Squeak-dev December 2017

squeak-dev@lists.squeakfoundation.org

34 participants
159 discussions

The Trunk: Compiler-eem.360.mcz
by commits＠source.squeak.org 24 Dec '17

24 Dec '17

Eliot Miranda uploaded a new version of Compiler to project The Trunk: http://source.squeak.org/trunk/Compiler-eem.360.mcz ==================== Summary ==================== Name: Compiler-eem.360 Author: eem Time: 24 December 2017, 11:17:18.883597 am UUID: 28fc60ef-5fe3-4e38-a15c-e266f8aa3f72 Ancestors: Compiler-mt.359 Fix the Decompiler to remove the nil node that it mistakenly generates following a returning if with a nil else. As a side-effect provide ParseNode>>nodesWithPreceedingStatementsDo:. Recategorize the BlockNode>>statements accessors. Fix a typo. Remove a trio of methods that are now in EToys. =============== Diff against Compiler-mt.359 =============== Item was changed: + ----- Method: BlockNode>>statements (in category 'accessing') ----- - ----- Method: BlockNode>>statements (in category 'equation translation') ----- statements ^statements! Item was changed: + ----- Method: BlockNode>>statements: (in category 'accessing') ----- - ----- Method: BlockNode>>statements: (in category 'equation translation') ----- statements: val statements := val! Item was removed: - ----- Method: MessageNode>>morphFromKeywords:arguments:on:indent: (in category 'tiles') ----- - morphFromKeywords: key arguments: args on: parent indent: ignored - - ^parent - messageNode: self - receiver: receiver - selector: selector - keywords: key - arguments: args - ! Item was added: + ----- Method: MethodNode>>preenNilNodeFollowingNilIfNode: (in category 'converting-private') ----- + preenNilNodeFollowingNilIfNode: aNilIfMessageNode + self nodesDo: + [:node| | statements indices | + (node isBlockNode + and: [(statements := node statements) includes: aNilIfMessageNode]) ifTrue: + [indices := (2 to: statements size) reject: + [:i| + (statements at: i) == NodeNil + and: [(statements at: i - 1) isNilIf]]. + node statements: (({1}, indices) collect: [:i| statements at: i])]]! Item was changed: ----- Method: MethodNode>>preenTempsConflictingWithBlockNode: (in category 'converting-private') ----- preenTempsConflictingWithBlockNode: temps + "Push temps that conflict with other blocks down into their narrowest enclosing block scope." - "Push temps that conflict with other bocks down into their narrowest enclosing block scope." temps do: [:tempVar| (self accept: (NarrowerVariableScopeFinder new ofVariable: tempVar)) ifNotNil: [:enclosingScope | self assert: enclosingScope isBlockNode. self nodesDo: [:node| ((node == self or: [node isBlockNode]) and: [node temporaries includes: tempVar]) ifTrue: [node temporaries: (node temporaries copyWithout: tempVar)]]. enclosingScope temporaries: enclosingScope temporaries, { tempVar }]]! Item was changed: ----- Method: MethodNode>>preenableNodes (in category 'converting-private') ----- preenableNodes "Answer a Dictionary from node or sequence of nodes to preen method selector for nodes in the tree that require post-processing after either a format or a decompile. Such issues are the variable for an ifNotNil: which is local to the ifNotNil: block but, due to the inlining of ifNotNil: appears to be declared at the outer level, and, similarly, a temporary variable that conflicts with one of the same name in a block when, were the variable declared local to some inlined block it would no longer conflict. The resulting dictionary is used to perform the value with the key (node or array) as argument to preen the tree." | preenableNodes priorBlocks priorVariables | preenableNodes := Dictionary new. priorBlocks := OrderedCollection new. priorVariables := Set new. + self nodesWithPreceedingStatementsDo: + [:node :preceedingStatementOrNil| | variable temps | - self nodesDo: - [:node| | variable temps | (node isMessageNode and: [node macroPrinter == #printIfNilNotNil:indent: and: [node receiver isMessageNode and: [node receiver selector key == #== and: [node receiver receiver isAssignmentNode and: [(variable := node receiver receiver variable) isTemp and: [variable isRemote not and: [variable isOnlySubnodeOf: node in: self]]]]]]]) ifTrue: [preenableNodes at: node put: #preenIfNotNilNode:. priorVariables add: variable]. node isBlockNode ifTrue: [temps := OrderedCollection new. node temporaries do: [:temp| priorBlocks do: [:aBlock| aBlock temporaries do: [:priorTemp| (priorVariables includes: priorTemp) ifFalse: [priorTemp key = temp key ifTrue: [temps addLast: priorTemp]]]]]. temps isEmpty ifFalse: [preenableNodes at: temps put: #preenTempsConflictingWithBlockNode:]. + priorBlocks addLast: node]. + (node == NodeNil + and: [preceedingStatementOrNil notNil + and: [preceedingStatementOrNil isMessageNode + and: [preceedingStatementOrNil isNilIf]]]) ifTrue: + [preenableNodes at: preceedingStatementOrNil put: #preenNilNodeFollowingNilIfNode:]]. - priorBlocks addLast: node]]. ^preenableNodes! Item was added: + ----- Method: ParseNode>>nodesWithPreceedingStatementsDo: (in category 'visiting') ----- + nodesWithPreceedingStatementsDo: aBinaryBlock + self accept: (ParseNodeWithPreceedingStatementEnumerator ofBlock: aBinaryBlock)! Item was added: + ParseNodeEnumerator subclass: #ParseNodeWithPreceedingStatementEnumerator + instanceVariableNames: 'preceedingStatement' + classVariableNames: '' + poolDictionaries: '' + category: 'Compiler-Support'! Item was added: + ----- Method: ParseNodeWithPreceedingStatementEnumerator>>ofBlock: (in category 'initialize-release') ----- + ofBlock: aBlock + "N.B. This enumerator visits a node before any of the node's children. + Hence, when enumewrating statements in a block, we can ensure that + the second argument to the block, the preceeding statement, is non-nil + only for top-level statements in the block by nilling out preceedingStatement + once the block is evaluated. Perhaps stronger would be to capture its value + in a temporary and nil it before evaluating, but this is good enough." + theBlock := [:node| + aBlock value: node value: preceedingStatement. + preceedingStatement := nil]! Item was added: + ----- Method: ParseNodeWithPreceedingStatementEnumerator>>ofBlock:select: (in category 'initialize-release') ----- + ofBlock: aBlock select: aSelectBlock + self ofBlock: aBlock. + theSelectBlock := aSelectBlock! Item was added: + ----- Method: ParseNodeWithPreceedingStatementEnumerator>>visitBlockNode: (in category 'visiting') ----- + visitBlockNode: aBlockNode + | savedPreceedingStatement | + (theSelectBlock isNil or: [theSelectBlock value: aBlockNode]) ifFalse: + [^nil]. + theBlock value: aBlockNode. + savedPreceedingStatement := preceedingStatement. + preceedingStatement := nil. + [aBlockNode statements do: + [:statement| + statement accept: self. + preceedingStatement := statement]] ensure: + [preceedingStatement := savedPreceedingStatement]! Item was removed: - ----- Method: VariableNode>>currentValueIn: (in category 'tiles') ----- - currentValueIn: aContext - - aContext ifNil: [^nil]. - ^((self variableGetterBlockIn: aContext) ifNil: [^nil]) value printString - - - ! Item was removed: - ----- Method: VariableNode>>variableGetterBlockIn: (in category 'tiles') ----- - variableGetterBlockIn: aContext - - | temps tempIndex ivars | - - (self type = 4 and: [self key isVariableBinding]) ifTrue: [ - ^[self key value] - ]. - aContext ifNil: [^nil]. - self isSelfPseudoVariable ifTrue: [^[aContext receiver]]. - self type = 1 ifTrue: [ - ivars := aContext receiver class allInstVarNames. - tempIndex := ivars indexOf: self name ifAbsent: [^nil]. - ^[aContext receiver instVarAt: tempIndex] - ]. - self type = 2 ifTrue: [ - temps := aContext tempNames. - tempIndex := temps indexOf: self name ifAbsent: [^nil]. - ^[aContext tempAt: tempIndex] - ]. - ^nil - !

1 0

FileChooserDialog usage?
by David T. Lewis 24 Dec '17

24 Dec '17

FileChooserDialog class>>openOn: works quite nicely, but the example given in the class comment leads to message not understood #opneOn:pattern:message: Can we get the class comment updated with the right expected usage? A class comment for DirectoryChooserDialog would be good too. I do like the actual file chooser/saver dialogs, they work nicely and are easy to use. I have a questionabout FileAbstractSelectionDialog>>suffixList:. Not all operating systems and file systems make use of the unixy 'name.suffix' convention. For example OpenVMS uses [Directory]FileName.Extension;Version where the trailing file version number would not fit nicely. I realize that nobody is using VMS for Squeak at the moment, so this is just an example. My question is, would it be better to use the actual match patterns (such as '*.txt') rather trying to construct the match pattern in suffixList: ? Dave

2 5

The Trunk: Tests-eem.389.mcz
by commits＠source.squeak.org 23 Dec '17

23 Dec '17

Eliot Miranda uploaded a new version of Tests to project The Trunk: http://source.squeak.org/trunk/Tests-eem.389.mcz ==================== Summary ==================== Name: Tests-eem.389 Author: eem Time: 23 December 2017, 2:56:56.034589 pm UUID: 2a2295d5-c146-4bd0-8082-b57e86c617a4 Ancestors: Tests-eem.388 Reenable the ImageSegmentTest if the VM is mature enough. =============== Diff against Tests-eem.388 =============== Item was changed: ----- Method: ImageSegmentTest class>>testSelectors (in category 'Accessing') ----- testSelectors + (Smalltalk isRunningSpur + and: [Smalltalk vmVMMakerVersion < 2295]) ifTrue: + "The ImageSegment Test is known to not work on older Spur VMs + with the prospect of crashing. #expectedFailures does not cut it here; + don't even try to run them" + [^#()]. + ^super testSelectors! - Smalltalk isRunningSpur ifTrue: [ - "The ImageSegment Test is known to not work on SPUR VMs - with the prospect of crashing. #expectedFailure does not cut it here, - don't even try to run them" - ^ #()]. - ^ super testSelectors!

1 0

The Trunk: System-eem.992.mcz
by commits＠source.squeak.org 23 Dec '17

23 Dec '17

Eliot Miranda uploaded a new version of System to project The Trunk: http://source.squeak.org/trunk/System-eem.992.mcz ==================== Summary ==================== Name: System-eem.992 Author: eem Time: 23 December 2017, 2:55:18.856019 pm UUID: 55484fa8-616c-49ea-aa18-c8ba79a76369 Ancestors: System-tpr.991 Provide an accessor to the VMMaker version number from which the VM was generated. =============== Diff against System-tpr.991 =============== Item was added: + ----- Method: SmalltalkImage>>vmVMMakerVersion (in category 'system attributes') ----- + vmVMMakerVersion + "Answer the version number of the VMMaker package from which the main VM was compiled." + "Smalltalk vmVMMakerVersion" + + | vmMakerID | + vmMakerID := self vmVersion substrings detect: [:token| token beginsWith: 'VMMaker']. + ^Integer readFrom: (vmMakerID subStrings: '.') last!

1 0

ArchiveViewer bug related to ZipNewFileMember>>rewindData
by tim Rowledge 23 Dec '17

23 Dec '17

Between 5.0 & 5.1 the archive viewer got broken; if you make an a zip archive from a file in the FileList and click on the file in the viewer list we get an error because the file stream is closed and so StandardFileStream>>primSize: fails. So far as I can tell from a quick look this stems from a change Chris M made to close the stream in ZipNewFileMember>>from: (dated 7/16/2015 15:29) - which seems like a pretty sensible thing to do to be honest. Most likely all that is needed is to reopen the stream as part of the rewind but maybe someone with a bit more recollection of how the inner workings of the zip viewer works might like to take a look. Maybe it just needs to use a few sends of the #openStreamWhile: that got added at the same time? As a wild guess, the file really needs to be open during the execution of #extractTo:from:to: in order to provide the data required by ArchiveViewer>>briefContents ? tim -- tim Rowledge; tim(a)rowledge.org; http://www.rowledge.org/tim The severity of the itch is proportional to the reach.

1 0

The Trunk: Tests-eem.388.mcz
by commits＠source.squeak.org 23 Dec '17

23 Dec '17

Eliot Miranda uploaded a new version of Tests to project The Trunk: http://source.squeak.org/trunk/Tests-eem.388.mcz ==================== Summary ==================== Name: Tests-eem.388 Author: eem Time: 23 December 2017, 11:08:38.390751 am UUID: b2681f27-cd71-426b-ab2d-0d331cd9fb24 Ancestors: Tests-nice.387 DecompilerTests: WeakSet>>scanForLoadedSymbol: no longer fails. Ensure Decompiler >> #initSymbols: is recompiled if it hasn't been. Its original bytecodes do cause a decompiler failure. Perhaps this should be in Compiler, but the old version of the method is perfectly serviceable. =============== Diff against Tests-nice.387 =============== Item was added: + (PackageInfo named: 'Tests') preamble: '"below, add code to be run before the loading of this package" + + "If recompiled, this method no longer creates an error if decompiled:" + ((''*| namedTemps i |*'' match: (Decompiler >> #initSymbols:) decompileString) + or: [''*| t2 t3 |*'' match: (Decompiler >> #initSymbols:) decompileString]) ifTrue: + [Decompiler recompile: #initSymbols: from: Decompiler]'! Item was changed: ----- Method: DecompilerTests>>decompilerFailures (in category 'utilities') ----- decompilerFailures "Here is the list of failures: either a syntax error, a hard error or some failure to decompile correctly. Collected via DecompilerTestFailuresCollector new computeFailures." "class name, selector, error class name or nil" ^ #( #(Behavior toolIconSelector: TestFailure) #(BrowserCommentTextMorph showPane SyntaxErrorNotification) #(CodeHolder getSelectorAndSendQuery:to:with: SyntaxErrorNotification) #(DecompilerTests testDecompileUnreachableParameter Error) #(HttpUrl checkAuthorization:retry: TestFailure) #(MCConfigurationBrowser post SyntaxErrorNotification) #(MVCToolBuilder setLayout:in: SyntaxErrorNotification) "same-name block-local temps in optimized blocks" #(ParseNodeBuilder variable: TestFailure) "Almost fully returning if's result in a nil node." #(PNGReadWriter copyPixelsGray: SyntaxErrorNotification) #(ProtoObjectTest testIfNilIfNotNil SyntaxErrorNotification) #(ProtoObjectTest testIfNotNil SyntaxErrorNotification) #(ProtoObjectTest testIfNotNilIfNil SyntaxErrorNotification) #(SHMCClassDefinition withAllSuperclassesDo: SyntaxErrorNotification) "same-name block-local temps in optimized blocks" #(TextURL actOnClickFor: TestFailure) #(TTContourConstruction segmentsDo: SyntaxErrorNotification) "Worth fixing; these two are mistaken conversion from a whileTrue: to a to:do: but the index is used outside the whileTrue:" + #(TTFontReader processHorizontalMetricsTable:length: SyntaxErrorNotification))! - #(TTFontReader processHorizontalMetricsTable:length: SyntaxErrorNotification) - #(WeakSet scanForLoadedSymbol: TestFailure))!

1 0

The Trunk: Network-eem.214.mcz
by commits＠source.squeak.org 23 Dec '17

23 Dec '17

Eliot Miranda uploaded a new version of Network to project The Trunk: http://source.squeak.org/trunk/Network-eem.214.mcz ==================== Summary ==================== Name: Network-eem.214 Author: eem Time: 23 December 2017, 10:24:05.025715 am UUID: c1c30a2b-ecc5-4f08-a54f-481ffdbebf77 Ancestors: Network-pre.213 Remove unsent methods which are holding onto an undecared variable (parameters). =============== Diff against Network-pre.213 =============== Item was removed: - ----- Method: MIMEDocument>>parameterNamed: (in category 'accessing') ----- - parameterNamed: parameterKey - - ^ parameters at: parameterKey! Item was removed: - ----- Method: MIMEDocument>>parameterNamed:ifAbsent: (in category 'accessing') ----- - parameterNamed: parameterKey ifAbsent: aBlock - - ^ parameters at: parameterKey ifAbsent: aBlock! Item was removed: - ----- Method: MIMEDocument>>privateParameters: (in category 'private') ----- - privateParameters: aDictionary - parameters := aDictionary!

1 0

Saving Projects Broken? Squeak5.1-16549-64bit.2.image
by John-Reed Maffeo 23 Dec '17

23 Dec '17

/Smalltalk/Squeak64/Squeak5.1-16549-64bit.2.image Squeak5.1 latest update: #16549 Current Change Set: WorkSpace Image format 68021 (64 bit) I have no experience with saving projects, but I thought I would try it and see how it works. The image I am using has four projects: 1. The root project that opens in a fresh image 2. A Morphic project 3.1 Another Morphic project and, 3.2 An MVC project in the 3.1 Morphic project. When trying to save 2, I get an error message. Note: I also tried this on a fresh image with the same results. a. Open new (as downloaded) image. b. Create Morphic project, make no changes, just open it. c. Try to save it >> error It looks like it is trying to process "(TheWorldMainDockingBar>>#toggleFullScreenOn: "a CompiledMethod(2777225)")" jrm

4 9

filename extensions for Form reading
by tim Rowledge 23 Dec '17

23 Dec '17

I’m looking at ImageMorph>>#readFromFile and find myself a bit puzzled. The initial code is simple enough but essentially lets a user enter any sort of filename whatsoever and then relies upon the Form>>fromFileNamed: method handling anything odd - which will be lots of fun if the file doesn’t exist (because the not-there-readonlyfile error is not handled at all) or is not actually an image file (because notifier). Sigh. I was intending to improve this by using a nice new FileChooserDialog which would show only image files by filtering on the typical file extensions, or at least the ones we can handle (not much point with any others). The problem then becomes how to find those extensions; it might look like ImageReadWriter class>>allTypicalFileExtensions would do it but if we look for usages things get oh so much muddier. In Form class>>#fileReaderServicesForFile:suffix: for example, we use #allTypicalFileExtensions and add ‘form’. And rather weirdly, ‘*’ too, which must surely be a mistake? FileList2 class>>#endingSpecs does a quite odd looking sequence of making an array of arrays where the contents of an array are immediately replaced by the results of ImageReadWriter allTypicalFileExtensions, making me wonder “why?” Imports>>#importImageDirectory: collects file extensions with "ImageReadWriter allTypicalFileExtensions add: ‘form’ “ So unless I've missed something, it would be more useful to implement Form class> allTypicalFileExtensions as ^ImageReadWriter allTypicalFileExtensions add: ‘form’. tim -- tim Rowledge; tim(a)rowledge.org; http://www.rowledge.org/tim Java: the best argument for Smalltalk since C++

2 1

The Trunk: CollectionsTests-eem.287.mcz
by commits＠source.squeak.org 22 Dec '17

22 Dec '17

Eliot Miranda uploaded a new version of CollectionsTests to project The Trunk: http://source.squeak.org/trunk/CollectionsTests-eem.287.mcz ==================== Summary ==================== Name: CollectionsTests-eem.287 Author: eem Time: 22 December 2017, 2:14:32.454096 am UUID: 4ae7ecb8-a1b6-474b-a90a-b35d04536b8a Ancestors: CollectionsTests-eem.286 Fix ByteArrayTest>>testFallbackReplaceFromToWith =============== Diff against CollectionsTests-eem.286 =============== Item was changed: ----- Method: ByteArrayTest>>testFallbackReplaceFromToWith (in category 'testing') ----- testFallbackReplaceFromToWith | teststring ba sz cm | + teststring := 'Test string' asByteArray. + sz := teststring byteSize. - teststring := 'Test string'. - sz := 'Test string' byteSize. ba := ByteArray new: sz. cm := SequenceableCollection compiledMethodAt: #replaceFrom:to:with:startingAt:. self shouldnt: [cm valueWithReceiver: ba arguments: {1. sz. teststring. 1}] raise: Exception description: 'Primitive 105 should be optional for ByteArray' !

2 1

← Newer
1
2
3
4
5
6
7
8
9
...
16
Older →

Jump to page: