Planet Squeak Code

- removed SHA256Plugin because SHA2Plugin covers SHA256

by Levente Uzonyi (leves@elte.hu) at March 24, 2020 12:11 AM

Fixed several potential buffer overflow errors. And other minor tweaks.

DESPlugin:
- validate the size of the arguments of the primitives
- use primitive failure codes
- use #methodReturnBool: to return booleans

DSAPlugin:
- validate the numebr of arguments of the primitives
- always return after primitive failure (ignored argument validation)
- use primitive failure codes
- use #methodReturnBool: to return booleans

MD5Plugin:
- validate the size of the arguments of the primitives
- not a SmartSyntaxInterpreterPlugin anymore
- moved endianness check into #decode:sizeInIntegers:
- added inline pragmas
- use #byteSwap32 in #decode:sizeInIntegers:
- removed unused variable: swap
- changes may have broken simulation

SHA2Plugin:
- use #stSizeOf: instead of #slotSizeOf:
- assign a value to w in smalltalk in #sha256ProcessBuffer:updatingHash: and #sha512ProcessBuffer:updatingHash: to avoid the compiler complaining about unassigned variables and to help simluation

by Levente Uzonyi (leves@elte.hu) at March 24, 2020 12:11 AM

Fix bad bad copy/past error in generateSignedShiftRight:on:indent:

This broke B2DPlugin and probably other things (see senders of signedBitShift: and >>>)

by Nicolas Cellier (nicolas.cellier.aka.nice@gmail.com) at March 23, 2020 11:41 PM

Relocated CipherThunkMaker and EncoderThunk to ThunkStack.

by Robert Withers (robert.withers@protonmail.com) at March 23, 2020 08:57 PM

Relocated CipherThunkMaker and EncoderThunk to ThunkStack.

by Robert Withers (robert.withers@protonmail.com) at March 23, 2020 08:57 PM

Relocated CipherThunkMaker and EncoderThunk to ThunkStack.

by Robert Withers (robert.withers@protonmail.com) at March 23, 2020 08:56 PM

Relocated CipherThunkMaker and EncoderThunk to ThunkStack.

by Robert Withers (robert.withers@protonmail.com) at March 23, 2020 08:56 PM

moved SessionIdentity to ThunkStack

by Robert Withers (robert.withers@protonmail.com) at March 23, 2020 08:52 PM

moved SessionIdentity to ThunkStack

by Robert Withers (robert.withers@protonmail.com) at March 23, 2020 08:52 PM

moved SessionIdentity to ThunkStack

by Robert Withers (robert.withers@protonmail.com) at March 23, 2020 08:52 PM

moved SessionIdentity to ThunkStack

by Robert Withers (robert.withers@protonmail.com) at March 23, 2020 08:52 PM

reorg

by Robert Withers (robert.withers@protonmail.com) at March 23, 2020 08:47 PM

reorg

by Robert Withers (robert.withers@protonmail.com) at March 23, 2020 08:47 PM

reorg

by Robert Withers (robert.withers@protonmail.com) at March 23, 2020 08:46 PM

reorg

by Robert Withers (robert.withers@protonmail.com) at March 23, 2020 08:46 PM

reorg

by Robert Withers (robert.withers@protonmail.com) at March 23, 2020 08:46 PM

reorg

by Robert Withers (robert.withers@protonmail.com) at March 23, 2020 08:46 PM

Trying to get the protocol selector functional. Am focused on SSL and ParrotTalk exclusively; no protocol seelctor.

by Robert Withers (robert.withers@protonmail.com) at March 23, 2020 06:53 PM

Trying to get the protocol selector functional. Am focused on SSL and ParrotTalk exclusively; no protocol seelctor.

by Robert Withers (robert.withers@protonmail.com) at March 23, 2020 06:53 PM

Trying to get the protocol selector functional. Am focused on SSL and ParrotTalk exclusively; no protocol seelctor.

by Robert Withers (robert.withers@protonmail.com) at March 23, 2020 06:53 PM

Trying to get the protocol selector functional. Am focused on SSL and ParrotTalk exclusively; no protocol seelctor.

by Robert Withers (robert.withers@protonmail.com) at March 23, 2020 06:53 PM

Eliminate infinite dialog loop while seeking author's initials.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at March 23, 2020 07:31 AM

Proposal: Add convenience method #runSimulated to BlockClosure. While this should not be used in production, it can be helpful for exploring and testing the simulation machinery.

March 22, 2020 07:25 PM

Extends tests for parser notifications and tests bugs that were fixed in Morphic-ct.1640 & Compiler-ct.423. Also tests cancel button that was added in Compiler-ct.422.

Depends on Regex-Core-ct.56.

Please review! I see it as a desirable goal to test as much of things like this that caused various bugs in the past. However, it was hard for me to find an appropriate trade-off between maximizing the test coverage and keeping the tests readable, not too prescriptive.

by Christoph Thiede (Christoph.Thiede@student.hpi.uni-potsdam.de) at March 22, 2020 07:19 PM

Fix a slip in the previous commit that only shows up during simulaiton.

by Eliot Miranda (eliot.miranda@gmail.com) at March 22, 2020 02:31 AM

Fixes and refactors interactive selection during parser notifications. In #queryUndefined, do not forget to reset the selection after the parser notification has been skipped. Extract selection-specific logic into #selectFrom:to:during:. See detailed bug report and solution description in Morphic-ct.1640. Load together with Morphic-ct.1640 only.

by Christoph Thiede (Christoph.Thiede@student.hpi.uni-potsdam.de) at March 21, 2020 11:09 PM

Reverts Morphic-ct.1606, which attempted to solve an issue around interactive parser selections on the wrong way and introduced a new regression.

To reproduce this new regression, in an interactive environment (no automatic variable declarations! u could use an empty browser, for example), try to printIt:
x:=6*7
Answer the UnknownVariable dialog with "declare method temp".
Expected output:
| x |
x:=6*7 42
Actual output:
| x |
42x:=6*7

Let me explain:

In Morphic-ct.1606, we reverted any selection change that occured during the interactive compilation. While this is fine for warnings that can be accepted only, by that change we neglected the fact that certain parser notifications actually modify the source and thus need to permanently adjust the selection. This is what we depend on for a correct functionality of #declareTemp:at: and others. See also #correct:from:to:.
Instead, the actual bug occurs in the parser, where #queryUndefined forgets to reset the selection after the UndefinedVariable warning has been skipped. Other error correction tools know how to do this correctly, for example #correctVariable:interval:. This bug will be fixed in straightly the next commit. Sorry for the trouble! Load together with Compiler-ct.423 only.

by Christoph Thiede (Christoph.Thiede@student.hpi.uni-potsdam.de) at March 21, 2020 11:09 PM

Proposal: Allow to cancel compilation if an unused variable is detected

This way it feels more consistent with UndefinedVariable and UndeclaredVariable.

by Christoph Thiede (Christoph.Thiede@student.hpi.uni-potsdam.de) at March 21, 2020 10:07 PM

Recategorize compiler exceptions

In particular, this fixes a very interesting method stamp of ParserNotification >> #setName: :D

by Christoph Thiede (Christoph.Thiede@student.hpi.uni-potsdam.de) at March 21, 2020 07:36 PM

Miscellaneous minor refactoring

- Use #ifError: and correct deprecated [:argWithoutBar] block syntax (see Compiler-ct.417)
- Fix a documetation typo
- Simplify some conditions

by Christoph Thiede (Christoph.Thiede@student.hpi.uni-potsdam.de) at March 21, 2020 07:04 PM

VMMaker.oscog-eem.2729: Allow setting a break block either in the coInterpreter or the cogit (so processor et al are in scope).

by Eliot Miranda (eliot.miranda@gmail.com) at March 19, 2020 03:25 AM

Interpreters: rip out the check alloc filler support in the interests of simplicity. Though useful in theory, this faclity has never been used successfully to find a bug in all the years it's been available (and we can always put it back).
Rip out the debugPrimCallStackOffset support. I can't remember what this was supposed to do. It is confusing.

Spur: Don't update become effect flags for identical oops that will later be filtered out in the loops oiver the arrays. Don't queue a WeakArray for finalization more than once (i.e. if its already in the queue there's no point adding it again).

Cogit: fix genExternalizePointersForPrimitiveCall & genLoadCStackPointersForPrimCall for the SPReg ~= NativeSPReg regime.
Localise code to the primitive generator invocation block in compilePrimitive

Slang: rewrite TAssignmentNode>>emitValueExpansionOn:level:generator: in the same style as the recent rewrite of TReturnNode>>emitValueExpansionOn:level:generator:

Simulation:
Fix simulating leaf calls in teh context of a primitive invocation. ARMv8 has generated cache flushing code which is executed potentilly during a become primitive. Making simulateLeafCallOf: restore stack,frame,and link registers prevents this breaking asserts in handleCallOrJumpSimulationTrap:.

Don't bother to print endless zeros on the rump C stack.

Allow settingt a break block either in teh coInterpreter or the cogit (so processor et al are in scope).

Fix receiver in canBeImmutable:.

Fix strncpy:_:_: for CArray receivers. Correctly simulate getModuleName in e.g. LargeIntegersPlugin.

Most of the above advance productizing the ARMv8 JIT.

by Eliot Miranda (eliot.miranda@gmail.com) at March 19, 2020 03:11 AM

Add abstract accessors for the link register on processors that have it.

by Eliot Miranda (eliot.miranda@gmail.com) at March 19, 2020 02:50 AM

Now that Slang is fixed by VMMaker.oscog-eem.2728 clean up some concretizers that end up using value expansions.

by Eliot Miranda (eliot.miranda@gmail.com) at March 19, 2020 02:29 AM

- fix: make ThirtyTwoBitRegister >> leftRotateBy: work when the argument is negative

by Levente Uzonyi (leves@caesar.elte.hu) at March 19, 2020 01:45 AM

- test ThirtyTwoBitRegister>>leftRotateBy: with negative arguments as well

by Levente Uzonyi (leves@caesar.elte.hu) at March 19, 2020 01:44 AM

- make postscript work when the package is loaded for the first time

by Levente Uzonyi (leves@elte.hu) at March 18, 2020 10:55 PM

I got all the protocol startup messages setup with a #process... and a #send... in the SSLSessionOperations, with the SSL startup protocol. One unique issue is that there are alternative protocol states, depending if the server wants a client certificate (for send and process) and whether the Certificate is Requested. So I am thinking of having several alternate state machines and pull in the appropriate one in #processClientHello:, on the server side, then send the request, receive the certificate and send a verify, I think.

by Robert Withers (robert.withers@protonmail.com) at March 18, 2020 09:09 PM

I got all the protocol startup messages setup with a #process... and a #send... in the SSLSessionOperations, with the SSL startup protocol. One unique issue is that there are alternative protocol states, depending if the server wants a client certificate (for send and process) and whether the Certificate is Requested. So I am thinking of having several alternate state machines and pull in the appropriate one in #processClientHello:, on the server side, then send the request, receive the certificate and send a verify, I think.

by Robert Withers (robert.withers@protonmail.com) at March 18, 2020 09:09 PM

Upgrade stable to 1.7
This introduces a package split of packed Matrix
This fix a few pb in least square problem resolution via SVD.
This fix a bogus dimension check in matrix left divide (A \ B).

March 18, 2020 08:50 PM

Upgrade stable to 1.7
This introduces a package split of packed Matrix
This fix a few pb in least square problem resolution via SVD.
This fix a bogus dimension check in matrix left divide (A \ B).

March 18, 2020 08:50 PM

Push the stable version to 1.13 for squeak 5.x and squeak6.x
Note that the moer thorough struct tests will require a recent VM on 64bit unixy (x86_64 SysV ABI)

This version mainly fix struct alignment, and re-initialization of auto-generated struct accessors when restarting on a different platform.

by Nicolas Cellier (nicolas.cellier.aka.nice@gmail.com) at March 18, 2020 08:23 PM

Push the stable version to 1.13 for squeak 5.x and squeak6.x
Note that the moer thorough struct tests will require a recent VM on 64bit unixy (x86_64 SysV ABI)

This version mainly fix struct alignment, and re-initialization of auto-generated struct accessors when restarting on a different platform.

March 18, 2020 07:24 PM

Fixup least square solveByDivideAndConquerSVD
The size of work arrays was not up to date.
Since we already use a query for lwork, we shall also do it for rwork and iwork.
Don't even invoke ilaenv whenever the query succeed. ilaenv is useful only for manual setting of work area dimensions.

March 18, 2020 07:13 PM

Oups, fixup floorLog:2 does not have same precedence than floorLog2 and require parentheses...

March 18, 2020 06:38 PM

Fixup for Squeak: there is no copyToHeap (not needed) nor floorLog2 in Squeak (not needed either because floorLog: is exact for exact operands).

March 18, 2020 06:33 PM

All the procedures for packed matrix have been moved in Smallapack-Packed

March 18, 2020 06:27 PM

Fixup for leastSquare problem via SVD: the singular values are real

March 18, 2020 06:24 PM

Fix bogus dimension check in matrix leftDivide (A \ B)
Add a withIndexCollect: implementation for Matrix: unlike other collections, the block take 3 parameters, rowIndex, columnIndex, and elementValue.

March 18, 2020 06:21 PM

Split off the tests.

by Robert Withers (robert.withers@protonmail.com) at March 18, 2020 05:57 PM

Split off the tests.

by Robert Withers (robert.withers@protonmail.com) at March 18, 2020 05:57 PM

- use #leftRotateBy: with positive values as Squeak's current version doesn't work with negative values
- postscript recalculates affected Rijndael tables

by Robert Withers (robert.withers@protonmail.com) at March 18, 2020 05:56 PM

checkpoint: setup SSL state machine in SessionOperations class >>#stateMap.

by Robert Withers (robert.withers@protonmail.com) at March 18, 2020 05:17 PM

checkpoint: setup SSL state machine in SessionOperations class >>#stateMap.

by Robert Withers (robert.withers@protonmail.com) at March 18, 2020 05:17 PM

- use #leftRotateBy: with positive values as Squeak's current version doesn't work with negative values
- postscript recalculates affected Rijndael tables

by Levente Uzonyi (leves@elte.hu) at March 18, 2020 04:46 PM

- updated CryptoHashFunctionTest to use the new HashFunction interface to create HashFunction instances
- fixed testHMACSHA512Spec. it was just a copy of testHMACSHA256Spec
- removed MD2 and MD4 tests
- removed #testMD5Implementations and #testSHA256Implementations because they didn't make sense with the new implementors
- removed register tests
- added all tests from the Hasher-Tests package

by Robert Withers (robert.withers@protonmail.com) at March 18, 2020 11:25 AM

Slang:
Fix a bug with value expansions. The original code elided the value[:value:]* send when inlining a literal block evaluation whose block didn't end in return. When dispatchConcretize, and all concretize mehtods invoked there-in were changed to answer the number of bytes of generated code, rather than each concretize mehtod assigning the number of bytes of generated code individually, this bug surfaced, and invalid code was produced. Amazing that this affected only ARMv5.

Nuke an obsolete pragma in an genForwardersInlinedIdenticalOrNotIf:.
Nuke an unsent method in ObjectMemory.

by Eliot Miranda (eliot.miranda@gmail.com) at March 18, 2020 01:34 AM

Adds PackagePaneBrowser fullOnPackage: and refines StandardToolSet browsePackage:.

Try it out! Morph package browse.

by Christoph Thiede (Christoph.Thiede@student.hpi.uni-potsdam.de) at March 17, 2020 07:27 PM

Revise message list menus and make "toggle break on entry" an actual toggle item. Improves multilingual support.

by Christoph Thiede (Christoph.Thiede@student.hpi.uni-potsdam.de) at March 17, 2020 06:48 PM

Proposal: Allow simple change sorters to become become dual sorters, and dual change sorters to be split into two separate change sorters. This commit also refactors DualChangeSorter in general.

by Christoph Thiede (Christoph.Thiede@student.hpi.uni-potsdam.de) at March 16, 2020 08:03 PM

Proposal: Implement #browse on BlockClosure and CompiledCode again.

thisContext sender arguments last browse.
thisContext sender arguments last method browse.

by Christoph Thiede (Christoph.Thiede@student.hpi.uni-potsdam.de) at March 16, 2020 07:47 PM

checkpoint porting...

by Robert Withers (robert.withers@protonmail.com) at March 16, 2020 06:22 PM

checkpoint porting...

by Robert Withers (robert.withers@protonmail.com) at March 16, 2020 06:22 PM

save

by Robert Withers (robert.withers@protonmail.com) at March 16, 2020 01:13 PM

save

by Robert Withers (robert.withers@protonmail.com) at March 16, 2020 01:13 PM

- removed the HA prefix
- generalized HashFunctionTest >> #hashFunctionClass. removed implementors in subclasses
- categorized a method

by Levente Uzonyi (leves@elte.hu) at March 15, 2020 11:44 PM

- removed the HA prefix

by Levente Uzonyi (leves@elte.hu) at March 15, 2020 11:44 PM

- increase test coverage so that only the class-side initializers are not covered by the tests

by Levente Uzonyi (leves@elte.hu) at March 15, 2020 11:44 PM

- HASHA224WithPlugin supports big-endian machines
- fixed the argument names of #primProcessBuffer:updatingHash:
- updated some comments

by Levente Uzonyi (leves@elte.hu) at March 15, 2020 11:44 PM

- validate the size of hashOop in SHA2Plugin >> primitiveSHA256ProcessBufferUpdatingHash

by Levente Uzonyi (leves@elte.hu) at March 15, 2020 11:43 PM

Merge from Kernel-eem.1317.
Add implementation of Behavior>>isImmediateClass for V3
In Object>>instVarAt:put: continue to use primitive 74, not 174 from oscog.

Name: Kernel-eem.1317
Author: eem
Time: 11 March 2020, 1:12:02.344414 pm
UUID: 3755843a-380e-4cbc-a498-b1f1a85f2906
Ancestors: Kernel-eem.1316

Add the core of read-only support, which will provide both for read-only literals and for efficient write-through cacheing to object databases such as gemstone.

This implementation differs slightly from Clément Béra's original implementatin for Pharo in that here ModificationForbidden always uses a mirror (either a Context or a Class) to attempt to modify the read-only
object in retryModification.

by Dave Lewis (lewis@mail.msen.com) at March 15, 2020 05:33 PM

Fix extent/resizing

March 14, 2020 05:59 PM

Actally fix a bogus failing assert over-zealously added to the VMMaker.oscog-eem.2724 commit.

by Eliot Miranda (eliot.miranda@gmail.com) at March 10, 2020 07:40 PM

Fix a bogus failing assert over-zealously added to the previous commit.

by Eliot Miranda (eliot.miranda@gmail.com) at March 10, 2020 07:25 PM

- use <inline: #always> instead of <inline: true> to avoid generating unnecessary functions
- rotateRight*:by:'s amount is usqInt to check for accidental negative values
- use optimized maj computation in buffer processors
- use #byteSwap32 and #byteSwap64 to swap bytes
- copy* methods only support big-endian encoding, because that's only what SHA2 needs
- improved validation of double word primitive arguments
- better method names

by Levente Uzonyi (leves@elte.hu) at March 10, 2020 05:27 PM

- fixed HASHA256WithSHA2PluginTest

by Levente Uzonyi (leves@elte.hu) at March 10, 2020 05:26 PM

- follow up the changes of SHA2Plugin in CryptographyPlugins-ul.19

by Levente Uzonyi (leves@elte.hu) at March 10, 2020 05:26 PM

updating dependencies

by Robert Withers (robert.withers@protonmail.com) at March 10, 2020 12:41 PM

updating dependencies

by Robert Withers (robert.withers@protonmail.com) at March 10, 2020 12:41 PM

updating dependencies

by Robert Withers (robert.withers@protonmail.com) at March 10, 2020 12:41 PM

updating dependencies

by Robert Withers (robert.withers@protonmail.com) at March 10, 2020 12:41 PM

updating dependencies

by Robert Withers (robert.withers@protonmail.com) at March 10, 2020 12:40 PM

updating dependencies

by Robert Withers (robert.withers@protonmail.com) at March 10, 2020 12:40 PM

updating dependencies

by Robert Withers (robert.withers@protonmail.com) at March 10, 2020 12:40 PM

updating dependencies

by Robert Withers (robert.withers@protonmail.com) at March 10, 2020 12:39 PM

updating dependencies

by Robert Withers (robert.withers@protonmail.com) at March 10, 2020 12:39 PM

updating dependencies

by Robert Withers (robert.withers@protonmail.com) at March 10, 2020 12:38 PM

updating dependencies

by Robert Withers (robert.withers@protonmail.com) at March 10, 2020 12:37 PM

updating dependencies

by Robert Withers (robert.withers@protonmail.com) at March 10, 2020 12:37 PM

- ensured common protocol (#processFinalBuffer:bitLength:)
- added SHA512 tests.

by Robert Withers (robert.withers@protonmail.com) at March 10, 2020 12:35 PM

- ensured common protocol (#processFinalBuffer:bitLength:)
- added SHA512 tests.

by Robert Withers (robert.withers@protonmail.com) at March 10, 2020 12:34 PM

SpurMemoryManager/SpurPlanningCompactor:
Fix one bug thrown up by Marcel's EphemeronLink example. The planning compactor failed to update objects in the mournQueue objStack.

Make sure prepareObjStacksForPlanningCompactor is actually invoked (!!).

Refactor a couple of obj stack routines to take an "also do contents" flag, printObjStack:printContents:, relocateObjStackForPlanningCompactor:andContents:, the last one fixing the bug in question.

Refactor runLeakCheckerForFreeSpace: into runLeakCheckerForFreeSpace:ignoring: to allow the most recenty allocated object in clone (the shallowCopy primitive) to be ignored, because its contents are not yet initialized when the free space check is done.

Make sure to run the free space integrity check as part of leak checking.

Fix a comment typo.

Simulator:
StackInterpreter: fix simulaiton of primitiveHighBit on 64-bits.
Make withSimulatorFetchPointerMovedAsideDo: monotonic, allowing the free space checker to be run in the context of the leak checker.

VMMaker initialization, hack fix initialization in the current Squeak Git support context.

by Eliot Miranda (eliot.miranda@gmail.com) at March 10, 2020 04:26 AM

Name: Collections-nice.879
Author: nice
Time: 7 March 2020, 10:03:26.526308 pm
UUID: 0c7a4aea-f6e4-4578-b279-9077302e13cd
Ancestors: Collections-mt.878

Fix copyReplaceAll:with: for a Symbol receiver

self assert: (#at:put: copyReplaceAll: ':' with: '_') = 'at_put_'.

Implementation did try to modify a copy of the Symbol, which was not possible.

by Dave Lewis (lewis@mail.msen.com) at March 10, 2020 12:29 AM

Name: Kernel-nice.1315
Author: nice
Time: 10 March 2020, 1:17:28.155986 am
UUID: 10c5ea97-f473-493c-8598-4bd06316ca6b
Ancestors: Kernel-nice.1314

Fix FullBlockClosure comparison.
Super would compare startpc which is only relevant in super (where startpc point to bytecode sequence in home method).
FullBlockClosure must compare their compiledBlock which live separately.

[:a|1]=[:b|1] = true.
[:a|2]=[:b|1] = false.

by Dave Lewis (lewis@mail.msen.com) at March 10, 2020 12:27 AM

Name: Kernel-nice.1314
Author: nice
Time: 9 March 2020, 12:11:55.571842 am
UUID: e657e32e-7763-49bc-9dec-035556544077
Ancestors: Kernel-nice.1313

Random changes retained by 5.3 release.

- use #substrings or #lines when simpler than #subStrings:
- protect top Context when we want to #findSimilarSender
(I had one such case in Debugger, but forgot which one)
- deprecate Integer>>destinationBuffer: this might belong to a package, but not to Kernel obviously.
- Fix SmallInteger>>digitsAsFloat fallback code, super only deals with magnitude, not sign

by Dave Lewis (lewis@mail.msen.com) at March 10, 2020 12:26 AM

Fix FullBlockClosure comparison.
Super would compare startpc which is only relevant in super (where startpc point to bytecode sequence in home method).
FullBlockClosure must compare their compiledBlock which live separately.

[:a|1]=[:b|1] = true.
[:a|2]=[:b|1] = false.

by Nicolas Cellier (nicolas.cellier.aka.nice@gmail.com) at March 10, 2020 12:17 AM

Fix the right shift: dont convert to usqInt a type longer than usqInt

by Nicolas Cellier (nicolas.cellier.aka.nice@gmail.com) at March 09, 2020 11:27 PM

Revert a hack fix now that Morphic-eem.1638 has fixed the issue properly.

by Eliot Miranda (eliot.miranda@gmail.com) at March 09, 2020 05:23 PM

Fix project publishing; acceptProjectDetails: may be presented with a nil details.

by Eliot Miranda (eliot.miranda@gmail.com) at March 09, 2020 05:22 PM

Compatibility to be able to publish MVC projects.

by Eliot Miranda (eliot.miranda@gmail.com) at March 09, 2020 05:20 PM

Update documentation for vmParameter 48, add documentation for vmParameters 72 to: 74

by Nicolas Cellier (nicolas.cellier.aka.nice@gmail.com) at March 09, 2020 02:06 PM

Deprecate Integer>>destinationBuffer:

No sender, sole implementor, no comment, strange timestamp.
This meets enough criteria.

by Nicolas Cellier (nicolas.cellier.aka.nice@gmail.com) at March 09, 2020 12:27 AM

Random changes retained by 5.3 release.

- use #substrings or #lines when simpler than #subStrings:
- protect top Context when we want to #findSimilarSender
(I had one such case in Debugger, but forgot which one)
- deprecate Integer>>destinationBuffer: this might belong to a package, but not to Kernel obviously.
- Fix SmallInteger>>digitsAsFloat fallback code, super only deals with magnitude, not sign

by Nicolas Cellier (nicolas.cellier.aka.nice@gmail.com) at March 09, 2020 12:11 AM

Fix instructionPrinter for full closure: indent might be un-initialized, so use the message send.

Sacralize EncoderForSistaV1 as the SecondaryBytecodeSetEncoderClass, in order to mitigate the consequences if ever someone had the idea to re-initialize CompiledCode (or subclasses).

by Nicolas Cellier (nicolas.cellier.aka.nice@gmail.com) at March 08, 2020 11:24 PM

Make some tests aware of FullBlockClosure.

by Nicolas Cellier (nicolas.cellier.aka.nice@gmail.com) at March 08, 2020 10:34 PM

Make comparing two near-zero numbers follow the same logic as comparing a near-zero number and zero.

by Stephane Rollandin (hepta@zogotounga.net) at March 08, 2020 02:22 PM

Make comparing two near-zero numbers follow the same logic as comparing a near-zero number and zero.

by Stephane Rollandin (hepta@zogotounga.net) at March 08, 2020 02:22 PM

Fix comparison of FullBlockClosures

Note: it's a bit misleading to get the compiledBlock into the startpc inst.var.
Note that other usage of startpc could as well be protected thru (self startpc).
I did not do it, because all other messages using the inst. var. are redefined in FullBlockClosure.

by Nicolas Cellier (nicolas.cellier.aka.nice@gmail.com) at March 08, 2020 12:24 PM

Raise the requirements for closeTo:

1) we expect a symmetric relation:
if I am close to aNumber, then aNumber is close to me.

2) we expect a respect of basic order relation:
if I am close to the furthest number, I should be close to the nearest too!

by Nicolas Cellier (nicolas.cellier.aka.nice@gmail.com) at March 08, 2020 11:48 AM

Attempt to let some tests pass even if using retractable scroll bars.

Note: the workaround is not ideal, because forcing the preferences triggers allSubInstancesDo: which is too sllooowwww for a Unit TestCase...

by Nicolas Cellier (nicolas.cellier.aka.nice@gmail.com) at March 07, 2020 10:52 PM

Fix recent breakage of MCDependencySorterTest

by Nicolas Cellier (nicolas.cellier.aka.nice@gmail.com) at March 07, 2020 10:20 PM

Let TestIndenting pass whatever custom tab stops.

In some image, I have set the tab stops like that:

(TextConstants at: #DefaultTabsArray) becomeForward: (32 to: 1536 by: 32) asArray.
(TextConstants at: #DefaultMarginTabsArray) becomeForward: ((32 to: 768 by: 32) collect: [:i | {i. i}]).

by Nicolas Cellier (nicolas.cellier.aka.nice@gmail.com) at March 07, 2020 10:15 PM

Revert one change accidentally included in the previous commit. This is a separate update for checking file system space on image save, and may be included in a future commit along with platform changes.

by David T. Lewis (lewis@mail.msen.com) at March 01, 2020 05:00 PM

VMMaker 4.18.1
Compatibility for oscog versions of the C preprocessor directives. Map oscog versions to the original implementations by fudging the argument lists in code generation. Add unit tests to ensure correct generation of directives and conditionals. Allows oscog branch methods to be loaded and generated without modification.

by David T. Lewis (lewis@mail.msen.com) at March 01, 2020 04:10 PM

empty log message

by Rikio Morikawa (rexmori1@mac.com) at March 01, 2020 12:38 PM

Browser printOut was writing chunk format into an html file making it unreadable. It now opens a help topic instead which is more useful. Perhaps, the menu item needs to be renamed to 'Help'.

March 01, 2020 12:01 PM

Fixing LSystem

March 01, 2020 09:54 AM

New upload RougeLike

by Karl Ramberg 2 (karlramberg@gmail.com) at March 01, 2020 09:50 AM

Misc cleanups, remove unreferenced method, eliminate code duplication, fix a comment typo.

by David T. Lewis (lewis@mail.msen.com) at February 29, 2020 11:15 PM

Housekeeping to keep JSCodeGenerator in sync with CCodeGenerator. Remove unused code as per VMMaker-dtl.410. No functional impact for JS.

by David T. Lewis (lewis@mail.msen.com) at February 29, 2020 07:02 PM

Housekeeping. - the MEMORY_ACCESS_IN_IMAGE macro is no longer used in support code for package MemoryAccess, so get rid of it. To use the Slang memory access rather than cpp macros in sqMemoryAccess.h, evaluate "MemoryAccess enable"..

by David T. Lewis (lewis@mail.msen.com) at February 29, 2020 06:55 PM

Name: Collections-ul.877
Author: ul
Time: 27 February 2020, 2:00:22.678898 pm
UUID: df2c680a-81dd-49bf-a1c2-a0fec288ce0f
Ancestors: Collections-mt.876

- recalculate tally in WeakIdentityDictionary >> #noCheckNoGrowFillFrom: to prevent it becoming greater than the maximum number of elements the dictionary can hold

by Dave Lewis (lewis@mail.msen.com) at February 29, 2020 02:46 PM

Name: System-mt.1137
Author: mt
Time: 28 February 2020, 2:26:26.058195 pm
UUID: dc9de28f-3e13-e94d-a21b-1fdbf6ca003a
Ancestors: System-mt.1136

Prepare 5.3rc3 --- Updates year in Squeak license. Properly clean up the clipboard.

by Dave Lewis (lewis@mail.msen.com) at February 29, 2020 02:44 PM

Complements Morphic-ct.1636 (protection against event-handling errors)

Why is this not in Morphic?

by Christoph Thiede (Christoph.Thiede@student.hpi.uni-potsdam.de) at February 29, 2020 12:41 PM

Adds protection against event-handling errors as proposed by Marcel et al. in [1]. I did not note any subjective performance impact, but we might want to write measures/tests for this. Please review!

Minimum example (don't execute this BEFORE loading this commit):

c := Morph newSubclass.
c compile: 'handleMouseMove: evt ^self error'.
m := c new.
m openInHand.

[1] http://forum.world.st/bug-in-a-ToolBuilder-Squeak5-3rc2-tp5112536p5112551.html

by Christoph Thiede (Christoph.Thiede@student.hpi.uni-potsdam.de) at February 29, 2020 12:39 PM

Proposal: Revive #errorOnStep detection. Introduce #setErrorOnStep to keep it (quite) easy to add non-morphic receivers into the steplist (such as ObjectExplorer). Hm, should we maybe define #setErrorOnStep as a method stub on Object?

by Christoph Thiede (Christoph.Thiede@student.hpi.uni-potsdam.de) at February 29, 2020 11:39 AM

Add 'clear' to Transcript window menu.

February 29, 2020 11:19 AM

Proposal: Add pick button to NewColorPickerMorph that allows to pick a color from the Display. With this enhancement, NCPM is finally superior to the old ColorPickerMorph and could now replace the latter by the default preferences.

by Christoph Thiede (Christoph.Thiede@student.hpi.uni-potsdam.de) at February 29, 2020 10:46 AM

Fixes MorphicHierarchy, which was broken because #world is nil before opening the list in the world.

by Christoph Thiede (Christoph.Thiede@student.hpi.uni-potsdam.de) at February 29, 2020 10:33 AM

Proposal: In Morph>>#balloonText, also try to invoke balloonSelector on model.

by Christoph Thiede (Christoph.Thiede@student.hpi.uni-potsdam.de) at February 29, 2020 10:29 AM

Rounding error

February 29, 2020 10:13 AM

Bumps Squeak version to 5.3rc3 --- Yay! :-)

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at February 28, 2020 01:28 PM

Prepare 5.3rc3 --- Clean up event stats during release building.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at February 28, 2020 01:27 PM

Prepare 5.3rc3 --- Updates year in Squeak license. Properly clean up the clipboard.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at February 28, 2020 01:26 PM

Prepare 5.3rc3 --- All those "----" entries (SNAPSHOT, QUIT, etc.) in the changes file should not be of type #doIt, because that can cause Compiler/Parser errors. Change it to #misc instead. We might want to remove any other " if begins with ----" checks later, too.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at February 28, 2020 01:25 PM

Prepare 5.3rc3 --- Fixes some texts in the welcome window.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at February 28, 2020 01:22 PM

- recalculate tally in WeakIdentityDictionary >> #noCheckNoGrowFillFrom: to prevent it becoming greater than the maximum number of elements the dictionary can hold

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at February 28, 2020 01:21 PM

Opaque pointers to B3D primitive objects are stored into a 4-bytes per word WordArray.
We must care of allocating enough words for storing nObjects pointers.

by nicolas cellier (nicolas.cellier.aka.nice@gmail.com) at February 27, 2020 07:58 PM

Fix B3DRasterizerPlugin for 64bits VM.

state.objects are stored into a (4-bytes) WordArray.
Unfortunately, every pointer will require 8-bytes (2 words) on a 64bits VM.

So we have to fix the calculation of state.nObjects.
On the other side (the image side), we will also need to fix the number of words allocated to hold the objects (#initObjects:)

by nicolas cellier (nicolas.cellier.aka.nice@gmail.com) at February 27, 2020 07:53 PM

SmoothStep testing with degreeSin

February 27, 2020 04:32 PM

- recalculate tally in WeakIdentityDictionary >> #noCheckNoGrowFillFrom: to prevent it becoming greater than the maximum number of elements the dictionary can hold

by Levente Uzonyi (leves@caesar.elte.hu) at February 27, 2020 01:01 PM

Name: Collections-mt.876
Author: mt
Time: 26 February 2020, 5:24:00.015527 pm
UUID: 282a3821-8c7b-974f-b426-2cc481055eab
Ancestors: Collections-ul.875

Prepare 5.3rc2 --- Fixes #isEmpty for weak collections.

by Dave Lewis (lewis@mail.msen.com) at February 26, 2020 11:41 PM

Name: Kernel-mt.1306
Author: mt
Time: 26 February 2020, 5:17:44.570527 pm
UUID: c36a6ca0-83a9-8c48-a66e-2e1f683b0391
Ancestors: Kernel-mt.1305

Prepare 5.3rc2 --- Class organization clean-up must be silent to not interfere with other clean-up code such as in ChangeSet.

by Dave Lewis (lewis@mail.msen.com) at February 26, 2020 11:39 PM

Name: Kernel-mt.1305
Author: mt
Time: 25 February 2020, 10:37:58.980786 am
UUID: 46b30358-e624-b44e-a36d-b6775af60202
Ancestors: Kernel-mt.1304

Fixes class-rename issue.

We need a name! Voting time is now:

- RemarkNotification
- LoggingNotification
- CommentNotification
- NoticeNotification (???)
- [You idea goes here.]

by Dave Lewis (lewis@mail.msen.com) at February 26, 2020 11:38 PM

No change, method not present in System.V3

Name: System-mt.1136
Author: mt
Time: 26 February 2020, 5:16:40.861527 pm
UUID: 162c3149-76a8-da4a-96c7-f5088c26dc0f
Ancestors: System-mt.1135

Prepare 5.3rc2 --- Removes obsolete query (autogenerated long time ago)

by Dave Lewis (lewis@mail.msen.com) at February 26, 2020 11:36 PM

Complements System-ct.1137 (clean up Clipboard). System should not depend on EToys.

February 26, 2020 06:47 PM

Prepare 5.3rc2 --- be even more aggressive when cleaning up Clipboard. Privacy. You know :)

February 26, 2020 06:46 PM

Prepare 5.3rc2 --- Renames the two existing change sets to "RootProject" and "HomeProject" to be both readable and file-out compatible. Fixes the (manual) check for undeclared symbols. Make sure that #clearCaches really cleans the code artifacts in the correct order. Adds some documentation for that. Also bumps version from rc1 to rc2.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at February 26, 2020 04:26 PM

Prepare 5.3rc2 --- Fixes #isEmpty for weak collections.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at February 26, 2020 04:24 PM

Prepare 5.3rc2 --- Marks expected failures for some Interval tests. Adds tests about #isEmpty in weak collections.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at February 26, 2020 04:21 PM

Prepare 5.3rc2 --- Makes the test for undeclared symbols more robust and now passing.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at February 26, 2020 04:20 PM

Prepare 5.3rc2 --- During release building, reset keyboard shortcuts for all kinds of editors. More importantly, reset the ChangeText/FindText caches, which might contain sensitive information.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at February 26, 2020 04:19 PM

Prepare 5.3rc2 --- Class organization clean-up must be silent to not interfere with other clean-up code such as in ChangeSet.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at February 26, 2020 04:17 PM

Prepare 5.3rc2 --- Removes obsolete query (autogenerated long time ago)

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at February 26, 2020 04:16 PM

Prepare 5.3rc2 --- Removes invalid reference to universe installer.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at February 26, 2020 04:14 PM

Tidy up change sets for release images. Empty the home project change set and give it a better name.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at February 26, 2020 10:16 AM

Interpreter: Change the signature of the core selector send breakpointing routines from taking the receiver to taking a class tag. (Requires corresponding changes in platforms/Cross/vm/dispdbg.h).
Add an accessor for breakLookupClassTag and add couldBeContext:. Comment fetchClassTagOf:. [This was all to track down a bug with ARMv8 do do with register save/restore across the call in ceScheduleScavengeTrampoline that caused contexts to appear where blocks were expected].

Cogit: rename the accessor for codeToDataDelta to getCodeToDataDelta to allow codeToDataDelta to be defined as 0 in the non DUAL_MAPPED_CODE_ZONE regime. Consequently find and fix a slip in NewspeakCogMethod class>>initialize. Don't attempt to take the address of 0 in the DUAL_MAPPED_CODE_ZONE regime.
Fix genLoadCStackPointer(s) to use NativeSPReg

Slang: add support for even and odd.

Simulator:
Extend machine code breakpojnt framework to allow CogMethodSurrogates to function as breakpoints.
Recategorize the breakpoint routines. Fix a slip in Integer>>addBreakpoint:.

by Eliot Miranda (eliot.miranda@gmail.com) at February 26, 2020 03:19 AM

ARMv8: Fix MoveAwRR & MoveRRAw for SP; SP is not allowed as a souerce or target reg in in LDP/STP, so substitute RISCTempReg.
genAlignCStackSavingRegisters:numArgs:wordAlignment: is a noop given the ARMv8 enforces SP alignment on all SP accesses.
Fix bad bug in genRestoreRegs:.

Add suitable inlining directives and simplify the processor-specific opcode generation routines.

by Eliot Miranda (eliot.miranda@gmail.com) at February 26, 2020 02:34 AM

empty log message

by Ryan Park (ryandavidpark@gmail.com) at February 25, 2020 07:05 PM

empty log message

by Ryan Park (ryandavidpark@gmail.com) at February 25, 2020 06:22 PM

empty log message

by Ryan Park (ryandavidpark@gmail.com) at February 25, 2020 06:20 PM

Bump the image version to Squeak5.3rc1.

by Fabio Niephaus (code@fniephaus.com) at February 25, 2020 02:28 PM

Fixes an issue with debugging a process at hand, which is used when entering MVC as recovery for Morphic. Example: "[ 3 + 4 ] newProcess debug".

After this fix, the strange re-draw glitches are gone. BUT, the debugger opens only after, for example, opening the world menu once. That's related to #addDeferredUIMessage: and not having the right controller at hand in #enterForEmergencyRecovery. Not that important for 5.3.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at February 25, 2020 02:12 PM

Eliminate infinite dialog loop while seeking author's initials.

February 25, 2020 09:59 AM

Fixes a minor regression with the window title font after UI theme changes.

Note that one could have swapped #setDefaultParamters and #replaceBoxes in #applyUserInterfaceTheme (as suggested by ct), BUT #replaceBoxes reads more specific than #setDefaultParameters, so I put it first. Also, the order of those calls should not make a difference.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at February 25, 2020 09:55 AM

Complements Kernel-mt.1305.

Fixes class-rename issue, which occurred due to catching ProgressNotification. And the missing #resume.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at February 25, 2020 09:39 AM

Fixes class-rename issue.

We need a name! Voting time is now:

- RemarkNotification
- LoggingNotification
- CommentNotification
- NoticeNotification (???)
- [You idea goes here.]

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at February 25, 2020 09:38 AM

We did not find a working fix for that bug.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at February 25, 2020 08:36 AM

We removed a dependency. Yay! :-)

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at February 25, 2020 08:35 AM

Extend CogFrameInspector to allow changing the stack pointer, and to force text generation when frame is old and out-of-bounds.

by Eliot Miranda (eliot.miranda@gmail.com) at February 25, 2020 06:18 AM

Robustness in printing register state and decorating assembly.

by Eliot Miranda (eliot.miranda@gmail.com) at February 25, 2020 05:17 AM

- Optimize for system compactness by ensuring the default internal array size of any HashedCollection is not initialized larger than it may ever need to be.
- Let #new: be used to define larger sizes than the minimum, and perform comparably with #new even if the minimum size is specified.

by Chris Muller (ma.chris.m@gmail.com) at February 25, 2020 04:35 AM

Fixes a small bug after dropping a class into a system category. Explicitly select the class to assure the correct system category is selected in the left pane.

by Christoph Thiede (Christoph.Thiede@student.hpi.uni-potsdam.de) at February 24, 2020 08:03 PM

Adds support for dragging methods on a class in a browser

Committed again and replaces Tools-ct.949 due to a stupid slip (forgot one method).

by Christoph Thiede (Christoph.Thiede@student.hpi.uni-potsdam.de) at February 24, 2020 07:55 PM

Adds support for dragging methods on a class in a browser

by Christoph Thiede (Christoph.Thiede@student.hpi.uni-potsdam.de) at February 24, 2020 07:50 PM

Adapt to VMMaker.oscog-eem.2720
Change return value for not-found of methodFor: from 0 to nil.

by Eliot Miranda (eliot.miranda@gmail.com) at February 24, 2020 05:58 PM

Cogit:
Change return value for not-found of methodFor: from 0 to nil.

#define codeToDataDelta as 0 for the non DUAL_MAPPED_CODE_ZONE regime.

Fix followForwardedLiteralsIn:, mapObjectReferencesInMachineCodeForXXX, and storeLiteral:atAnnotatedAddress:using: for the DUAL_MAPPED_CODE_ZONE regime.

Inlcude the method in question in armPrintDualZoneAnomalies output.

Fix printMethodFieldForPrintContext: to always print the method oop frst and the cogMethod, if any, afterwards.

Fix asserts in frameCallerContext: and mapStackPages when a frame context may be forwarded as part of scavenging.

Slang: Fix sizeOfIntegralCType: for the more complex defnition of codeToDataDelta.

by Eliot Miranda (eliot.miranda@gmail.com) at February 24, 2020 05:58 PM

ARMv8: Fix genLoadCStackPointers for the NativeSPReg ~= SPReg regime.

by Eliot Miranda (eliot.miranda@gmail.com) at February 24, 2020 05:00 PM

Minor, non-exhaustive refactorings in the Browser:

- Don't reinvent the shouldCopy wheel in #dropOnMessageCategories:at:
- Improve multilingual support
- Brackets, spaces and use of #ifError:

by Christoph Thiede (Christoph.Thiede@student.hpi.uni-potsdam.de) at February 24, 2020 03:09 PM

Proposal: If no message is selected in a browser, browse senders of class name instead

by Christoph Thiede (Christoph.Thiede@student.hpi.uni-potsdam.de) at February 24, 2020 03:03 PM

Makes "Image Preferences" work again in the system reporter tool by ignoring Deprecation warnings.

This is only a quickfix for the release. We might want to think about whether to mark deprecated preferences visually or maybe even skip them in the report.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at February 24, 2020 02:20 PM

Remove obsoleted instance variables from TransferMorph

by Christoph Thiede (Christoph.Thiede@student.hpi.uni-potsdam.de) at February 24, 2020 01:51 PM

Adds a test for #containsPoint: in PolygonMorph. See the discussion on the mailing list: http://forum.world.st/Bug-in-PolygonMorph-gt-gt-filledForm-td5112218.html

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at February 24, 2020 01:03 PM

ARMv8:
Fix an assert in emitMoveCw:intoR:at:.
Fix rewriteImm[19|26]JumpBefore:target: to not assign to opcode.
Fix a trailing nil that was creeping into assembly decoration.

by Eliot Miranda (eliot.miranda@gmail.com) at February 23, 2020 10:42 PM

empty log message

February 23, 2020 06:42 PM

My Spec tutorial practice for Spec

by Rikio Morikawa (rexmori1@mac.com) at February 23, 2020 09:12 AM

Fixes broken balloon helps on halo handles

Please review.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at February 18, 2020 01:49 PM

Fixes an issue with event dispatching used (only) in halo, which affects the balloon texts for those halo handles. Adds documentation of some tricky parts.

Not sure why MenuMorph is so clingy about its keyboard help. Manually delete that if the menu gets closed.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at February 18, 2020 01:48 PM

Extends and realigns version of #findFirst: and #findLast:.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at February 18, 2020 08:25 AM

Minor spelling/formatting fix in a progress label.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at February 18, 2020 08:12 AM

Small fix in the wizard's labels when installing extra packages. Make all labels appear in the same form, which starts with an uppercase letter and is (more or less) title case.

(Note that, after the release, we might want to re-design #findFeatures in String. That extra #asLowercase destroys information from the original text. But maybe that was intended after all to better cluster features.)

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at February 18, 2020 08:09 AM

Fix bug in LargeInteger division: instantiation of quotient (quo) may fail and thus require a proper guard.

It may happen: I did report some crash when testing huge integer division at http://smallissimo.blogspot.com/2019/05/tuning-large-arithmetic-thresholds.html

Fix a copy/paste glitch JumpCarry/JumpNoCarry in #setsConditionCodesFor:
We only ever use JumpZero so far, so it's benign.

by Nicolas Cellier (nicolas.cellier.aka.nice@gmail.com) at February 18, 2020 07:51 AM

Fix alternative code generation for highBit via BSR for Spur32:

Previous opcode is a Label, so we cannot test if it setsConditionCodesFor: JumpNegative.
We have to force a CmpCq:R:.
For X64, it's OK, previous opcode is an arithmetic shift.

Fix simulation of primitiveHighBit by sending #numSmallIntegerTagBits to the objectMemory which knows this kind of memory layout detail.

While at it, change the order of BSR registers maskReg -> destReg, like any other CogRTLOpcodes, rather than destReg <- maskReg.

by Nicolas Cellier (nicolas.cellier.aka.nice@gmail.com) at February 17, 2020 08:44 PM

Only notify addition/removal of a repo if there actually was a change

by Tony Garnock-Jones (tonyg@leastfixedpoint.com) at February 17, 2020 01:36 PM

GitRepositoryCache clearCaches

by Tony Garnock-Jones (tonyg@leastfixedpoint.com) at February 17, 2020 08:17 AM

GitTaglikeObject contributor

by Tony Garnock-Jones (tonyg@leastfixedpoint.com) at February 17, 2020 08:15 AM

Decode strings as UTF-8 instead of ASCII (!!!).

by Tony Garnock-Jones (tonyg@leastfixedpoint.com) at February 17, 2020 08:15 AM

Save Morphic-PunchedCards-dtl.6 as package PunchedCards to avoid conflict with Morphic package. No changes other than package name.

Name: Morphic-PunchedCards-dtl.6
Author: dtl
Time: 27 August 2004, 1:37:57 pm
UUID: 603455d4-7c3f-4557-81d1-246012050f70
Ancestors: Morphic-PunchedCards-dtl.5

Name: Morphic-PunchedCards-dtl.3
Author: dtl
Time: 23 August 2004, 9:53:16 pm
UUID: 99fb9f01-03ec-4e45-9051-89adafbad992
Ancestors: Morphic-PunchedCards-dtl.2

'From Squeak3.7beta of ''1 April 2004'' [latest update: #5923] on 23 August 2004 at 5:54:49 am'!
Change Set: PunchedCard-new
Date: 21 August 2004
Author: David T. Lewis

Updated Bob Arning's PunchedCard morph for Squeak 3.7, and added a few embellishments.

Recategorized methods
Added #handlesKeyboard:
Added programmable punching
Added #openWith:
Added #punchFromString:
Speeded up Punchy class
Added keyboard handlers for <home>, <end>, left arrow, right arrow

by Dave Lewis (lewis@mail.msen.com) at February 16, 2020 03:28 PM

GitTagLike messageSummary

by Tony Garnock-Jones (tonyg@leastfixedpoint.com) at February 14, 2020 08:33 AM

No change, method was apparently already moved to Tools in this update stream.

Name: Collections-topa.873
Author: topa
Time: 11 February 2020, 1:29:50.690328 pm
UUID: 45abc260-998e-4352-89cb-2e33b4f4b59c
Ancestors: Collections-ul.872

Fix browser-dependency of Collections

by Dave Lewis (lewis@mail.msen.com) at February 13, 2020 11:31 PM