Squeak-dev August 2011

squeak-dev@lists.squeakfoundation.org

55 participants
171 discussions

Daily Commit Log
by commits＠source.squeak.org 19 Aug '11

19 Aug '11

Changes to Trunk (http://source.squeak.org/trunk.html) in the last 24 hours: http://lists.squeakfoundation.org/pipermail/packages/2011-August/004935.html Name: Compiler-eem.212 Ancestors: Compiler-eem.211 Add a useful validation check to the closure analysis. Insist that local variables have at most one defining scope. =============================================

1 0

The Trunk: Compiler-eem.212.mcz
by commits＠source.squeak.org 19 Aug '11

19 Aug '11

Eliot Miranda uploaded a new version of Compiler to project The Trunk: http://source.squeak.org/trunk/Compiler-eem.212.mcz ==================== Summary ==================== Name: Compiler-eem.212 Author: eem Time: 18 August 2011, 4:32:55.72 pm UUID: 7aa3630c-0547-4f85-821b-c1a9bb4dbbd5 Ancestors: Compiler-eem.211 Add a useful validation check to the closure analysis. Insist that local variables have at most one defining scope. =============== Diff against Compiler-eem.211 =============== Item was changed: ----- Method: TempVariableNode>>definingScope: (in category 'code generation (closures)') ----- definingScope: scopeBlock "<BlockNode>" + definingScope ifNotNil: + [self error: 'temp has more than one defining scope. This is probably a parser error']. definingScope := scopeBlock!

1 0

anyone ever use google maps with squeak?
by Lawson English 18 Aug '11

18 Aug '11

Hi all. I'm trying to use the google maps web API from squeak, but I'm stuck at square one (as usual). I can't seem to get https working properly to access a google map server. The closest I have gotten is using the HttpsSocket class and utilities found in Cobalt (squeak 3.8) but that doesn't quite work either. a:= HttpsUrl absoluteFromText: 'https://maps.googleapis.com/maps/api/place/search/json?location=52.36973,5.…'. Transcript show: a retrieveContents contents. It works in a browser, but not with any version of squeak I have tried using any code I have tried. The URL I'm trying to access is: 'https://maps.googleapis.com/maps/api/place/search/json?location=52.36973,5.…' Suggestions on how to get it through squeak? I'll take any version, any flavor at this point. Caveat: I just installed Mac OS X Lion, which may be complicating things somehow. Thanks. Lawson

3 11

Alien bug fix
by Eliot Miranda 18 Aug '11

18 Aug '11

Hi All, I've just published Alien-eem.14 to http://www.squeaksource.com/Alien. This fix needs to be integrated into each dialect's version of Alien. The bug causes VM crashes due to trying to free stale pointers from previous runs. Alien-eem.14 Author: eem Time: 18 August 2011, 10:34:20.024 am UUID: 39bedad1-6159-4ba4-be35-277df2a25977 Ancestors: Alien-eem.13 Fix startUp: discarding of stale values from Alien's GCMallocedAliens class var. -- best, Eliot

1 0

The Inbox: System-hsj.454.mcz
by commits＠source.squeak.org 18 Aug '11

18 Aug '11

A new version of System was added to project The Inbox: http://source.squeak.org/inbox/System-hsj.454.mcz ==================== Summary ==================== Name: System-hsj.454 Author: hsj Time: 18 August 2011, 11:50:35.656 am UUID: 35f24b27-934a-4a28-a012-c140a395fe0f Ancestors: System-hsj.453 Fixed reference to Pharo-specific PrimitiveFailed error. =============== Diff against System-hsj.453 =============== Item was changed: ----- Method: SmalltalkImage>>maxExternalSemaphores (in category 'vm parameters') ----- maxExternalSemaphores "The size of array in some VM's where external signals for semaphores in externalObjects are handled. Essentially, if a semaphore is registered in externalObjects outside its bounds, they will not be signalled." + ^[self vmParameterAt: 49] on: Error do: [:ex | ex return: nil]! - ^[self vmParameterAt: 49] on: PrimitiveFailed do: [:ex | ex return: nil]!

1 0

The Inbox: System-hsj.453.mcz
by commits＠source.squeak.org 18 Aug '11

18 Aug '11

A new version of System was added to project The Inbox: http://source.squeak.org/inbox/System-hsj.453.mcz ==================== Summary ==================== Name: System-hsj.453 Author: hsj Time: 18 August 2011, 11:47:23.92 am UUID: 2e054529-adf7-4ac2-8156-53d736d9edfd Ancestors: System-hsj.452 Reverted back to maxExternalSemaphores, it is in fact a more accurate name. Added more comments Changed all uses of ProtectTable lock, and removed it from definition. =============== Diff against System-hsj.452 =============== Item was changed: Object subclass: #ExternalSemaphoreTable instanceVariableNames: '' + classVariableNames: 'ProtectAdd ProtectRemove' - classVariableNames: 'ProtectAdd ProtectRemove ProtectTable' poolDictionaries: '' category: 'System-Support'! + !ExternalSemaphoreTable commentStamp: 'HenrikSperreJohansen 8/18/2011 11:25' prior: 0! - !ExternalSemaphoreTable commentStamp: 'HenrikSperreJohansen 8/17/2011 14:52' prior: 0! By John M McIntosh johnmci(a)smalltalkconsulting.com This class was written to mange the external semaphore table. When I was writing a Socket test server I discovered various race conditions on the access to the externalSemaphore table. This new class uses class side methods to restrict access using two mutex semaphores, one for removal and one for additions to the table. It seemed cleaner to deligate the reponsibility here versus adding more code and another class variable to SystemDictionary + Note that in Smalltalk recreateSpecialObjectsArray we still directly play with the table. + Henrik Sperre Johansen + The name is somewhat of a misnomer; the table can be used for any objects, not just semaphores. + That is its main usage though, so a split which deals with semaphores and other external objects differently + (In the same underlying table) is not currently worth it. + Therefore, while in general not all users will care if the table is above a certain size, we still guard against adding more objects than the limit above which external signals would be lost (on some VMs.) + ! - - Note that in Smalltalk recreateSpecialObjectsArray we still directly play with the table.! Item was changed: ----- Method: ExternalSemaphoreTable class>>clearExternalObjects (in category 'accessing') ----- clearExternalObjects "Clear the array of objects that have been registered for use in non-Smalltalk code." + "Only lock additions, removals executing in parallel would have little effect on the resulting array" + - "Only protect adds, concurrent removals would make little difference as to how empty the table is..." ProtectAdd critical: [ self unprotectedExternalObjects: Array new]. ! Item was changed: ----- Method: ExternalSemaphoreTable class>>collectionBasedOn:withRoomFor: (in category 'private') ----- + collectionBasedOn: externalObjects withRoomFor: anObject + "Called if no slots to put anObject in have been found in externalObjects " + "Return a externalObject collection which does, either: + - Same collection with some slots freed up by finalization logic + - A larger array, which has replaced the parameter as canonical externalObject array. + An error must be raised if this method is incapable of fulfilling its duties" + + | newObjects newSize | + "grow linearly" + newSize := externalObjects size +20. + (self freedSlotsIn: externalObjects ratherThanIncreaseSizeTo: newSize) + ifTrue: [newObjects := externalObjects] - collectionBasedOn: aCollection withRoomFor: anObject - "Return a collection with room to store anObject, either aCollection itself with free slots, or a larger copy" - | sz newObjects newSize | - sz := aCollection size. - newSize := sz +20. "grow linearly" - (self freedSlotsIn: aCollection ratherThanIncreaseSizeTo: newSize) - ifTrue: [newObjects := aCollection] ifFalse: [ + newObjects := externalObjects species new: newSize. + newObjects replaceFrom: 1 to: externalObjects size with: externalObjects startingAt: 1. - newObjects := aCollection species new: newSize. - newObjects replaceFrom: 1 to: sz with: aCollection startingAt: 1. self unprotectedExternalObjects: newObjects.]. ^newObjects! Item was changed: ----- Method: ExternalSemaphoreTable class>>externalObjects (in category 'accessing') ----- externalObjects + "Not really sure why this is protected, once called you are out of protection of the locks anyways, and any use of the object is dangerous... + Only additions can potentially change the actual array in use though, so only lock that." + ^ProtectAdd critical: [self unprotectedExternalObjects].! - ^ProtectTable critical: [Smalltalk specialObjectsArray at: 39].! Item was changed: ----- Method: ExternalSemaphoreTable class>>freedSlotsIn:ratherThanIncreaseSizeTo: (in category 'private') ----- freedSlotsIn: externalObjects ratherThanIncreaseSizeTo: newSize "In some VM's, the external object table has a max size, which has to be increased for vm to reference them correctly." "In that case, try to gc to free slots first before actually increasing the max size" "Return whether I ended up freeing slots by GC'ing, or one should increase the size of " + ^Smalltalk vm maxExternalSemaphores - ^Smalltalk vm maxExternalObjects ifNotNil: [:maxSize | (maxSize < newSize) and: + [| needToGrow | - [| freedEnoughSlots | Smalltalk garbageCollect. "Do we have free slots now? If not, performing the GC didn't help and we still have to grow." + needToGrow := externalObjects includes: nil. - freedEnoughSlots := externalObjects includes: nil. + needToGrow - freedEnoughSlots ifTrue: ["If we did GC, warn we had to gc so actions could be taken if appropriate." + Transcript cr; show: TimeStamp now printString; cr; + show: 'WARNING: Had to GC to make room for more external objects.'; cr; + show: 'If this happens often, it would be a good idea to either:'; cr; + tab; show: '- Raise the maxExternalObjects size.'; cr; + tab; show: '- Write your code to explicitly release them rather than wait for finalization.'] + ifFalse: [Smalltalk vm maxExternalSemaphores: newSize]. + needToGrow]] - Transcript cr; show: TimeStamp now printString, - ' - WARNING: Had to GC to make room for more external objects. - If this happens often, it would be a good idea to either: - - Raise the maxExternalObjects size. - - Write your code to explicitly release them rather than wait for finalization.' ] - ifFalse: [Smalltalk vm maxExternalObjects: newSize]. - freedEnoughSlots]] ifNil:[false]! Item was changed: ----- Method: ExternalSemaphoreTable class>>safelyRegisterExternalObject: (in category 'accessing') ----- safelyRegisterExternalObject: anObject "Register the given object in the external objects array and return its index. If it is already there, just return its index." | objects firstEmptyIndex obj sz newObjects | + objects := self unprotectedExternalObjects. - objects := Smalltalk specialObjectsArray at: 39. "find the first empty slot" + firstEmptyIndex := + (self slotFor: anObject in: objects) + ifNil: ["if object has no empty slots, we need to get a collection which does" + objects := self collectionBasedOn: objects withRoomFor: anObject. + self slotFor: anObject in: objects.]. - firstEmptyIndex := self slotFor: anObject in: objects. - - "if object has no empty slots, we need to get a collection which does" - firstEmptyIndex = 0 ifTrue: [ - objects := self collectionBasedOn: objects - withRoomFor: anObject. - firstEmptyIndex := self slotFor: anObject in: objects.]. - objects at: firstEmptyIndex put: anObject. ^ firstEmptyIndex ! Item was changed: ----- Method: ExternalSemaphoreTable class>>slotFor:in: (in category 'private') ----- slotFor: anObject in: aCollection + "Return the first empty slot, or nil if there is none" - "find the first empty slot" "The following was written in an atomic fashion using special methods with no suspension points, not sure if on purpose, but keeping it that way for now. Uses should be protected by the ProtectTable semaphore anyhow, but... it's too much work to reason 100% about it" | firstEmptyIndex | 1 to: aCollection size do: [:i | | obj | obj := aCollection at: i. obj == anObject ifTrue: [^ i]. "object already there, just return its index" (firstEmptyIndex == nil and: [obj == nil]) ifTrue: [firstEmptyIndex := i]]. + ^firstEmptyIndex! - ^firstEmptyIndex ifNil: [0]! Item was removed: - ----- Method: SmalltalkImage>>maxExternalObjects (in category 'vm parameters') ----- - maxExternalObjects - "The size of array where external objects are registered. Only in Cog" - self isRunningCog ifFalse: [^nil]. - ^self vmParameterAt: 49! Item was removed: - ----- Method: SmalltalkImage>>maxExternalObjects: (in category 'vm parameters') ----- - maxExternalObjects: aSize - "This method should never be called as result of normal program execution. - If it is however, handle it differently: - - In development, signal an error to promt user to set a bigger size at startup immediately. - - In production, accept the cost of potentially unhandled interrupts, but log the action for later review. - - See comment in maxExternalObjectsSilently: why this behaviour is desirable," - - "Can't find a place where development/production is decided. - Suggest Smalltalk image inProduction, but use an overridable temp meanwhile." - |inProduction| - - self isRunningCog ifFalse: [^0]. - - inProduction := false. - ^"Smalltalk image" inProduction - ifFalse: [self error: 'Not enough space for external objects, set a larger size at startup!!'] - ifTrue: [ - self maxExternalObjectsSilently: aSize. - Transcript cr; - show: 'WARNING: Had to increase size of external object table due to many concurrently in use!!'; cr; - show: 'You should increase this size at startup using #maxExternalObjectsSilently:'; - cr; - show: 'Current table size: ', self maxExternalObjects printString]! Item was removed: - ----- Method: SmalltalkImage>>maxExternalObjectsSilently: (in category 'vm parameters') ----- - maxExternalObjectsSilently: aSize - "Changes the size of array where external objects are registered. - The size can only grow, and will always be the next power of two larger than the parameter. - - The intended use is to set the table size to some adequate maximum as part of a non-resuming image startUp. - - Setting this at any time other than start-up can potentially lose requests. - i.e. during the realloc new storage is allocated, the old contents are copied and then pointers are switched. - Requests occurring during copying won't be seen if they occur to indices already copied. " - - self isRunningCog ifFalse: [^0]. - "The vm-header field is a short, maximum 64k entries. Well, on most platforms anyways " - (aSize < 0 or: [aSize > 16rFFFF]) ifTrue: [^DomainError signalFrom: 0 to: 16rFFFF]. - ^self vmParameterAt: 49 put: aSize! Item was changed: ----- Method: SmalltalkImage>>maxExternalSemaphores (in category 'vm parameters') ----- maxExternalSemaphores + "The size of array in some VM's where external signals for semaphores in externalObjects are handled. + Essentially, if a semaphore is registered in externalObjects outside its bounds, they will not be signalled." + ^[self vmParameterAt: 49] on: PrimitiveFailed do: [:ex | ex return: nil]! - "The size of table where external semaphores are registered. Only in Cog" - self isRunningCog ifFalse: [^nil]. - ^self vmParameterAt: 49! Item was changed: ----- Method: SmalltalkImage>>maxExternalSemaphores: (in category 'vm parameters') ----- maxExternalSemaphores: aSize + "This method should never be called as result of normal program execution. + If it is however, handle it differently: + - In development, signal an error to promt user to set a bigger size at startup immediately. + - In production, accept the cost of potentially unhandled interrupts, but log the action for later review. - "Changes the size of table where external semaphores are registered. - The size can only grow, and will always be the next power of two larger than the parameter. + See comment in maxExternalObjectsSilently: why this behaviour is desirable," + + "Can't find a place where development/production is decided. + Suggest Smalltalk image inProduction, but use an overridable temp meanwhile." + |inProduction| - Setting this at any time other than start-up can potentially lose requests. - i.e. during the realloc new storage is allocated, t - he old contents are copied and then pointers are switched. - Requests occurring during copying won't be seen if they occur to indices already copied. - The intended use is to set the table to some adequate maximum at start-up" + self maxExternalSemaphores ifNil: [^0]. + + inProduction := false. + ^"Smalltalk image" inProduction + ifFalse: [self error: 'Not enough space for external objects, set a larger size at startup!!'] + ifTrue: [ + self maxExternalSemaphoresSilently: aSize. + Transcript show: 'WARNING: Had to increase size of semaphore signal handling table due to many external objects concurrently in use'; cr; + show: 'You should increase this size at startup using #maxExternalObjectsSilently:'; cr; + show: 'Current table size: ', self maxExternalSemaphores printString]! - self isRunningCog ifFalse: [^0]. - "The vm-header field is a short, maximum 64k entries. Well, on most platforms anyways " - (aSize < 0 or: [aSize > 16rFFFF]) ifTrue: [^DomainError signalFrom: 0 to: 16rFFFF]. - ^self vmParameterAt: 49 put: aSize! Item was added: + ----- Method: SmalltalkImage>>maxExternalSemaphoresSilently: (in category 'vm parameters') ----- + maxExternalSemaphoresSilently: aSize + "Changes the size of array where external signals for semaphores in externalObjects are handled.. + The size can only grow, and will always be the next power of two larger than the parameter. + + The intended use is to set the table size to some adequate maximum as part of a non-resuming image startUp. + + Setting this at any time other than start-up can potentially result in lost signals during reallocation. + i.e. Requests handled during copying og signals from old to new array won't be seen if they occur to indices already copied, before pointers to the new and old arrays are switched. " + + self maxExternalSemaphores ifNil: [^0]. + "The vm-header field where the size is stored is a short, maximum 64k entries. + Well, on most platforms anyways " + (aSize < 0 or: [aSize > 16rFFFF]) ifTrue: [^DomainError signalFrom: 0 to: 16rFFFF]. + ^self vmParameterAt: 49 put: aSize!

1 0

Daily Commit Log
by commits＠source.squeak.org 18 Aug '11

18 Aug '11

Changes to Trunk (http://source.squeak.org/trunk.html) in the last 24 hours: http://lists.squeakfoundation.org/pipermail/packages/2011-August/004934.html Name: Morphic-cmm.568 Ancestors: Morphic-nice.567 - Fixed Auto Enclose preference. - Restored ability to use Control in lieu of Alt+Shift for enclosing with (, {, or double-quote. =============================================

1 0

Wiki after com-swiki.
by Alan Grimes 17 Aug '11

17 Aug '11

om I've had a great time with Com-swiki. I mean, how can you beat "just click the green box" as a startup procedure! =P but com-swiki hasn't been maintained in 6 years and it's starting to show its age. (66.92.168.49). I think it is time for me to start looking at an update. I am running it on an Athlon 800, 384mb ram. I am migrating my network, as much as possible, to IPv6 compatibility, in anticipation of getting on the new internet in a few decades time... I am only beginning the process of migration. I'll need a replacement platform, which will probably require a few hunderd times more effort to set up. =\ Then I'll need to export and transition my content from the few pages I have right now. Anyway, I'm really sad to see Com Swiki go, it's one of the very few pieces of Good Software I've had the pleasure of using in recent years. -- E T F N H E D E D Powers are not rights.

1 0

Squeak Oversight Board minutes – 8/16/11
by Chris Cunnington 17 Aug '11

17 Aug '11

Attending: Chris Muller, Bert Freudenberg, Colin Putney, Randal Schwartz, Levente Uzonyi, Jecel Assumpacao Jr., Chris Cunnington, Craig Latta * Craig Latta discussed an early release of Spoon; there was a Spoon debugging bull session * Spoon project category on Mantis requested, and created * trademark policy - what if somebody creates a product with the name Squeak that we disagree with? issues around that were explored * It's the job of the board to manage the name "Squeak." That implies managing the domain name squeak.org (among others), the services provided there, the mailing list, official releases of the Squeak product etc. We do that based on the moral authority that comes out of the election process, since we don't have a legal ownership of the name. We're also making explicit our historical position of not objecting to derivative names, such as Squeak Germany, SiliconSqueak, Stable Squeak, SqueakSource and the like. * the future of SqueakSource was discussed

1 0

The Inbox: System-hsj.452.mcz
by commits＠source.squeak.org 17 Aug '11

17 Aug '11

A new version of System was added to project The Inbox: http://source.squeak.org/inbox/System-hsj.452.mcz ==================== Summary ==================== Name: System-hsj.452 Author: hsj Time: 17 August 2011, 4:14:49.888 pm UUID: 9b7158fc-3a52-42f1-b44d-11264a79420b Ancestors: System-ul.451 Guard against max external object table size on Cog. The default is to raise an error if this were to occur. (see #maxExternalObjects:) Split ExternalSemaphoreTable mutex into separate ones for addition and removal. While removal may seem not needed, it is for the case where 1 thread tries to add, while 2 threads try to remove the same object. (If not, the added object could end up being removed unintentionally) This allows removal to take place while another process tries to add one, f.ex. as a result of trying to make free slots rather than grow by doing GC to trigger finalizers. (see #freedSlotsIn:ratherThanIncreaseSizeTo:) Both methods referred to above write warnings to the transcript, there may be better suited places to report these in Squeak. =============== Diff against System-ul.451 =============== Item was changed: Object subclass: #ExternalSemaphoreTable instanceVariableNames: '' + classVariableNames: 'ProtectAdd ProtectRemove ProtectTable' - classVariableNames: 'ProtectTable' poolDictionaries: '' category: 'System-Support'! + !ExternalSemaphoreTable commentStamp: 'HenrikSperreJohansen 8/17/2011 14:52' prior: 0! - !ExternalSemaphoreTable commentStamp: '<historical>' prior: 0! By John M McIntosh johnmci(a)smalltalkconsulting.com + This class was written to mange the external semaphore table. When I was writing a Socket test server I discovered various race conditions on the access to the externalSemaphore table. This new class uses class side methods to restrict access using two mutex semaphores, one for removal and one for additions to the table. It seemed cleaner to deligate the reponsibility here versus adding more code and another class variable to SystemDictionary - This class was written to mange the external semaphore table. When I was writing a Socket test server I discovered various race conditions on the access to the externalSemaphore table. This new class uses class side methods to restrict access using a mutex semaphore. It seemed cleaner to deligate the reponsibility here versus adding more code and another class variable to SystemDictionary + + Note that in Smalltalk recreateSpecialObjectsArray we still directly play with the table.! Item was changed: ----- Method: ExternalSemaphoreTable class>>clearExternalObjects (in category 'accessing') ----- clearExternalObjects "Clear the array of objects that have been registered for use in non-Smalltalk code." + "Only protect adds, concurrent removals would make little difference as to how empty the table is..." + ProtectAdd critical: [ + self unprotectedExternalObjects: Array new]. - - ProtectTable critical: [Smalltalk specialObjectsArray at: 39 put: Array new]. ! Item was added: + ----- Method: ExternalSemaphoreTable class>>collectionBasedOn:withRoomFor: (in category 'private') ----- + collectionBasedOn: aCollection withRoomFor: anObject + "Return a collection with room to store anObject, either aCollection itself with free slots, or a larger copy" + | sz newObjects newSize | + sz := aCollection size. + newSize := sz +20. "grow linearly" + (self freedSlotsIn: aCollection ratherThanIncreaseSizeTo: newSize) + ifTrue: [newObjects := aCollection] + ifFalse: [ + newObjects := aCollection species new: newSize. + newObjects replaceFrom: 1 to: sz with: aCollection startingAt: 1. + self unprotectedExternalObjects: newObjects.]. + ^newObjects! Item was added: + ----- Method: ExternalSemaphoreTable class>>freedSlotsIn:ratherThanIncreaseSizeTo: (in category 'private') ----- + freedSlotsIn: externalObjects ratherThanIncreaseSizeTo: newSize + "In some VM's, the external object table has a max size, which has to be increased for vm to reference them correctly." + "In that case, try to gc to free slots first before actually increasing the max size" + "Return whether I ended up freeing slots by GC'ing, or one should increase the size of " + ^Smalltalk vm maxExternalObjects + ifNotNil: [:maxSize | + (maxSize < newSize) and: + [| freedEnoughSlots | + Smalltalk garbageCollect. + "Do we have free slots now? If not, performing the GC didn't help and we still have to grow." + freedEnoughSlots := externalObjects includes: nil. + + freedEnoughSlots + ifTrue: ["If we did GC, warn we had to gc so actions could be taken if appropriate." + Transcript cr; show: TimeStamp now printString, + ' + WARNING: Had to GC to make room for more external objects. + If this happens often, it would be a good idea to either: + - Raise the maxExternalObjects size. + - Write your code to explicitly release them rather than wait for finalization.' ] + ifFalse: [Smalltalk vm maxExternalObjects: newSize]. + freedEnoughSlots]] + ifNil:[false]! Item was changed: ----- Method: ExternalSemaphoreTable class>>initialize (in category 'initialize') ----- initialize + ProtectAdd := Semaphore forMutualExclusion. + ProtectRemove := Semaphore forMutualExclusion! - ProtectTable := Semaphore forMutualExclusion! Item was changed: ----- Method: ExternalSemaphoreTable class>>registerExternalObject: (in category 'accessing') ----- registerExternalObject: anObject + ^ ProtectAdd critical: [self safelyRegisterExternalObject: anObject] - ^ ProtectTable critical: [self safelyRegisterExternalObject: anObject] ! Item was changed: ----- Method: ExternalSemaphoreTable class>>safelyRegisterExternalObject: (in category 'accessing') ----- safelyRegisterExternalObject: anObject "Register the given object in the external objects array and return its index. If it is already there, just return its index." | objects firstEmptyIndex obj sz newObjects | objects := Smalltalk specialObjectsArray at: 39. "find the first empty slot" + firstEmptyIndex := self slotFor: anObject in: objects. - firstEmptyIndex := 0. - 1 to: objects size do: [:i | - obj := objects at: i. - obj == anObject ifTrue: [^ i]. "object already there, just return its index" - (obj == nil and: [firstEmptyIndex = 0]) ifTrue: [firstEmptyIndex := i]]. + "if object has no empty slots, we need to get a collection which does" - "if no empty slots, expand the array" firstEmptyIndex = 0 ifTrue: [ + objects := self collectionBasedOn: objects + withRoomFor: anObject. + firstEmptyIndex := self slotFor: anObject in: objects.]. - sz := objects size. - newObjects := objects species new: sz + 20. "grow linearly" - newObjects replaceFrom: 1 to: sz with: objects startingAt: 1. - firstEmptyIndex := sz + 1. - Smalltalk specialObjectsArray at: 39 put: newObjects. - objects := newObjects]. objects at: firstEmptyIndex put: anObject. ^ firstEmptyIndex ! Item was added: + ----- Method: ExternalSemaphoreTable class>>slotFor:in: (in category 'private') ----- + slotFor: anObject in: aCollection + "find the first empty slot" + "The following was written in an atomic fashion using special methods with no suspension points, not sure if on purpose, but keeping it that way for now. + Uses should be protected by the ProtectTable semaphore anyhow, but... it's too much work to reason 100% about it" + + | firstEmptyIndex | + 1 to: aCollection size do: [:i | | obj | + obj := aCollection at: i. + obj == anObject ifTrue: [^ i]. "object already there, just return its index" + (firstEmptyIndex == nil and: [obj == nil]) ifTrue: [firstEmptyIndex := i]]. + ^firstEmptyIndex ifNil: [0]! Item was added: + ----- Method: ExternalSemaphoreTable class>>unprotectedExternalObjects (in category 'private') ----- + unprotectedExternalObjects + ^Smalltalk specialObjectsArray at: 39! Item was added: + ----- Method: ExternalSemaphoreTable class>>unprotectedExternalObjects: (in category 'private') ----- + unprotectedExternalObjects: aCollection + ^Smalltalk specialObjectsArray at: 39 put: aCollection! Item was changed: ----- Method: ExternalSemaphoreTable class>>unregisterExternalObject: (in category 'accessing') ----- unregisterExternalObject: anObject + ProtectRemove critical: [self safelyUnregisterExternalObject: anObject] - ProtectTable critical: [self safelyUnregisterExternalObject: anObject] ! Item was added: + ----- Method: SmalltalkImage>>maxExternalObjects (in category 'vm parameters') ----- + maxExternalObjects + "The size of array where external objects are registered. Only in Cog" + self isRunningCog ifFalse: [^nil]. + ^self vmParameterAt: 49! Item was added: + ----- Method: SmalltalkImage>>maxExternalObjects: (in category 'vm parameters') ----- + maxExternalObjects: aSize + "This method should never be called as result of normal program execution. + If it is however, handle it differently: + - In development, signal an error to promt user to set a bigger size at startup immediately. + - In production, accept the cost of potentially unhandled interrupts, but log the action for later review. + + See comment in maxExternalObjectsSilently: why this behaviour is desirable," + + "Can't find a place where development/production is decided. + Suggest Smalltalk image inProduction, but use an overridable temp meanwhile." + |inProduction| + + self isRunningCog ifFalse: [^0]. + + inProduction := false. + ^"Smalltalk image" inProduction + ifFalse: [self error: 'Not enough space for external objects, set a larger size at startup!!'] + ifTrue: [ + self maxExternalObjectsSilently: aSize. + Transcript cr; + show: 'WARNING: Had to increase size of external object table due to many concurrently in use!!'; cr; + show: 'You should increase this size at startup using #maxExternalObjectsSilently:'; + cr; + show: 'Current table size: ', self maxExternalObjects printString]! Item was added: + ----- Method: SmalltalkImage>>maxExternalObjectsSilently: (in category 'vm parameters') ----- + maxExternalObjectsSilently: aSize + "Changes the size of array where external objects are registered. + The size can only grow, and will always be the next power of two larger than the parameter. + + The intended use is to set the table size to some adequate maximum as part of a non-resuming image startUp. + + Setting this at any time other than start-up can potentially lose requests. + i.e. during the realloc new storage is allocated, the old contents are copied and then pointers are switched. + Requests occurring during copying won't be seen if they occur to indices already copied. " + + self isRunningCog ifFalse: [^0]. + "The vm-header field is a short, maximum 64k entries. Well, on most platforms anyways " + (aSize < 0 or: [aSize > 16rFFFF]) ifTrue: [^DomainError signalFrom: 0 to: 16rFFFF]. + ^self vmParameterAt: 49 put: aSize!

1 0

← Newer
1
...
4
5
6
7
8
9
10
...
18
Older →

Jump to page: