime and Duration
> changes.  I also removed the changes to TestRunner as I don't think 
> they
> belong here.
> Note that the description below will be incorrect in reference to these
> classes.  - Ken Causey
> original comment continues:
> This change set add ANSI compatibility to Squeak 3.2.
> Now TestRunner with ANSI tests loaded show (2777 run, 2753 passed, 24
> failed, 0 errors).
> 17 classes (8 new / 9 modified)
> New classes:
> ArithmeticError, DateAndTime, Duration, FloatingPointException,
> RuleDate, RuleIndexDate, RuleSelectionCodeDate, ScaledDecimal
> Modified Classes:
> BlockContext (added method #argumentCount)
> Character (added method #codePoint)
> Character class (added method #codePoint:)
> Float (2 methods added, 3 methods #modified)
> Float class (added method #one)
> Fraction (1 method added, 1 method modified)
> Fraction class (added method #one)
> Integer (6 methods added, 1 method modified)
> Integer class (added method #one)
> Number (7 methods added, 2 methods modified)
> Number class (modified method #readRemainderOf:from:base:withSign:)
> PositionableStream (modified method #skip:)
> String (3 methods added)
> TestRunner (2 methods modified)
> ZeroDivide is now subclass of ArithmeticError
>
> 5845BitstreamVeraSerifInstall -- Ned Konz -- 18 March 2004
> Installs the text style 'BitstreamVeraSerif' from a compressed MIME
> encoding in the postscript.
> Requires Ned's latest endianness fix as a prerequisite."
>
> 5846BitstreamVeraSansInstall -- Ned Konz -- 18 March 2004
> Installs the text style 'BitstreamVeraSans'
> from a compressed MIME encoding in the postscript."
>
> 5847BitstreamVeraSansMonoInstall -- Ned Konz -- 18 March 2004
> Installs the text style 'BitstreamVeraSansMono'
> from a compressed MIME encoding in the postscript."
>
> 5848UtilitesAuthorName-fc -- Frank Caggiano -- 12 March 2004
> Resubmitted, Ned pointd out that the original CS  was incompatable with
> a recent change in ProjectLauncher>>startUp. This SC uses the latest
> version of ProjectLauncher>>startUp.
> Makes AuthorName behave the same as AuthorInitials before it is set for
> the first time in a image.  That is it is set to the empty string as is
> AuthorInitials."
>
> 5849TextMorphSelectionFix-rr -- Romain Robbes -- 23 March 2004
> V3 : make the cursor disappear when there is no selection, and
> the TextMorph is not focused.
> V2 : provides more visual feedback by changing the selection
> color upon focus change, as suggested by Hernán Tylim.
> Makes the selected text in a TextMorph not
> disappear when the morph loses focus,
> by removing the call to releaseEditor in
> TextMorph>>keyboardFocusChange:  .
> This makes the TextMorphs' behavior more
> consistent with text panes in the host platforms,
> and makes Squeak a more consistent with itself,
> as this will allow all menus to behave according to
> the menuKeyboardControl preference, including
> the ones in a paragraph.
> Also included are the changes in PluggableTextMorph and
> StrikeFont, to allow menus to be invoked modally, and
> thus having keyboard control if the preference is set"
>
> 5850MPEGPlyrSubtitles-dgd-asm -- Alejandro Magistrello & Diego Gomez
> Deck -- 8 March 2004
> This changeset includes a lot of fixes/improvements to MPEGPlayer where
> the most important are the support for subtitles files (format .sub) 
> and
> the full-screen option.
> List of changes:
> - support for subtitles of .sub format (see comment in class
> MPEGSubtitles)
> - full-screen mode
> - the quit button is always present
> - eToys vocabulary for #videoFileName and #subtitlesFileName.  Using
> this new tiles with scripts in projects (or un book pages) in 'opening'
> mode we can create presentation with video in a multiplatform-way.
> - yellowbutton menu (usefull in full screen)
> - keyboard controls (usefull in full screen)
> - new UpdatingTextMorph class used to display the subtitles
> IMPORTANT: The instances of MPEGPlayer created before this changeset
> _will_ work including all the new stuff with the exception of the
> button-row that _will not_ get updated.
> Thanks to Alejandro Magistrello for the first version of this work!
>
> 5851DependentsSizeFix-nk -- Ned Konz -- 11 March 2004
> I noticed in the ChangeSorter that if you add a postscript, then close
> the editor window and try to remove the postscript that you can't.
> Further examination showed that there is code in 
> postscriptHasDependents
> that says
> 	self postscript dependents size > 0
> which returned true even though the dependents slots had been nil'd 
> out.
> Though counting dependents is probably questionable, this change set
> makes DependentsArray>>size report the number of actual dependents 
> (i.e.
> does not count nils).
>
> 5852HiliteCurrentFonts-laza -- Alexander at Lazarevic.de -- 25 March 2004
> - changes the Standard List Font menu so that the currently in use 
> fonts
> and point sizes are highlighted
> - changes the Standard System Font menu so that it can be pinned and
> stays up for consecutive font selections
> (removed extra unused method with 'hilight' in the selector -dew)
>
> 5853TextStyleToChangeSet-nk -- Ned Konz -- 15 March 2004
> Lets you write a TextStyle into a ChangeSet.
> Will also read a family from a directory containing .sf2 files.
> Could also work for TTCFonts, but requires endianness fixes first.
> v2 (15 Mar): more complete font replacement.
>
> 5854DeselectGrabbedMenu-spfa -- Stephane Rollandin -- 13 March 2004
> When we open a menu and grab or duplicate one of its item via the halo,
> we get an highlighted item in the hand and there is no easy way to
> deselect it without triggering its action.
> This changeset modifies two methods of MenuItemMorph so that a menu 
> item
> grabbed or duplicated gets deselected.
>
> 5855PSPointSizeFix-nk -- Ned Konz -- 25 March 2004
> Postscript output from Squeak has been in pixel units. However, we're
> outputting requests for fonts in point units, making the fonts appear
> 3/4 as big as they should on output.
> This change set fixes that problem by also requesting fonts in pixel
> units.
>
> 5856q20renameVariable-yo -- Yoshiki Ohshima -- 18 January 2004, revised
> 3/17/04
> 			
> Derived from Squeakland updates 0198renameVariable-yo.cs and
> 0203langChgFix-sw.cs.  Integrated with 3.7a/5816 by Scott Wallace 10
> March 04.
> Make the rename variable feature work.  As long as the variable is not
> used in textually coded scripts, the cases like it is referenced in the
> other player's script, multiple occurrence in more than one script, 
> etc,
> are solved.
> This update also fixes the long-standing bug that the menu-title of the
> slot-info-button could show an internal slot symbol (e.g.
> useGradientFill) rather than an external variable name (e.g.
> gradientFill).
> Some refinements and fixes added by Scott Wallace 7 March 04."
>
> 5857q21scriptNameChoices-sw -- Scott Wallace -- 21 December 2003, 11
> March 2004
> Adapted for Squeak 3.7a from Squeakland update 
> 0201scriptNameChoices-sw.
> Fixes the script-name-choice popups and arrow-actions (on script-name
> tiles in viewers and scriptors) so that...
> S  Choice list is now dynamically determined; thus even script-names
> established after the tile was created will now be offered.
> S  The default choice on a new script-name tile is now #emptyScript
> (Yoshiki's suggestion), rather than an actual user-script selector 
> which
> may or may not actually exist in a given player.
> S  Script-names for scripts that have *parameters* are no longer 
> offered
> in the pop-ups of script-name-valued tiles.
> S  The tell-all commands are bulletproofed against run-time MNU errors
> in cases where a script of the given name does not exist in a player
> being sent the request.
> S  Appropriate help-message supplied on a script-name tile."
>
> 5858q22easterEggs-sw -- Scott Wallace -- 10 March 2004, 18 March 2004
> Adapted for Squeak 3.7a from Squeakland updates 0202easterEggs-sw and
> 0205varName-sw.  There was one babel-related collision to be fixed up.
> Abolishes a few long-standing odd-ball wordings in the etoy
> user-interface.
> S  Makes the 'no help available' backstop message a little less
> whimsical.
> S  Makes the default names offered for launching new variables in a
> viewer be all of the form var1, var2, etc.
> S  Provides appropriate balloon help on symbol-list tiles."
>
> 5859q23referencesFix-sw -- Scott Wallace -- 15 March 2004
> Adopted from Squeakland update 0204referencesFix-sw
> Fixes two bugs occasionally seen with etoy tiles, as reported by 
> Michael
> Rueger:
> *  Sometimes no internal name in References was being handed out for an
> object.
> *  Sometimes an internal name that coincided with a global was being
> handed out.
> An example scenario for recreating one of the bugs:
> *  Get a Slider
> *  Rename it to Slider (sic!)
> *  Now refer to that object in the script of another object.
> *  The code in that script will be bound to the global 'Slider' object
> (a class) because a name registered in the References table was not
> handed out for this object."
>
> 5860browseObject -- Masashi Umezawa -- 11 March 2004
> Add #browse method to Object.
> The method is now on tools-browser package to minimize dependency.
> Object>>browseHierarchy delegates SystemNavigation to keep consistency.
> Simple SUnit tests are also included.
>
> 5861addHorzScrollbar37v4-sps -- Steven Swerling -- 23 March 2004
> (version changes at the bottom of this changeset preamble)
> Repost of old changeset for adding horizontal scroll bars to
> ScrollPanes. Updated and cleaned up for 3.7 (and only for 3.7). The
> scroll pane now ignores the preference for showing scrollbars only when
> needed -- this is now the default. If you don't like it for a 
> particular
> pane, you can call one of the #alwaysShow... or #hideXIndefinitely
> options for the pane (see the method category, 'accessing options').
> Calls to some obsolete methods that assume 1 scrollbar are now 
> forwarded
> to the corresponding method for the vertical scrollbar, see the
> 'obsolete' method category.
> WARNING: The last thing you need is for a bug in ScrollPane, as it 
> could
> prevent the Debugger from functioning. I don't know of any in this 
> code,
> but better safe then sorry. So consider this a beta. Please tinker
> around with your applications to make sure everything is working ok
> before saving your image.
> CAVEAT: When you file in this changeset, it will add an hScrollBar to
> all existing ScrollPanes -- the color will be off on these. When you
> open new windows, the horz scrollbar will take on the appropriate 
> color.
> Tested on a fresh 3.7 image w/ updates up to #5764.
> v1: initial release.
> v2: redirected calls to Font>>widthOfString to use new method
> #widthOfStringOrText:. If it's a string or text without any sort of
> emphasis, it returns the string width. If it's a text, it adds about 
> 20%
> to the string width and returns that. It does not calculate the precise
> text width this version, which is deferred.
> v3: Added Doug Way's fix for divide-by-zero bug on resize. Fixed
> dissappearing hScrollbar in browsers. Replaced #hiddenScrollbars
> preference with #alwaysShowVScrollbar and #alwaysShowHScrollbar, both 
> of
> which default to false.
> v4: Changed #alwaysShowVScrollbar default to true. -dew
>
> 5862MonitorComments-fbs
>
> 5863BrowserCommentUpdate-nk -- Ned Konz -- 29 March 2004
> Makes the separate comment pane update when you edit the class comment
> in the single comment pane (via the '?' button)."
>
> 5864EventSensorInterpt-nk -- Ned Konz -- 18 March 2004
> CS 5784NoEventSensorProcess-ar broke the ability to break out of a 
> tight
> UI loop.
> This CS fixes that problem by checking from time to time whether it has
> been too long since the last time that events were polled.
> It also lays the groundwork for eventual use of the input semaphore by
> Morphic by setting a flag if the input semaphore is actually being
> signaled by the VM.
>
> 5865DefaultChangeSetDir-nk -- Ned Konz -- 24 March 2004
> If you try to set the default change set directory to the full name of
> the default directory, you get a walkback.
> Also, if you try to set it to a directory whose name begins with the
> name of the default directory but is not a subdirectory, the relative
> path is calculated wrong.
> This changeset fixes both of those problems.
>
> 5866MVCEventSensorFixes-nk -- Ned Konz -- 16 March 2004
> Fixes the MVC hang upon switching between views that was introduced by
> CS 5784NoEventSensorProcess-ar.
> This was caused by waitNoButton and other tight delay loops calling
> anyButtonPressed, which was returning mouseButtons without having a
> chance to actually update that variable.
> v1 (16 March)
> v2 (17 March): more refactoring; made all tests and waits consistent.
>
> 5867DateTimeFactoredPrint-nk -- Ned Konz -- 30 March 2004
> Makes all the Kernel-Chronology-Tests test cases pass in a 5816 image.
> This requires the following prerequisites:
> 	accurateDateAndTimeNowV2-avi (21 February, Avi Bryant)
> 	testChronologyHours12-brp (12 March, Brent Pinkney)
> 	chronologyTimeStampNow-brp (12 March, Brent Pinkney)
> v2 (30 March 2004):
> - sped up DateAndTime comparisons considerably (very noticeable on 
> BFAV)
> - sped up DateAndTime creation (ticks:offset:)
> - added default initialization of Duration to 0
> - sped up Duration class>>seconds: and Duration class>>zero
> - fixed broken test cases
> - added timezone saving/restoring to test cases
> - removed linefeeds
> - removed some assumptions about strict monotonicity of Timestamps from
> test cases
>   (which is appropriate for this version of the Timestamp code, but may
> need to
>   be revisited if the Timestamp logic is changed).
> v1 (12 March 2004):
> - Factors DateAndTime>>printOn: into separate parts for easier use.
> - Omits leading space by default in printOn:
> - Adds alternative printOn:withLeadingSpace:
> - Fixes DateAndTime test cases
> - Omits recompilation of Deprecated>>defaultAction from test cases (why
> was this done?)
> - Fixes the spelling of 'Greenwich'
> - Adds a class comment to Date
> - Changes TimeZone default to report UTC rather than GMT
>
> 5868AdvanceTo37Beta -- Doug Way -- 1 April 2004
> This is the last update for version 3.7alpha.  In its postscript is an
> executable statement that will set the version forward to
> Squeak3.7beta."
>
>
>
>
> <This automated message brought to you by the UpdateIncorporationTool.>
> <See the UpdateIncorporationTool package on SqueakMap for details.>
>




More information about the Squeak-dev mailing list