- Pharo-dev - lists.pharo.org

[Pharo-project] VM team update

by David T. Lewis

Folks, We are in the midst of a flurry of activity to update the virtual machines for Mac, Windows and Unix in support of upcoming Squeak and Pharo releases. These VMs all will include Eliot Miranda's block closure support, as well as a large number of updates, bug fixes, and enhancements. There have been many significant updates to all of the major platforms, and I will not attempt to do justice to all of these. That said, the revision logs for the VMMaker package provide an overview of many of the changes, so I am attaching an update log below. It is worth noting here that John McIntosh has announced a new code base for Mac VMs (in addition to his iPhone VM work) with expanded 64-bit support. This will be the basis for future Mac VM releases following the current round of updates. Following is a change log of the VMMaker package updates, starting with current version, and going back to the introduction of block closure support in early 2009. - Dave ------- Name: VMMaker-dtl.169 Author: dtl Time: 10 April 2010, 3:43:27.612 pm Ancestors: VMMaker-ar.168 VMMaker 4.0.3 Add version identification primitives - primitiveInterpreterSourceVersion answers the version of VMMaker that was used to generate the VM source code. This is a string like '4.0.3'. - primitivePlatformSourceVersion answers the Subversion version level if $PLATFORM_SOURCE_VERSION has been defined, or fails otherwise. This should be a string like '2172'. - primitiveVMVersion answers an identifier for the VM version (VMMaker plus Subversion) if $VM_VERSION is defined, or fails otherwise. For the Unix VMs, this is a string like '4.0.3-2172'. - primitiveImageFormatVersion answers an integer identifier for the image format. This is the number that is stored in the image file header to identify image fomat (32/64 bit word size, block closure support). Rename some code generation methods from #writeXXX to #emitXXX for consistency with other method naming. Fix VMMaker file list processing to ignore Subversion administrative files (previously only CVS files were ignored). ------- Name: VMMaker-ar.168 Author: ar Time: 31 March 2010, 9:46:52 am Ancestors: VMMaker-ar.167 Bump VMMaker version to 4.0.2. ------- Name: VMMaker-ar.167 Author: ar Time: 31 March 2010, 9:44:42 am Ancestors: VMMaker-JMM.166 Fix SoundPlugin>>primSoundRecordSamplesInto again. Performs subtraction in the primitive, calls external code with the offset ALWAYS set to zero. Should be rationalized by removing the arg from the support code call. ------- Name: VMMaker-JMM.166 Author: JMM Time: 28 March 2010, 4:57:23 pm Ancestors: VMMaker-dtl.119, VMMaker-dtl.122, VMMaker-dtl.135, VMMaker-dtl.139, VMMaker-dtl.153, VMMaker-dtl.156, VMMaker-ar.160, VMMaker-ar.165, VMMaker-dtl.165 Merge some 64bit fixes primitiveAsyncFileOpen: filename is not an (int) address on 64bit systems MIDIPlugin &portName is not an int RePlugin>>recvrExtraPtr 64bits NULL is not int JPEGReadWriter2Plugin not 64bit clean primJPEGReadImage: primJPEGWriteImage: primitiveShowHostWindow: dispBits is unsigned char* Interpreter Ensure microsecond clock is 64bit statGCTime,statFullGCMSecs,statIGCDeltaTime,statIncrGCMSecs are all sqLong startTime,statGCTime as sqLong primitiveVMParameter has microsecond clock times 64bit readImageFromFile: 32 clean desiredHeapSize is unsigned ------- Name: VMMaker-dtl.165 Author: dtl Time: 28 March 2010, 6:54:54 pm Ancestors: VMMaker-dtl.164 Fix declaration of readImageFromFile:HeapSize:StartingAt: as per John's note at http://lists.squeakfoundation.org/pipermail/vm-dev/2010-January/003614.html ------- Name: VMMaker-dtl.164 Author: dtl Time: 28 March 2010, 1:44:25 pm Ancestors: VMMaker-eem.163 VMMaker 4.0.1 Reference Mantis 7458: [Vm-dev] microsecond timing for GC work Add primitiveMicrosecondClock and microsecond GC instrumentation by John McIntosh. Add primitiveUtcWithOffset. Fix signature of Interpreter>>dumpImage: (pointer declared as int). The new primitives require support in the platform code. Default implementations are provided to allow compilation without these platforms updates, see CCodeGenerator>>writeDefaultMacrosOn:. Without platform updates, the GC instrumentation falls back to millisecond precision, the primitiveMicrosecondClock primitive answers (incorrectly) a millisecond value, and primitiveUtcWithOffset fails the primitive. ------- Name: VMMaker-eem.163 Author: eem Time: 22 March 2010, 5:15:57 pm Ancestors: VMMaker-ar.162 SoundPlugin: fix bounds bug in primitiveSoundRecordSamples. The call to snd_RecordSamplesIntoAtLength neglected to subtract the start index from the size of the buffer to be written into. ------- Name: VMMaker-ar.162 Author: ar Time: 15 March 2010, 4:02:31 am Ancestors: VMMaker-dtl.161 Fix loading of image segments that are from older, but compatible image versions. ------- Name: VMMaker-dtl.161 Author: dtl Time: 14 March 2010, 9:19:53 am Ancestors: VMMaker-ar.160 VMMaker 4.0.0 Change #versionString from 3.x.x to 4.0.0. This is to align version numbering with the Squeak 4.0 release, and does not reflect a functional change in the VM. Change layout calculation in VMMakerTool to produce acceptable cosmetic layout for Squeak trunk, Pharo and other images. Font size calculation in Squeak trunk seems inconsistent with other images, so this change may be only a workaround for a bug in Squeak trunk. ------- Name: VMMaker-ar.160 Author: ar Time: 10 March 2010, 10:57:23 am Ancestors: VMMaker-dtl.159 Add David's fix for removing the inline definition of error(). ------- Name: VMMaker-dtl.159 Author: dtl Time: 24 February 2010, 9:15:48 am Ancestors: VMMaker-ar.158 VMMaker 3.11.14 Fix (- 16) syntax problem in BalloonEnginePlugin>>shortRunLengthAt:from: Bump versionString to 3.11.14 to reflect this change plus FFIPlugin updates incorporated in the last MC update. ------- Name: VMMaker-ar.158 Author: ar Time: 11 February 2010, 12:25:42 pm Ancestors: VMMaker-dtl.157 Merge ExternalForm primitives from Josh. ------- Name: VMMaker-dtl.157 Author: dtl Time: 29 January 2010, 12:14:34 pm Ancestors: VMMaker-dtl.156 VMMaker 3.11.13 Reference Mantis 0007461: Bulletproof BitBltSimulation against failure in #lockSurfaces Add patch from Josh Gargus to fix bug where #warpBits does not check the return value of #lockSurfaces, and can potentially crash when execution continues after the failed lock attempt. ------- Name: VMMaker-dtl.156 Author: dtl Time: 11 January 2010, 5:39:24 am Ancestors: VMMaker-dtl.155 VMMaker 3.11.12 Reference Mantis 7103: Playing sounds in linux 64 bits causes a vm segmentation fault Add changes to permit SoundPlugin to work on 64 bit platforms. Corresponding changes are required for platform support code. Platforms sources are currently at Subversion revision level 2147. Later revision levels will contain the patches to match this version of VMMaker. See Mantis 7103 for details. ------- Name: VMMaker-dtl.155 Author: dtl Time: 20 December 2009, 10:22:54 am Ancestors: VMMaker-dtl.154 VMMaker 3.11.11 Fix Interpreter>>byteSizeOf: to use #bytesPerWord rather than '4'. ------- Name: VMMaker-dtl.154 Author: dtl Time: 2 December 2009, 11:34:41 am Ancestors: VMMaker-dtl.153 Updated class comment for MiscPrimitivePlugin with information provided by Andreas in http://lists.squeakfoundation.org/pipermail/vm-dev/2009-December/003514.html ------- Name: VMMaker-dtl.153 Author: dtl Time: 27 November 2009, 4:49:50 am Ancestors: VMMaker-dtl.152 VMMaker 3.11.10 Reference Mantis 7421: Bug in Interpreter>>primitiveNextPut: Bug found and fixed by Henrik Johansen <henrik.s.johansen(a)veloxit.no> ------- Name: VMMaker-dtl.152 Author: dtl Time: 17 November 2009, 8:10:32 am Ancestors: VMMaker-dtl.151 VMMaker 3.11.9 Update KlattSynthesizerPlugin to refer to separately packaged pool variables, removing dependency on complete Speech package. ------- Name: VMMaker-dtl.151 Author: dtl Time: 14 November 2009, 9:54:55 am Ancestors: VMMaker-dtl.150 Add "VMMaker updateFromServer" for updating from a Monticello configuration ------- Name: VMMaker-dtl.150 Author: dtl Time: 11 November 2009, 9:13:25 pm Ancestors: VMMaker-ar.149 Add missing isExpression instance variable to TSendNode, apparently lost from dusty change set when integrating CPP macro generation for ObjectMemory>>isIntegerValue: ------- Name: VMMaker-ar.149 Author: ar Time: 9 November 2009, 10:41:59 am Ancestors: VMMaker-mtf.148 Small tweak for VMMaker appearance observed in latest trunk image. Makes the input fields solid gray to avoid bleeding the labels into the input fields. ------- Name: VMMaker-mtf.148 Author: mtf Time: 5 November 2009, 10:56:58 am Ancestors: VMMaker-dtl.147 no more underscore assignments ------- Name: VMMaker-dtl.147 Author: dtl Time: 1 November 2009, 12:00 am Ancestors: VMMaker-dtl.146 VMMaker 3.11.8 Reference Mantis 7407: BitBlt. Incorrect alpha values for several rules. Add VMMaker-BitBlt-AlphaFixes-jmv-M7407.cs bitblt alpha fixes by Juan Vuletich ------- Name: VMMaker-dtl.146 Author: dtl Time: 29 October 2009, 12:00 am Ancestors: VMMaker-dtl.145 VMMaker 3.11.7 Reference Mantis 7407: BitBlt. Incorrect alpha values for several rules. Changes and fixes provided by Juan Vuletich: Add missing memory access methods to BitBltSimulator and CArrayAccessor, required for bitblt simulation since the original Squeak 64 bit work but apparently omitted by accident. Add BitBltSimulationTest with two test methods moved from BitBltTest. The corresponding tests should be removed from BitBltTest at a later date (follow up required). Add package category VMMaker-Tests as a home for BitBltSimulationTest and hopefully for other tests in the future. ------- Name: VMMaker-dtl.145 Author: dtl Time: 25 October 2009, 12:00 am Ancestors: VMMaker-dtl.144 VMMaker 3.11.6 Fix incorrect use of zero index in #replaceLastOccurrence:with: that caused Slang inlining failure when running on a 64-bit image. ------- Name: VMMaker-dtl.144 Author: dtl Time: 24 October 2009, 12:00 am Ancestors: VMMaker-dtl.143 Add a "Rescan" button to VMMakerTool to scan the image, adding any new plugins and removing those no longer available. Allows a VMMakerTool configuration to be updated e.g. after loading a new plugin from SqueakSource. Per suggestion from Ian. ------- Name: VMMaker-dtl.143 Author: dtl Time: 20 October 2009, 12:00 am Ancestors: VMMaker-dtl.142 VMMaker 3.11.5 Reference Mantis 7405: Array new: SmallInteger maxVal broken. Change Set: ObjectAllocationSizeFixes-dtl-M7405 Date: 20 October 2009 Author: David T. Lewis Requests for very large object allocation result in various failures including VM crashes and incorrect results from primitive calls. This change set adds checks for arithmetic overflow in calculation of size requests, and limits the maximum size of a #growObjectMemory request to 16r7FFFFFFF to permit use of signed 32-bit integer arithmetic in the support code. Also included are some variable type declarations and removal of a now unnecessary #oop:isLessThan: call. There may be additional bugs in the support code, which is expected to fail gracefully even if someone requests an object memory size increase of 16r80000000 or more. In testing on a Linux VM, allocation of objects of size greater than 950MB appears to be possible, given enough physical memory to permit the allocation to succeed. I do not have sufficient real memory on my system to verify the actual results, although I can confirm that this is an effective way to bring a Linux system to its knees. ------- Name: VMMaker-dtl.142 Author: dtl Time: 7 October 2009, 12:00 am Ancestors: VMMaker-dtl.141 VMMaker 3.11.4 Provide hooks for Alien FFI, add #callInterpreter exported for external calls into #interpret and turn off inlining for #interpret. ------- Name: VMMaker-dtl.141 Author: dtl Time: 7 October 2009, 12:00 am Ancestors: VMMaker-dtl.140 Use #oop:isGreaterThanOrEqualTo: in #sufficientSpaceAfterGC: for consistency and readability. No functional change to generated C code. ------- Name: VMMaker-dtl.140 Author: dtl Time: 28 September 2009, 12:00 am Ancestors: VMMaker-dtl.139 Implement #imageFormatVersion in InterpreterSimulator to allow the simulator to load images with closure bytecodes. Change comment in ObjectMemory>>bytesPerWord to avoid confusion with C pointer terminology. Fix typo in InterpreterSimulator>>openOn:extraMemory: to change spelling of 'incomaptible'. Sorry to eliminate such a marvellous word from the image. I had hoped for an etymological link to the field of motivational finance, but alas it seems to be a common misspelling. ------- Name: VMMaker-dtl.139 Author: dtl Time: 16 September 2009, 9:43:58 pm Ancestors: VMMaker-dtl.138 Cosmetic fix - better timestamp format for comments in generated source files. ------- Name: VMMaker-dtl.138 Author: dtl Time: 2 September 2009, 8:24:56 am Ancestors: VMMaker-dtl.137 Add VMMMAKER_VERSION macro to the vm/interp.h header file to identify approximate source code version of generated sources (per ar suggestion) ------- Name: VMMaker-dtl.137 Author: dtl Time: 24 August 2009, 7:33:10 am Ancestors: VMMaker-dtl.136 VMMaker 3.11.3 Use UIManager for directory selection if possible, or fall back to FileList2 for older images. Update required for Pharo images. ------- Name: VMMaker-dtl.136 Author: dtl Time: 22 August 2009, 3:23:29 pm Ancestors: VMMaker-dtl.135 VMMaker 3.11.2 Reference Mantis 7266: SerialPlugin doesn't handle arbitrary nodes names Add support for named serial devices for SerialPlugin. The new primitives and support code are by Severin Lemaignan <severin.lemaignan(a)laas.fr>, with portions reimplemented here by dtl. Also changes method signatures to declare pointers correctly, permitting SerialPlugin to work on 64-bit host platforms. Additional platform code changes are required to match. ------- Name: VMMaker-dtl.135 Author: dtl Time: 18 August 2009, 7:49:41 am Ancestors: VMMaker-dtl.134 Reference Mantis 7384: GeniePlugin does not compile on closure-enabled images Eliminate two temporary variables from the GeniePlugin primitive to permit it to compile on closure-enabled images. ------- Name: VMMaker-dtl.134 Author: dtl Time: 14 August 2009, 10:15:48 am Ancestors: VMMaker-dtl.133 VMMaker 3.11.1 Reference Mantis 5238: Integer bug on 64bit image/32bit VM, DateAndTime now ==> -4391-12-26T16:18:33+00:00. See Mantis report for tests and performance evaluation. Provide an implementation of #isIntegerValue: that is both fast and correct on all combinations of 32 bit and 64 bit image and host. The implementation of #isIntegerValue: by tpr is fast, but does not work for a 64 bit image. The implementation of #isIntegerValue: by ikp is slower (especially on 32 bit hosts), but works correctly on 64 bit images. This is an implementation of #isIntegerValue: that uses conditional compilation to select the fastest correct implemention for the given image and host word size. Eventually this method will need to change again when the range of SmallInteger is increased for 64 bit images. ------- Name: VMMaker-dtl.133 Author: dtl Time: 14 August 2009, 10:01:20 am Ancestors: VMMaker-dtl.132 VMMaker 3.11.0 Add support for C preprocessor statements in Slang. Three new methods are supported: Object>>isDefined:inSmalltalk:comment:ifTrue:ifFalse: Object>>isDefined:inSmalltalk:comment:ifTrue: Object>>preprocessorExpression: Conditional preprocessor directives can be used to direct compilation of statement lists (the common usage), or they may be embedded within C expressions. See Mantis 5238 for change set VMMakerCppTranslation-dtl, merged into current VMMaker ------- Name: VMMaker-dtl.132 Author: dtl Time: 13 August 2009, 1:19:02 am Ancestors: VMMaker-dtl.131 VMMaker 3.10.8 Reference Mantis 7377: VM Crashes when setting port on destroyed socket (Win32) Patch (provided by Andreas) adds a guard test in SocketPlugin>>primitiveSocket:listenOnPort: to prevent calling sqSocketListenOnPort() when the primitive has already failed. ------- Name: VMMaker-dtl.131 Author: dtl Time: 10 August 2009, 10:45:01 pm Ancestors: VMMaker-dtl.130 VMMaker 3.10.8 Reference Mantis 7376: Bug in BitBlt (rgbMul sets alpha to 0) Patch provided by Juan Vuletich. Rule 37 sets alpha to zero if the destination is 32 bit. This affects antialiased StrikeFonts. This is related to the stuff I (jmv) fixed in April (0007327), but it is a different bug in a different method, that I didn't find that time: #partitionedMul:with:nBits:nPartitions: ignores the last argument, and acts as if it was always 3. With this fix, the second bitblt pass for aa strike fonts will not be needed anymore if text is black, meaning a small performance improvement, aside from correct bitblt behavior. ------- Name: VMMaker-dtl.130 Author: dtl Time: 4 August 2009, 11:36:14 pm Ancestors: VMMaker-dtl.129 Change VMMakerTool selection box label from '64-bit word' to '64-bit image'. ------- Name: VMMaker-dtl.129 Author: dtl Time: 4 August 2009, 10:04:21 pm Ancestors: VMMaker-dtl.128 The '64 bit VM?' selection box in VMMakerTool can be confusing for first-time VM builders. This change attempts to clarify the label and balloon help. Change label on selection box from '64 bit VM?' to '64-bit word', and balloon help from 'Set to build a 32 bit VM or a 64 bit VM' to 'Build (32- or 64-bit) VM to run 64-bit images. Default is false to support normal 32-bit images.'. ------- Name: VMMaker-dtl.128 Author: dtl Time: 3 August 2009, 7:46:17 pm Ancestors: VMMaker-dtl.127 Reference Mantis 7374: Inlining may cause broken C code A more general solution to TMethod-renameVars-dtl-M7374 that does better substring substitution when swizzling the C declaration. ------- Name: VMMaker-dtl.127 Author: dtl Time: 1 August 2009, 11:26:23 pm Ancestors: VMMaker-dtl.126 Reference Mantis 7374: Inlining may cause broken C code Change Set: TMethod-renameVars-dtl-M7374 Date: 1 August 2009 Author: David T. Lewis When renaming local variables for an inlined method, ensure that only the variable name is affected. Previously the expression declaration 'int i' was renamed to 'i1nt i1' due to naive use of copyReplaceAll. ------- Name: VMMaker-dtl.126 Author: dtl Time: 18 July 2009, 10:06:01 pm Ancestors: VMMaker-dtl.125 Correct misspelling of "receiver" in several method comments, noted by Eliot. ------- Name: VMMaker-dtl.125 Author: dtl Time: 5 July 2009, 10:52:01 pm Ancestors: VMMaker-dtl.124 VMMaker 3.10.7 Move some C code generator methods from kernel and collections packages into VMMaker package. Janatorial, see Mantis 7221. Follow up required: These methods will need to be removed from the kernel and collections packages at a later date. ------- Name: VMMaker-dtl.124 Author: dtl Time: 1 July 2009, 7:14:48 pm Ancestors: VMMaker-dtl.123 VMMaker 3.10.6 Update to patches for Mantis 7260: some Float comparison primitives don't handle Nan correctly Change Set: M7260-bytecodePrimGreaterEqual-update-dtl Date: 1 July 2009 Author: David T. Lewis One too many double negatives in the last patch. Reported by Bert and confirmed by Andreas and Nicolas, see http://lists.squeakfoundation.org/pipermail/vm-dev/2009-July/002780.html ------- Name: VMMaker-dtl.123 Author: dtl Time: 12 June 2009, 7:35:20 am Ancestors: VMMaker-dtl.122 VMMaker 3.10.5 Reference Mantis 7260: some Float comparison primitives don't handle Nan correctly Change Set: M7260-primitiveSmallIntegerCompareNan-Patch-nice Date: 4 January 2009 Author: nice This is a patch for bug http://bugs.squeak.org/view.php?id=7260 2 >= Float nan should be false. It used to be true... This patch applies to tricky handling of some special arithmetic bytecode when receiver is a SmallInteger. Change Set: M7260-primitiveFloatCompareNan-Patch-nice Date: 4 January 2009 Author: nice This is a patch for bug http://bugs.squeak.org/view.php?id=7260 2.0 >= Float nan should be false. It used to be true... ------- Name: VMMaker-dtl.122 Author: dtl Time: 5 May 2009, 7:55:55 pm Ancestors: VMMaker-dtl.121 Change Set: ClosureVMPopKiller-M7349 Date: 5 May 2009 Author: John M McIntosh Fix for mantis 7349, extra pop unbalances stack, causes closure VM to die ------- Name: VMMaker-dtl.121 Author: dtl Time: 30 April 2009, 10:56:31 pm Ancestors: VMMaker-dtl.120 Fix two methods with assignment to block temp arguments, reported on vm-dev list: <http://lists.squeakfoundation.org/pipermail/vm-dev/2009-April/002607.html> > On Thu, Apr 30, 2009 at 8:43 AM, Igor Stasenko <siguctua at gmail.com> wrote: >> >> There are multiple issues, when loading VMMaker into closure-aware >> image (10292). >> >> two #needsToRegenerateInterpreterFile methods >> storing into block args. Fix is easy and quick. ------- Name: VMMaker-dtl.120 Author: dtl Time: 12 April 2009, 12:51:35 pm Ancestors: VMMaker-dtl.119 VMMaker 3.10.4 Reference Mantis 6987: signed32BitValueOf:, signed64BitValueOf: etc. broken Change Set: Interpreter-signed32BitIntegerFor-dtl-M6987 Date: 12 April 2009 Author: David T. Lewis This is a fix for #signed32BitIntegerFor: on 64-bit images. It adds a declaration for the method parameter as type int, such that the passed value is coerced from (possibly 64 bit) sqInt to int. This provides the correct behavior when used on a 64 bit image. This change set should be applied in addition to the changes in Int32Fixes-dtl-M6987. If the 64-bit FFI changes for M7237 are also applied, then the FFITestCase tests in M6987 are all green for 64-bit host and 64-bit image. ------- Name: VMMaker-dtl.119 Author: dtl Time: 1 April 2009, 10:45:59 pm Ancestors: VMMaker-ar.118 VMMaker 3.10.2 Reference Mantis 0007327: Bug in BitBlt: rgbMin and other set alpha to 0 Fix provided by Juan Vuletich (jmv) http://lists.squeakfoundation.org/pipermail/vm-dev/2009-March/002473.html. Confirmed by Andreas Raab http://lists.squeakfoundation.org/pipermail/vm-dev/2009-March/002474.html. ------- Name: VMMaker-ar.118 Author: ar Time: 28 March 2009, 12:27:45 am Ancestors: VMMaker-dtl.117 Updated SocketPlugin for IPv6 primitives. ------- Name: VMMaker-dtl.117 Author: dtl Time: 12 March 2009, 11:15:44 pm Ancestors: VMMaker-dtl.116 VMMaker version 3.10.2 Reference Mantis 0006987: signed32BitValueOf:, signed64BitValueOf: etc. broken Change Set: Int32Fixes-dtl-M6987 Update #signed32BitValueOf: based on Andreas fix plus extra range checks to fail prim on invalid values. Add type cast to #primitiveIntegerAt, required for 64 bit image. Change Set: Int64Fixes-dtl-M6987 Update #signed64BitIntegerFor: and #signed64BitValueOf: to work properly for full range of 64 bit signed integers. Moved several primitives from "sound primitives" to "array primitives", removed category "sound primitives". See IntegerArrayTestCase-dtl-M6987 for tests. ------- Name: VMMaker-dtl.116 Author: dtl Time: 23 February 2009, 9:29:58 am Ancestors: VMMaker-dtl.115 VMMaker 3.10.1 Fix type cast problem in #pushClosureCopyCopiedValuesBytecode. Stored value should be SmallInteger oop, not a C pointer. Change from: (self cCoerce: localIP to: 'sqInt') to: (self oopForPointer: localIP) Fix confirmed by eem ------- Name: VMMaker-dtl.115 Author: dtl Time: 22 February 2009, 3:21:29 pm Ancestors: VMMaker-dtl.113, VMMaker-eem.114 VMMaker 3.10 Update to permit loading in vanilla Squeak. Changes are: - Change #primitiveDoNamedPrimitiveWithArgs to use #var:declareC: rather than pragma. - Remove unreference instance variables in Interpreter. - Reorganize method categories in Interpreter (hopefully as per eem intent). - Restore ordering of class vars in various classes (inadvertently reordered by Montecello). - Update versionString to '3.10'. - Fix comment typo in #activateNewClosureMethod: Note: one additional update will follow, to address type cast issue in #pushClosureCopyCopiedValuesBytecode. ------- Name: VMMaker-eem.114 Author: eem Time: 18 February 2009, 5:42:18 pm Ancestors: VMMaker-dtl.113 Closure-enabled Interpreter. Interpreter supports 5 new bytecodes: (138 pushNewArrayBytecode) (140 pushRemoteTempLongBytecode) (141 storeRemoteTempLongBytecode) (142 storeAndPopRemoteTempLongBytecode) (143 pushClosureCopyCopiedValuesBytecode) and loses (138 143 experimentalBytecode) and gains new closure primitives (200 primitiveClosureCopyWithCopiedValues) (201 205 primitiveClosureValue) (206 primitiveClosureValueWithArgs) (210 primitiveAt) "Compatibility with Cog StackInterpreter Context primitives" (211 primitiveAtPut) "Compatibility with Cog StackInterpreter Context primitives" (212 primitiveSize) "Compatibility with Cog StackInterpreter Context primitives" (221 222 primitiveClosureValueNoContextSwitch) 210-212 are needed for eventual StackVM support sice there-in ContextPart must use special at:, at:put: and size primitives for correct hiding of stack frames. Compatibility for images. Interpreter gains inst var imageFormatVersionNumber which is initialized to the standard 6802/68000 value in Interpreter class>>declareCVarsIn:. If the closure creation bytecode 143 is evaluated imageFormatVersionNumber changes to the closure value 6804/68002. Its value is written in snapshots. So if 143 is used the image version stamp changes and if not, it stays the same. imageFormatForwardCompatibilityVersion defines the enw values and is used in readableImage: to load both curfrent and new (closure) format images. VM gains a primitive unrelated to the closure primitives (218 primitiveDoNamedPrimitiveWithArgs) allows the debugger to evaluate external primitives without resorting to the horrible hack of destructively updating a method in ProtoObject (which is non-reentrant and may mean the debugger can't debug itself). c.f. (118 primitiveDoPrimitiveWithArgs). This primitive uses argumentCountOfMethodHeader: primitiveIndexOfMethodHeader: and roomToPushNArgs: which can profitably be used elsewhere. I'll leave this to you. There is also a reorganization; a much saner categorization in Interpreter. I can't code with the essentially random current one. Hence alas many "source same but rev changed". Best regards Eliot -------

April 11, 2010

Re: [Pharo-project] About getting HashedCollection (another dream happening)

by StÃ©phane Ducasse

Hi levente/igor this hashedCollection is still on my radar. Now I was wondering if I should work first on Set with nil or HashedCollection. Any idea/suggestion? Stef On Mar 14, 2010, at 7:02 PM, Levente Uzonyi wrote: > On Sun, 14 Mar 2010, stephane ducasse wrote: > >> Hi levente and others >> >> I always wanted to have Dictionary not be a subclass of Set and you did it. >> Now when you introduced that in Squeak, we were busy. >> But now I'm so found of this change (like other Smalltalk -> SmalltalkImage current --- which we stopped in the >> middle because lack of momentum and mindsharing) that I would like to integrate it into Pharo. >> Do you have any specific recommandations (like not shooting in our own foot)? > > IIRC this is what I did: > - created a copy of Set named HashedCollection > - changed it's category to Collections-Abstract > - changed Set's superclass to HashedCollection while removed it's instance variables > - removed Set specific methods from HashedCollection > - copied Dictionary specific methods from Set to Dictionary > - changed Dictionary's superclass to HashedCollection > - removed Dictionary specific methods from Set > - cleaup (code that assumes that Dictionary is a Set may be everywhere in the image, for example: Set rehashAllSets won't rehash Dictionaries now, etc) > > But if you change these classes now, you'll have to update AndrÃ©s' changes. That's why I told you earlier to add those changes before touching anything else in the Set hierarchy. The weak dictionary related parts have to be updated already. > > > Levente > >> >> Stef >> >> >> _______________________________________________ >> Pharo-project mailing list >> Pharo-project(a)lists.gforge.inria.fr >> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project > _______________________________________________ > Pharo-project mailing list > Pharo-project(a)lists.gforge.inria.fr > http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project

April 11, 2010

[Pharo-project] news about 1.0

by StÃ©phane Ducasse

Hi guys mariano, marcus and adrian got some push to get 1.0 out. I was just on cc: I think that we made a mistake not to roll that publicly via the mailing-list (we will do that for 1.1). Now this is holidays time here, mariano is going to get married and adrian just moved so no internet connection. Tomorrow I will sync with marcus to know if I can help to get this 1.0 version out. If I remember the mails I saw, marcus was looking for the BitBltPluging binaries for the one click image on other platforms than mac os. Stef

April 11, 2010

Re: [Pharo-project] Open Pharo debugger from a Seaside wallback

by Facundo Vozzi

Yes, it's seems to be that issue but its solution didn't repair my problem. I think that the problem is relationated with a network configuration issue so i will continue trying in that way. Thank you Lukas. Facundo On Sun, Apr 11, 2010 at 5:24 PM, Lukas Renggli <renggli(a)gmail.com> wrote: > This seems to be a known problem: > > > http://permalink.gmane.org/gmane.comp.lang.smalltalk.squeak.seaside/20959 > > Lukas > > 2010/4/11 Facundo Vozzi <facundov79(a)gmail.com>: > > Hi people, > > I'm using Pharo RC2 with Seaside 2.8.4 version and it's very slow when > > I trying to debugging a seaside application. In other words, I press the > > debug link on the seaside application walback and it take to Pharo over > 40 > > seconds to open de imagen debugger. I tried with Seaside 3 developer > image > > and it was the same situation. > > I'm using pharo on 32 bits Windows 7 in a Core 2 Duo and 2 GB ram memory. > > Has anybody experimented the same problem or behavior? > > Tranks, > > Facundo > > ps: Please, excuse my poor english. > > > > _______________________________________________ > > Pharo-project mailing list > > Pharo-project(a)lists.gforge.inria.fr > > http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project > > > > > > -- > Lukas Renggli > www.lukas-renggli.ch > > _______________________________________________ > Pharo-project mailing list > Pharo-project(a)lists.gforge.inria.fr > http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project > -- Facundo Vozzi InfOil S.A. Project Leader (+54-11) 4542-9999 x108 fvozzi(a)infoil.com.ar

April 11, 2010

Re: [Pharo-project] Open Pharo debugger from a Seaside wallback

by Lukas Renggli

This seems to be a known problem: http://permalink.gmane.org/gmane.comp.lang.smalltalk.squeak.seaside/20959 Lukas 2010/4/11 Facundo Vozzi <facundov79(a)gmail.com>: > Hi people, > I'm using Pharo RC2 with Seaside 2.8.4 version and it's very slow when > IÂ trying to debugging a seaside application. In other words, I press the > debugÂ link on the seaside application walback and it take to Pharo over 40 > seconds to open de imagen debugger. I tried with Seaside 3 developer image > and it was the same situation. > I'm using pharo on 32 bits Windows 7 in a Core 2 Duo and 2 GB ram memory. > Has anybody experimented the same problem or behavior? > Tranks, > Facundo > ps: Please, excuse my poor english. > > _______________________________________________ > Pharo-project mailing list > Pharo-project(a)lists.gforge.inria.fr > http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project > -- Lukas Renggli www.lukas-renggli.ch

April 11, 2010

[Pharo-project] [update 1.1] #11312

by StÃ©phane Ducasse

11312 ----- - Issue 2292: add #next:putAll:startingAt: to SocketStream. Thanks pmm - http://code.google.com/p/pharo/issues/detail?id=2292 - Issue 1011: Monticello snapshot browser does not display class side definitions. Thanks lukas. - Fix Code recovery menu: thanks alain - Issue 1908: beginsWithAnyOf: / endsWithAnyOf: stef

April 11, 2010

Re: [Pharo-project] A port of Pharo: Redline Smalltalk ...

by StÃ©phane Ducasse

Richard Nicolas checked systematically that add: method returns their argument. - Issue 2171: add consistency in MC, Exception and other for the tests: this a list of possible tests that would be worth to migrate to pharo. - Issue 2238: tests for FileStream behavior - Issue 1920: Better PureBehaviorTests - Issue 1934: some tests for WeakRegistry Let me know if you need advices. Stef On Apr 11, 2010, at 6:16 PM, Richard Durr wrote: > I'd like to contribute, but I am not good enough for the difficult > ones to come to pleasing solutions, and I somehow don't feel the other > ones accessable for me :/ > > On Sat, Apr 10, 2010 at 9:21 PM, StÃ©phane Ducasse > <stephane.ducasse(a)inria.fr> wrote: >> Hi richard >> >> some changes are changes coming from squeak and in such a case there is the squeak change >> It does not mean that we can integrate the change as it is. >> Thanks for looking at fixes and bug entries >> >> Stef >> >> On Apr 10, 2010, at 7:52 PM, Richard Durr wrote: >> >>> I am still trying to grok how this issue-fixing works. For example, in >>> "Difficulty:Easy" there are some issues which consist mostly of >>> ChangeSets of something. O_o >> >> O_o >> >> Nice :) >> >> >>> >>> On Tue, Apr 6, 2010 at 10:50 AM, StÃ©phane Ducasse >>> <stephane.ducasse(a)inria.fr> wrote: >>>> Yes this is the vision we have. >>>> Now if people would help for simple things in pharo we would have more time for such issues. >>>> We got several people doing PhD on modules and related. We studied namespace and other kind of >>>> points so we have an idea of what could be a solution but it should be implemented >>>> tested..... >>>> >>>> Stef >>>> >>>> >>>> On Apr 6, 2010, at 1:14 AM, Richard Durr wrote: >>>> >>>>> No, it's not. I would be far more valuable to have modularized images >>>>> though: Just like Next/Apples Nibs or Bundle-mechanism. These >>>>> image-modules would be edited from a development-image-bundle. They >>>>> could be used to form stand-alone applications already clean of the >>>>> development code or libraries that contain only a diff to the base >>>>> classes, these library-images could then be combined to form more >>>>> complex images. They could also contain their own namespaceâ¦ and they >>>>> would replace the non-modelling packaging mechanism, but I digressâ¦ ^^ >>>>> >>>>> | packageWorkedOn | >>>>> packageWorkedOn := Browser newPackageNamed: 'Calculator'. >>>>> packageWorkedOn importPackage: 'Seaside'. >>>>> "Create code in the Browserâ¦" >>>>> packageWorkedOn saveAsMonticello. >>>>> packageWordedOn saveAsPackage. >>>>> packageWorkedOn saveAsApplication. >>>>> >>>>> or something like that â¦ ^^ >>>>> >>>>> On Mon, Apr 5, 2010 at 9:08 PM, nullPointer <epicfan(a)gmail.com> wrote: >>>>>> >>>>>> jarober says: >>>>>> >>>>>>> The reason these "Smalltalk for .NET" and "Smalltalk for the JVM" >>>>>>> projects never seem to come off is simple - Smalltalk isn't just flat >>>>>>> text in an editor. Smalltalk is the entire interactive environment. >>>>>>> It would be fairly simple to get a syntax parser, but it wouldn't be >>>>>>> Smalltalk. It would be Ruby or Python with Smalltalk syntax. >>>>>>> Somewhat useful perhaps, but not really Smalltalk. >>>>>> >>>>>> :( >>>>>> >>>>>> Hopefully it would have sufficient knowledge to be able to do something >>>>>> thus. >>>>>> >>>>>> Sorry, but I believe that it is the way that had to take commercial >>>>>> Smalltalk. It surprises to me that Cincom and company do not see it. >>>>>> >>>>>> Regards. >>>>>> -- >>>>>> View this message in context: http://n4.nabble.com/A-port-of-Pharo-Redline-Smalltalk-tp1751884p1751963.ht… >>>>>> Sent from the Pharo Smalltalk mailing list archive at Nabble.com. >>>>>> >>>>>> _______________________________________________ >>>>>> Pharo-project mailing list >>>>>> Pharo-project(a)lists.gforge.inria.fr >>>>>> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project >>>>>> >>>>> >>>>> _______________________________________________ >>>>> Pharo-project mailing list >>>>> Pharo-project(a)lists.gforge.inria.fr >>>>> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project >>>> >>>> >>>> _______________________________________________ >>>> Pharo-project mailing list >>>> Pharo-project(a)lists.gforge.inria.fr >>>> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project >>>> >>> >>> _______________________________________________ >>> Pharo-project mailing list >>> Pharo-project(a)lists.gforge.inria.fr >>> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project >> >> >> _______________________________________________ >> Pharo-project mailing list >> Pharo-project(a)lists.gforge.inria.fr >> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project >> > > _______________________________________________ > Pharo-project mailing list > Pharo-project(a)lists.gforge.inria.fr > http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project

April 11, 2010

Re: [Pharo-project] add #next:putAll:startingAt: to SocketStream

by StÃ©phane Ducasse

thanks for the reminder Philippe I just integrated your fixes. Could you have a look at the entry pointed by levente? Stef On Apr 11, 2010, at 9:46 PM, Levente Uzonyi wrote: > On Sun, 11 Apr 2010, Philippe Marschall wrote: > >> Hi >> >> While I was optimizing some Seaside code I found out that SocketStream >> does not understand #next:putAll:startingAt: [1]. It's quite a helpful >> method that allows one to write a subcollection of a collection to a >> stream without making a copy. In our case it would allow the codec to >> directly write to the socket stream. >> >> Anyway what I wanted to say is this is fixed in Network-Kernel-pmm.29 ;-) >> >> [1] http://code.google.com/p/pharo/issues/detail?id=2292 > > FYI: http://code.google.com/p/pharo/issues/detail?id=2244 > > > Levente > >> >> Cheers >> Philippe >> >> >> _______________________________________________ >> Pharo-project mailing list >> Pharo-project(a)lists.gforge.inria.fr >> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project >> > > _______________________________________________ > Pharo-project mailing list > Pharo-project(a)lists.gforge.inria.fr > http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project

April 11, 2010

Re: [Pharo-project] add #next:putAll:startingAt: to SocketStream

by Levente Uzonyi

On Sun, 11 Apr 2010, Philippe Marschall wrote: > Hi > > While I was optimizing some Seaside code I found out that SocketStream > does not understand #next:putAll:startingAt: [1]. It's quite a helpful > method that allows one to write a subcollection of a collection to a > stream without making a copy. In our case it would allow the codec to > directly write to the socket stream. > > Anyway what I wanted to say is this is fixed in Network-Kernel-pmm.29 ;-) > > [1] http://code.google.com/p/pharo/issues/detail?id=2292 FYI: http://code.google.com/p/pharo/issues/detail?id=2244 Levente > > Cheers > Philippe > > > _______________________________________________ > Pharo-project mailing list > Pharo-project(a)lists.gforge.inria.fr > http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project >

April 11, 2010

[Pharo-project] Open Pharo debugger from a Seaside wallback

by Facundo Vozzi

Hi people, I'm using Pharo RC2 with Seaside 2.8.4 version and it's very slow when I trying to debugging a seaside application. In other words, I press the debug link on the seaside application walback and it take to Pharo over 40 seconds to open de imagen debugger. I tried with Seaside 3 developer image and it was the same situation. I'm using pharo on 32 bits Windows 7 in a Core 2 Duo and 2 GB ram memory. Has anybody experimented the same problem or behavior? Tranks, Facundo ps: Please, excuse my poor english.

April 11, 2010