Planet Squeak Code

Name: Collections-topa.809
Author: topa
Time: 28 October 2018, 8:16:44.149252 pm
UUID: 65e8b64d-f325-47d8-8f83-ab4a9793ba9f
Ancestors: Collections-eem.808

Add #histogramOf: convenience + test

by Dave Lewis (lewis@mail.msen.com) at December 09, 2018 01:52 PM

Name: Collections-eem.808
Author: eem
Time: 22 October 2018, 4:54:43.989805 pm
UUID: c6b15e95-b365-4ea0-8211-71ec7f55084b
Ancestors: Collections-ul.807

Faster ByteString/ByteArray hashing. The VM will accept either 1 or 2 args for the primitiveStringHash primitive so it can be used on the instance size, avoiding a little overhead.

by Dave Lewis (lewis@mail.msen.com) at December 09, 2018 01:52 PM

Se agrego testQuick a OrderedCollection

Se clasifico testSection dentro de *WASM

testeando con hntrie.wasm haciendo en Terminal

cd /Users/edgardecleene/Documents/MagallanesSites/SqueakWASM/wabt

wasm2wat hntrie.wasm -o hntrie.wast

Seleccionando el ultimo en SWasmLivingBook e intentando compilar se ve que el tamaño del archivo difiere

Por esto se implemento testQuick para ver rapidamente en que seccion está el problema

December 09, 2018 10:38 AM

empty log message

December 09, 2018 10:24 AM

Let's make basicInspect robust to crazy numbers.

With http://www.squeaksource.com/ArbitraryPrecisionFl.html
I can create crazy numbers like this:
((1 asArbitraryPrecisionFloatNumBits: 100) timesTwoPower: 1e100)

But then I can't print to decimal form (a bit too many bytes to be allocated....).

The problem when debugging such monster is that basicInspect is not basicEnough: it insists on printing the Number in the inspector title (label)!

If we invoke basicInspect, we generally want something basic. Even if we just inspect, having a label spanning more than 100 chars is vain...

So let's be robust.

by Nicolas Cellier (nicolas.cellier.aka.nice@gmail.com) at December 09, 2018 10:17 AM

Add DateAndTimeReferenceStreamTest to verify reference stream serialization for DateAndTime and Date. Tests are written in a traditional Chronology image, and will pass for both Spur and V3 Squeak images. When DateAndTime class is updated for UTCDateAndTime, saved instances must be compatible with the original serialization format because serialiized instances are used extensively in Monticello, and compatibility with saved sources is required. Therefore it is necessary that these tests pass after installation of UTCDateAndTime.

by David T. Lewis (lewis@mail.msen.com) at December 09, 2018 01:28 AM

Allow for nil offset in DateAndTime>>readDataFrom:size: as may be the case in 'browse origin' when seaching a 'browse revisions' version history. If offset is nil, set localOffsetSeconds to zero.

by Dave Lewis (lewis@mail.msen.com) at December 09, 2018 12:04 AM

Fix reference stream serialization for DateAndTime in UTC format. Instances are always serialized in a cononical format corresponding to the traditional implementation with four instance variables. When materializing, alway read the cononical data and initiaize to the new format with two instance variables.

by Dave Lewis (lewis@mail.msen.com) at December 09, 2018 12:03 AM

Verify reference stream serialization for DateAndTime and Date. Tests are written in a Cronology-Classic image and should pass in a Chronology-UTC image.

by Dave Lewis (lewis@mail.msen.com) at December 08, 2018 11:27 PM

FIx to avoid project view windowl hiding while displaying a server check dialog.

December 08, 2018 12:42 PM

Correct a typo in asMinimalDecimalFraction comment

December 07, 2018 09:32 PM

Use a LRUCache for pi and (2 ln) because they are frequently used.

December 07, 2018 09:31 PM

Categorizes uncategorized messages in the Sound package (1250 methods to go).

by Patrick Rein (patrick.rein@gmail.com) at December 07, 2018 03:27 PM

Categorizes unclassified methods in the Tests package, also recategorizes some test methods from testing to tests but not all (another door of the message categories advent calendar)

by Patrick Rein (patrick.rein@gmail.com) at December 07, 2018 12:27 PM

Categorizes the methods of MCToolWindowBuilder as they were still missing.

by Patrick Rein (patrick.rein@gmail.com) at December 07, 2018 12:22 PM

More advances on categorizing all the messages. This time the Monticello classes (except for one class for which I am really unsure what the messages are actually for).

by Patrick Rein (patrick.rein@gmail.com) at December 07, 2018 08:58 AM

...and fix the signature of alloca.

by Eliot Miranda (eliot.miranda@gmail.com) at December 06, 2018 11:13 PM

Tidy up the SerialPlugin. FIx a bug in primitiveSerialPortCloseByName:

by Eliot Miranda (eliot.miranda@gmail.com) at December 06, 2018 11:03 PM

Don't hide the ProjectViewMorph and it's owner when loading

December 06, 2018 08:13 PM

Similar to the commit in Multilingual, this commit categorizes un-categorized methods and moves tests from testing to tests.

by Patrick Rein (patrick.rein@gmail.com) at December 06, 2018 07:26 PM

In the spirit of a more approachable Squeak this commit classifies all unclassified methods in Multilingual and re-categorizes some misplaced methods. There are also two tiny refactorings in this commit.

by Patrick Rein (patrick.rein@gmail.com) at December 06, 2018 07:22 PM

Added tests for OrderedSet forged from SetTest and OrderedDictionaryTest.

by Levente Uzonyi (leves@elte.hu) at December 05, 2018 06:12 PM

Simple OrderedSet based on OrderedDictionary's implementation, but the order instance variable is an OrderedCollection, which simplifies a few things and makes #removeFirst and #removeLast O(1) (amortized).

by Levente Uzonyi (leves@elte.hu) at December 05, 2018 06:12 PM

empty log message

December 04, 2018 04:47 PM

Fixes categories for point tests. Adds tests for infix rectangle creation.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at December 02, 2018 10:52 AM

Adds an infix version for Rectangle >> #center:extent: for convenience.

The name #inflate: is subject to discussion. :-)

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at December 02, 2018 10:50 AM

Deprecations from Morphic-mt.1468

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at December 02, 2018 10:11 AM

Two minor fixes in the realm of BorderedMorph:
1) Fixes a small issue with translucent borders.
2) Deprecates two set* messages.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at December 02, 2018 10:09 AM

- Provide #asChronologySeconds for compatibility with old Chronology-format applications.
- Let #ticks:offset: account for possibility of DateAndTime's with nil 'offset'.

by Chris Muller (ma.chris.m@gmail.com) at December 02, 2018 02:21 AM

empty log message

December 01, 2018 09:21 PM

Recategorized tests into the tests category (sorry could not resist). No other changes.

by Patrick Rein (patrick.rein@gmail.com) at December 01, 2018 04:30 PM

Refactors the FileContentsBrowserTest to not copy the browser creation and adds a test case for browsing change sets which does not yet show the problematic behavior.

by Patrick Rein (patrick.rein@gmail.com) at December 01, 2018 04:23 PM

A minor refactoring of FileContentsBrowser instance creation methods in preparation of a testing and debugging of missing methods when browsing change set files.

by Patrick Rein (patrick.rein@gmail.com) at December 01, 2018 04:18 PM

StackInterpreter:
Fix a bug where a reference in a married context in a base frame woudl prevent garbage collection. The same issue is fixed for normal marriage/divorce of contexts, but was not handled in makeBaseFrameFor:. Thanks to Ryan Macnak for identifying both bug and fix.

Fikx a typo. Recategorise some tests and add a test for the iussue above.

by Eliot Miranda (eliot.miranda@gmail.com) at December 01, 2018 12:02 AM

- Changes to be able to materialize legacy DateAndTime's even after loading Dave Lewis' UTCDateAndTime. This required addition of yet another post-materialization hook, #maPostMaterializeFrom:using:. That hook is called every time (so no need to also implement #maWantsPostMaterialization) with an empty implementation on ProtoObject, so I also don't think we'll need #maPostMaterialize anymore either, since they can now simply implement the new one (and use the arguments or not).
Further, even ones that initialized in #setUpPreAndPostProcessing can use this method.
- This version saves DateAndTime's as a String (as I wish I'd have done from the beginning), making it incompatible with legacy serialized DateAndTime's. I will now attempt another iteration that saves in the legacy format.

by Chris Muller (ma.chris.m@gmail.com) at November 30, 2018 02:43 AM

Fix a bug in SHMCClassDefinition>>sharedPools that answered nil as the pool for a missing pool, and hence broke khighlighting in the Monticello package browser when browsing older version s with different pool dictionary definitions.

by Eliot Miranda (eliot.miranda@gmail.com) at November 29, 2018 10:00 PM

#isContext: is sent to self instead of objectMemory

Fix a couple of typo's in:

- CoInterpreter>>ceNonLocalReturn:
- StackInterpreter>>commonReturn

where #isContext: is sent to self instead of objectMemory.

by Eliot Miranda (eliot.miranda@gmail.com) at November 29, 2018 07:38 PM

#isContext: is sent to self instead of objectMemory

Fix a couple of typo's in:

- CoInterpreter>>ceNonLocalReturn:
- StackInterpreter>>commonReturn

where #isContext: is sent to self instead of objectMemory.

by Alistair Grant (akgrant0710@gmail.com) at November 29, 2018 05:42 PM

Remove a mistaken halt from the postscript of the previous commit.

by Eliot Miranda (eliot.miranda@gmail.com) at November 24, 2018 11:19 PM

Redefine LargePositiveInteger hash for compatibility between 32-bit and 64-bit systems.

by Eliot Miranda (eliot.miranda@gmail.com) at November 24, 2018 09:44 PM

Beginning some tests to ensure messages being sent from menus actually exist in default receivers

November 24, 2018 07:50 AM

Copy a method from MVC's ParagraphEditor to Morphic's TextEditor. #printerSetup is sent by StringHolder's shifted yellow button menu and so can be sent to both ParagraphEditor (in MVC) and TextEditor (in Morphic).

November 24, 2018 07:43 AM

Add 'alphabetize' to Package Pane Browser's class category menu, as is found on standard Browser.

November 24, 2018 07:33 AM

- Suggestion that #basicClass should be inlined while #class should be a message send, so that Proxy's can be supported.
- If so, then #xxxClass can be banished.
- With #xxxClass banished, the Squeak code that called it can be written normally, simply as "class".

by Chris Muller (ma.chris.m@gmail.com) at November 24, 2018 05:12 AM

Prevent DockingBar menu item labels from being edited, by default. This hides three possible defects in current item label editing behavior: the editor appears with its position slightly off, it only allows one keypress before automatically accepting the input, and if a menubar icon is shift-clicked then it is erased and replaced by text input by the user. Note that the menubar can be easily restored to its default contents by clicking Extras->Rebuild Menus, or evaluating: TheWorldMainDockingBar updateInstances

November 23, 2018 02:23 AM

add the main repository url (in-place edit of 3.0 baseline)

November 22, 2018 06:07 PM

Sista:
Make non-local return safe in the presence of Sista oiuter-context-less blocks; i.e. throw cannotReturn: when a home context cannot be found instead of crashing.

Simulation:
Fix Spur's go-faster memcpy:_:_: for OSProcessPlugin>>#copyBytesFrom:to:length:

by Eliot Miranda (eliot.miranda@gmail.com) at November 21, 2018 12:14 AM

Define #= and #hash for Message such that distinct Messages with the identical selector and lookupClass, and literalEqual: arguments are equal.

by Eliot Miranda (eliot.miranda@gmail.com) at November 20, 2018 09:43 PM

Add tests for Message equality

by Eliot Miranda (eliot.miranda@gmail.com) at November 20, 2018 09:41 PM

Fix ClosureExtractor (embeddedBlockClosures) for exotic Sista blocks.

by Eliot Miranda (eliot.miranda@gmail.com) at November 20, 2018 09:39 PM

Fix a slip in Sista block creation

by Eliot Miranda (eliot.miranda@gmail.com) at November 20, 2018 09:34 PM

Altered Interval>>hash to increase entropy

by Chris Cunningham (cunningham.cb@gmail.com) at November 20, 2018 07:29 PM

Added states and #processXXX: methods for Hello_v3_7 and Response_v3_7 in the protocol selector. Preparation to support v3.7 without the protocol negotiation messages, creating an implicit 3-message handshake based upon ASN1 decoding of the specified header.

by Alan Pinch (alan.c.pinch@gmail.com) at November 20, 2018 06:42 AM

Added states and #processXXX: methods for Hello_v3_7 and Response_v3_7 in the protocol selector. Preparation to support v3.7 without the protocol negotiation messages, creating an implicit 3-message handshake based upon ASN1 decoding of the specified header.

by Robert Withers (robert.withers@protonmail.com) at November 20, 2018 06:42 AM

reorg of standard messages to PT-base

by Alan Pinch (alan.c.pinch@gmail.com) at November 20, 2018 03:56 AM

reorg of standard messages to PT-base

by Robert Withers (robert.withers@protonmail.com) at November 20, 2018 03:56 AM

Merge cmm.31 and cmm.32.

by Chris Muller (ma.chris.m@gmail.com) at November 20, 2018 12:35 AM

- Now supporting #imageThenFilesystem for squeaksource.com.
- Use a SSFilesystem to #loadMonticello:from: even if it isn't the primary storage for maximum speed, since it's just bytes, it's not a DB object.
- If no Magma stores are used in a CompositeStorage, then stay with Squeak ReferenceStream for serializer, otherwise we must use the Ma Serializer to serialize the model retrieved from Magma to a 'data.obj' file since we are using WriteBarrier (to speed up commits) which changed the class of the object.
- Let SSImagePersistence save the image as frequently as every 5 minutes, and make it configurable via #maximumSaveFrequency:.
- Since 48-bit Magma cannot commit more than 4M objects in a single commit, so break the initial Repository creation up into multiple commits. Sigh... Can't wait to upgrade Magma back to 64-bit!
- Now calling SSRepository class>>#shutDown when shutting down. Not sure why this was changed not to do this back in 2015...
- To support ImagePersistence, run.st script now has to check if the image was saved with the server in a running state.

by Chris Muller (ma.chris.m@gmail.com) at November 20, 2018 12:22 AM

Add deny:equals: to complement assert:equals:

by Eliot Miranda (eliot.miranda@gmail.com) at November 19, 2018 05:53 PM

refactor

by Robert Withers (robert.withers@protonmail.com) at November 19, 2018 01:42 PM

refactor

by Alan Pinch (alan.c.pinch@gmail.com) at November 19, 2018 01:42 PM

added and integrated ParrotTalk-v3.7, with a SessionProtocolSelector that distinguishes between v3.6 and v3.7, then manipulates the stack to install the correct SessionOperations for each version.

by Alan Pinch (alan.c.pinch@gmail.com) at November 19, 2018 01:40 PM

added and integrated ParrotTalk-v3.7, with a SessionProtocolSelector that distinguishes between v3.6 and v3.7, then manipulates the stack to install the correct SessionOperations for each version.

by Robert Withers (robert.withers@protonmail.com) at November 19, 2018 01:40 PM

Fix slip in DateAndTime class>>ticks:offset:, do not ignore the offset value.

by Dave Lewis (lewis@mail.msen.com) at November 18, 2018 03:24 PM

completing how to read a Json object

November 18, 2018 11:19 AM

Added how to read a Json object

November 18, 2018 11:15 AM

empty log message

November 18, 2018 10:59 AM

New test to expose critical bug in the proposed UTCDateAndTime.

by Chris Muller (ma.chris.m@gmail.com) at November 18, 2018 04:40 AM

OSProcess moved to be as a dependency of #base (a.k.a., Ma-Core) since it has become generally useful at that level (e.g., DirectoryEntryFile>>#linuxPermissions).

by Chris Muller (ma.chris.m@gmail.com) at November 17, 2018 07:24 PM

switched to using doesFrameEmbedding for thunks.

by Alan Pinch (alan.c.pinch@gmail.com) at November 17, 2018 02:59 PM

switched to using doesFrameEmbedding for thunks.

by Robert Withers (robert.withers@protonmail.com) at November 17, 2018 02:59 PM

switched thunk doesPop and doesPush over to doesFrameEmbedding, a better description of how frames are nested than pop/push, protocol used with the stack thunks.

by Alan Pinch (alan.c.pinch@gmail.com) at November 17, 2018 02:57 PM

switched thunk doesPop and doesPush over to doesFrameEmbedding, a better description of how frames are nested than pop/push, protocol used with the stack thunks.

by Robert Withers (robert.withers@protonmail.com) at November 17, 2018 02:57 PM

Utility methods for determining how much RAM is used, and how much is available to the running image.

by Chris Muller (ma.chris.m@gmail.com) at November 16, 2018 11:53 PM

refactored ParrotThunkMaker

by Robert Withers (robert.withers@protonmail.com) at November 16, 2018 05:04 PM

refactored ParrotThunkMaker

by Alan Pinch (alan.c.pinch@gmail.com) at November 16, 2018 05:04 PM

added default pop substqack return in #popStackUpTo:

by Alan Pinch (alan.c.pinch@gmail.com) at November 16, 2018 05:03 PM

added default pop substqack return in #popStackUpTo:

by Robert Withers (robert.withers@protonmail.com) at November 16, 2018 05:03 PM

Make intervals that are #= have the same hash.
Also, slight adjustment to #=.

by Chris Cunningham (cunningham.cb@gmail.com) at November 16, 2018 06:07 AM

Do NOT move to trunk!
Various option for fixing the internval hash issue related to collectionsTests cbc.296.

Some explanation to go to mailing list - this is so interested folks can try out the code.

by Chris Cunningham (cunningham.cb@gmail.com) at November 16, 2018 05:55 AM

Test for #hash and #= bugs. In anticipation of fixing these.

by Chris Cunningham (cunningham.cb@gmail.com) at November 16, 2018 05:47 AM

Due to the addition of the SendFramesBuffer, replacing the BufferListLayer, now a fully formed stack can be created on the call side, with the RavenTerminal pushed on top of the Session, still supporting early message sending prior to full connection. On the answer side, the creating of the scope creates the RaveenTerminal but it is not yet pushed. That occurs in reactToSession:, which ensures the head of the stack is not the RavenTerminal thunk, priot to pushing on top of the Session.

by Alan Pinch (alan.c.pinch@gmail.com) at November 16, 2018 01:39 AM

Due to the addition of the SendFramesBuffer, replacing the BufferListLayer, now a fully formed stack can be created on the call side, with the RavenTerminal pushed on top of the Session, still supporting early message sending prior to full connection. On the answer side, the creating of the scope creates the RaveenTerminal but it is not yet pushed. That occurs in reactToSession:, which ensures the head of the stack is not the RavenTerminal thunk, priot to pushing on top of the Session.

by Robert Withers (robert.withers@protonmail.com) at November 16, 2018 01:39 AM

Refactored the stack to use a SendFramesBuffer thunk so mjessages can be sent to the connection immediately and are buffered until connection completes. Due to the possibility that there are thunks above the session, as is now the case with Raven, added behavior to #popStackUpTo: to pop a substack of thunks to a specified thunk, as well as a #pushStack:, using #reverse, to push a substack of thunks onto the receiving stack. This behavior allows a stack gaining crypto thunks to pop stack to the session, then pop the SendFramesBuffer thunk, then add the crypto thunks (immigration, cipher, customs, encoder) then push the substack including session back onto the subject stack and then flush the buffer thunk downcall: through the session. Renamed the FrameBuffer to ReceivingFrameBuffer.

by Alan Pinch (alan.c.pinch@gmail.com) at November 16, 2018 01:33 AM

Refactored the stack to use a SendFramesBuffer thunk so mjessages can be sent to the connection immediately and are buffered until connection completes. Due to the possibility that there are thunks above the session, as is now the case with Raven, added behavior to #popStackUpTo: to pop a substack of thunks to a specified thunk, as well as a #pushStack:, using #reverse, to push a substack of thunks onto the receiving stack. This behavior allows a stack gaining crypto thunks to pop stack to the session, then pop the SendFramesBuffer thunk, then add the crypto thunks (immigration, cipher, customs, encoder) then push the substack including session back onto the subject stack and then flush the buffer thunk downcall: through the session. Renamed the FrameBuffer to ReceivingFrameBuffer.

by Robert Withers (robert.withers@protonmail.com) at November 16, 2018 01:33 AM

Adds a test for morph border initialization

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at November 15, 2018 11:16 AM

Fixes small bug in Morph border initialization.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at November 15, 2018 11:14 AM

Name: System-eem.1045
Author: eem
Time: 31 October 2018, 3:38:52.935624 pm
UUID: aece8105-e853-4c18-a38e-6872b4cd1440
Ancestors: System-fn.1044

Sort by name within bins when printing the results of a SpaceTally.

by Dave Lewis (lewis@mail.msen.com) at November 15, 2018 03:36 AM

Name: System-fn.1044
Author: fn
Time: 15 October 2018, 10:11:52.454872 am
UUID: 1339fe15-c8ba-4e80-86ff-eb58ad8ffb56
Ancestors: System-pre.1043

Fixes dot inconsistency.

by Dave Lewis (lewis@mail.msen.com) at November 15, 2018 03:34 AM

Name: Collections-ul.807
Author: ul
Time: 22 October 2018, 9:27:33.883624 pm
UUID: c2efe147-56f8-43ff-a08b-f42390390418
Ancestors: Collections-eem.806

- introduced KeyedSet >> #put: which works like #add:, but replaces existing objects with the argument as Dictionary's #at:put: does
- unified the usage of primitiveStringHash by adding it to ByteSymbol and by using String's implementation as fallback, which works for ByteArrays too

by Dave Lewis (lewis@mail.msen.com) at November 15, 2018 02:25 AM

Name: Compiler-eem.395
Author: eem
Time: 15 October 2018, 12:37:16.203042 pm
UUID: 2f7c8f3a-4e2d-413b-8651-b05b5683f89d
Ancestors: Compiler-eem.394

Implement markerOrNil: for the SistaV1 bytecode set and firm up both implementations so that e.g.
aMethod
self fooClass new
is not interpreted as a marker method (because #new is a special selector and hence does not cayse a literal to be allocated).

by Dave Lewis (lewis@mail.msen.com) at November 15, 2018 02:22 AM

Fixes a minor hick-up from the refactoring of Array>>#with* to brace notation with one point missing in one of the array.

Also categorizes a number of previously uncategorized messages.

by Patrick Rein (patrick.rein@gmail.com) at November 14, 2018 02:41 PM

If an explicit MC version is specified for Installer, load that version. Otherwise load the most recent version of the applicable package or package branch.

Previously, in the case of a repository containing both 'Chronology-Core-dtl.3' and 'Chronology-Core-dtl.30', the following will incorrectly load version 30 rather than version 3.

(Installer ss project: 'UTCDateAndTime') install: 'Chronology-Core-dtl.3'

With this update, the correct version is loaded when explicitly specified, otherwise the most recent version of Chronology-Core (or branch therof) will be loaded.

Unit test coverage is in Tests-dtl.403.

by David T. Lewis (lewis@mail.msen.com) at November 14, 2018 01:05 AM

If an explicit MC version is specified for Installer, load that version. Otherwise load the most recent version of the applicable package or package branch.

by David T. Lewis (lewis@mail.msen.com) at November 14, 2018 01:03 AM

If an explicit MC version is specified for Installer, load that version. Otherwise load the most recent version of the applicable package or package branch.

Previously, in the case of a repository containing both 'Chronology-Core-dtl.3' and 'Chronology-Core-dtl.30', the following will incorrectly load version 30 rather than version 3.

(Installer ss project: 'UTCDateAndTime') install: 'Chronology-Core-dtl.3'

With this update, the correct version is loaded when explicitly specified, otherwise the most recent version of Chronology-Core (or branch therof) will be loaded.

by David T. Lewis (lewis@mail.msen.com) at November 14, 2018 01:01 AM

In the case of a repository containing both 'Chronology-Core-dtl.3' and
'Chronology-Core-dtl.30', the following will incorrectly load version 30 rather
than version 3.

(Installer ss project: 'UTCDateAndTime') install: 'Chronology-Core-dtl.3'

A workaround is to reverse the order of sorting in InstallerMonticello>>mcSortFileBlock, such that 'Chronology-Core-dtl.3' will be detected before 'Chronology-Core-dtl.30' in InstallerMonticello>>mcThing.

A possibly better fix is to change InstallerMonticello>>mcDetectFileBlock: to do string equality checks rather than #beginsWith: comparisons. This appears to be correct, but might have side effects, since I do not know the original rationale for using #beginsWith:.

This change implements the possibly better fix.

by David T. Lewis (lewis@mail.msen.com) at November 14, 2018 01:00 AM

Turn MCDefinitionIndex into a thin wrapper over a KeyedSet and speed it up part #2:

- removed migration code
- categorized all methods
- made use of #definitionsDo:

Bonus:
- replaced uses of Array >> #with:* with brace arrays throughout the package
- replaced Array new with Array empty where possible
- sped up MCScanner a bit

by Levente Uzonyi (leves@elte.hu) at November 13, 2018 11:16 PM

Turn MCDefinitionIndex into a thin wrapper over a KeyedSet and speed it up part #1:

- class side #definitions: allocates a large enough collection to hold the elements in its argument
- introduced #definitionsDo: to avoid unnecessary collection creation when enumerating the definitions
- removed unnecessary class-side #new
- added code which works when the value of definitions is a KeyedSet
- postscript migrates all instances to use KeyedSet for definitions

by Levente Uzonyi (leves@elte.hu) at November 13, 2018 11:14 PM

Begin to phase out 'published' flag unless we can remember what the original purpose of this level of detail in the model was for.

by Chris Muller (ma.chris.m@gmail.com) at November 13, 2018 09:34 PM

Make SMReleaseBrowser support a multi-selection of Squeak Version categories, not just one. The SMServer will still pick just one to be the "primary" Squeak version, which is a questionable behavior.

by Chris Muller (ma.chris.m@gmail.com) at November 13, 2018 09:33 PM

- Allow releases which are not necessarily published to be selected when the UI selection was at the package level.
- Fix the problem of Release scripts seemingly not updating, due to not refreshing the local cache after uploading the script to the server.

by Chris Muller (ma.chris.m@gmail.com) at November 13, 2018 03:32 AM

Update for new OSProcess convenience method.

by Chris Muller (ma.chris.m@gmail.com) at November 12, 2018 11:09 PM

- Replace missing method which broke 'browse revisions'. OOPS!!!
- Improved handling and logging of failed SSFilesystem saves.

by Chris Muller (ma.chris.m@gmail.com) at November 12, 2018 10:28 PM

Eliminate a confusing shadowing in Context>>send:super:numArgs:

by Eliot Miranda (eliot.miranda@gmail.com) at November 12, 2018 03:06 AM

Simplify b3dInitializeRasterizerState, reducing the number of primitiveFailed calls and eliminating cCode:'s. Make B3D_NO_ERROR a class var.

by Eliot Miranda (eliot.miranda@gmail.com) at November 12, 2018 01:35 AM

ensureClassPool is /much/ nicer to use if it answers the class pool

by Eliot Miranda (eliot.miranda@gmail.com) at November 11, 2018 11:08 PM

Let DateAndTime>>asSeconds answer an integer count of seconds, and provide #asExactSeconds to answer seconds since Squeak epoch with available precision retained. An integer seconds count for #asSeconds is consistent with expected usage, and #asExactSeconds represents exact duration as measured in seconds. Thus:

| aDateAndTime |
aDateAndTime := DateAndTime now.
aDateAndTime = (DateAndTime fromSeconds: aDateAndTime asExactSeconds). "==> true"
aDateAndTime = (DateAndTime fromSeconds: aDateAndTime asSeconds). "==> false"

Also remove unreferenced instance variables from ChronologyConstants. These may have been left over from an early refactoring but have been in Squeak since early days. Remove them now.

by Dave Lewis (lewis@mail.msen.com) at November 11, 2018 10:52 PM

CustomHelp subclass: #SRESqueakReverseEngineering

November 10, 2018 05:07 PM

Add accessors for the sendMouseWheelEvents flag that instructs the VM to send mouse wheel events as mouse wheel events, not keynboard arrow keys. Needs an up-to-date VM (and on WIndows needs the VM code to be written).

Fix a typo

by Eliot Miranda (eliot.miranda@gmail.com) at November 10, 2018 04:46 AM

Include description of the flags in VM parameter 65.

by Eliot Miranda (eliot.miranda@gmail.com) at November 10, 2018 04:43 AM

And sendWheelEvents must be global (not in VM's vars struct).

by Eliot Miranda (eliot.miranda@gmail.com) at November 10, 2018 03:21 AM

Add support for a sendWheelEvents flag that persists in the image header and is settable via vmParameterAt: 48 put: ...32...

by Eliot Miranda (eliot.miranda@gmail.com) at November 10, 2018 03:09 AM

Allow an updater to read its update maps from a local directory repository. Previously only HTTP repositories were supported. A local repository may be useful for testing an update stream prior to publishing update maps to a public location.

by David T. Lewis (lewis@mail.msen.com) at November 10, 2018 12:52 AM

MCVersionInspector has an 'Adopt' button to allow the selected version to be added to the parent or parents of a working version. Provide a 'Reparent' button to allow a selected version to become the sole parent of a working version. Adjust defaultExtent to accomodate the additional button.

Motivated by the exercise of preparing to copy Chronology-Core versions from one repository into equivalent Chronology-Core.UTC versions in another repository (possibly trunk). In this scenario, it is helpful to be able to reparent a newly loaded Chronology-Core version from one repository to be the child of the last saved Chronology-Core.UTC version in another. The '.UTC' suffix in this case is intended to branch those versions from the trunk update stream, allowing later merge to trunk with branch history preserved.

by David T. Lewis (lewis@mail.msen.com) at November 10, 2018 12:05 AM

Merge Monticello-dtl.685 from inbox with Monticello-eem.685 in trunk.

by David T. Lewis (lewis@mail.msen.com) at November 10, 2018 12:04 AM

Reviewed the production image serving squeaksource.com to identify any remaining fixes not already applied in this image. Found just two:
- SSImagePersistence (which is no longer used or needed) should be sure to save the image from the UI process, and write an appropriate entry to the log.
- Start and stop the StatisticsThread when starting/stopping the server.

by Chris Muller (ma.chris.m@gmail.com) at November 09, 2018 10:50 PM

Reviewed the production image serving squeaksource.com to identify any remaining fixes not already applied in this image. Found just two:
- SSImagePersistence (which is no longer used or needed) should be sure to save the image from the UI process, and write an appropriate entry to the log.
- Start and stop the StatisticsThread when starting/stopping the server.

by Chris Muller (ma.chris.m@gmail.com) at November 09, 2018 10:48 PM

If an explicit MC version is specified for Installer, load that version. Otherwise load the most recent version of the applicable package or package branch.

by Chris Muller (ma.chris.m@gmail.com) at November 09, 2018 07:57 PM

If an explicit MC version is specified for Installer, load that version. Otherwise load the most recent version of the applicable package or package branch.

by Chris Muller (ma.chris.m@gmail.com) at November 09, 2018 07:46 PM

If an explicit MC version is specified for Installer, load that version. Otherwise load the most recent version of the applicable package or package branch.

Previously, in the case of a repository containing both 'Chronology-Core-dtl.3' and 'Chronology-Core-dtl.30', the following will incorrectly load version 30 rather than version 3.

(Installer ss project: 'UTCDateAndTime') install: 'Chronology-Core-dtl.3'

With this update, the correct version is loaded when explicitly specified, otherwise the most recent version of Chronology-Core (or branch therof) will be loaded.

November 09, 2018 01:14 AM

Slang
Fix mis-handling of trailing boolean on inlining in conditionals when trailing boolean would cause the conditional to be taken. e.g. in
self tryCopyingBitsQuickly ifTrue:
[^nil].
tryCopyingBitsQuickly ends with ^true but the in lining code handled the trailing ^true by taking the fall-through path past the ^nil.

Comnsequently fix a regression in the BitBlt primitive caused by cleanups in VMMaker.oscog-eem.2461.

by Eliot Miranda (eliot.miranda@gmail.com) at November 08, 2018 08:50 PM

- A significant simplification: Personal SqueakSource no longer insists on running as user 'squeaksource', but simply as whatever user is logged in during the installation.
This makes sense since this is "Personal SqueakSource". It greatly simplifies the installation scripts and reduces user impact.
It does requires the scripts to be generated at runtime, however. So the 'exportscripts' command is now part of the setup, however since that requires Squeak, 'configsys' must be run before that to install the squeak.conf file to enable Squeak to run. So, we now only export #configsys and #exportscripts.
- Old repair scripts moved to a deprecated category so scripts won't be generated for them. They're useful examples of first-class Repair's, but no longer needed.
- Updated the adminEmail to today's box-admins address.

by Chris Muller (ma.chris.m@gmail.com) at November 08, 2018 08:47 PM

- Set background process priorities explicitly instead of relative to #activePriority, so it can be understood at compile time.
- Updates to configsys script for new Squeak VM.
- Database-based repositories keep their "MC bytes" in with the rest of the database objects, so they don't need to #save again after having committed the updates to their domain model associated with committing a new MC Version.
This should fix the timeouts we've been experiencing on source.squeak.org especially with the large VMMaker packages.
- Fix for deploying a new Personal SqueakSource when no legacy data.obj is provided (SSRepository class>>#ensureRepositoryIfNoneCreateNewNamed:administeredBy: and SSRepository>>#initialize).

by Chris Muller (ma.chris.m@gmail.com) at November 08, 2018 08:46 PM

- Feature for creating zips. Use ZipArchive>>#maAddFile:as: instead of #addFile:as:, and the original Linux permissions of the zipped files will be preserved when they're unzipped in the target.
Making this feature required being able to get the Linux permissions of the file, which I don't think is possible through the Squeak VM, so I use a new method, DirectoryEntryFile>>#lsOutput, which uses a new convenience method on OSProcess to scrape from an ls -l listing. Kinda crazy, but it works.

by Chris Muller (ma.chris.m@gmail.com) at November 08, 2018 08:03 PM

In the case of a repository containing both 'Chronology-Core-dtl.3' and
'Chronology-Core-dtl.30', the following will incorrectly load version 30 rather
than version 3.

(Installer ss project: 'UTCDateAndTime') install: 'Chronology-Core-dtl.3'

A workaround is to reverse the order of sorting in InstallerMonticello>>mcSortFileBlock, such that 'Chronology-Core-dtl.3' will be detected before 'Chronology-Core-dtl.30' in InstallerMonticello>>mcThing.

A possibly better fix is to change InstallerMonticello>>mcDetectFileBlock: to do string equality checks rather than #beginsWith: comparisons. This appears to be correct, but might have side effects, since I do not know the original rationale for using #beginsWith:.

This change implements the possibly better fix.

November 08, 2018 02:24 AM

Convert Ctrl-left/right to mouse wheel events to fix
"VM sending keystrokes on mouse-wheel?"
http://forum.world.st/VM-sending-keystrokes-on-mouse-wheel-td5088162.html
using Tom Beckmann's idea.

Use the generated wheeelLeft and wheelRight
events in ScrollPane to scroll sideways if horizontal
scrollbars are not suppressed by
alwaysHideHScrollbar preference.

November 07, 2018 08:33 PM

Robustness for missing times in time stamps, needed to see the history of certain Balloon packages.

by Eliot Miranda (eliot.miranda@gmail.com) at November 07, 2018 02:35 AM

Replace deprecated clone with shallowCopy.

by Eliot Miranda (eliot.miranda@gmail.com) at November 07, 2018 01:24 AM

Replace deprecated clone with shallowCopy

by Eliot Miranda (eliot.miranda@gmail.com) at November 07, 2018 01:22 AM

Allow an updater to read its update maps from a local directory repository. Previously only HTTP repositories were supported. A local repository may be useful for testing an update stream prior to publishing update maps to a public location.

November 06, 2018 06:05 PM

MCVersionInspector has an 'Adopt' button to allow the selected version to be added to the parent or parents of a working version. Provide a 'Reparent' button to allow a selected version to become the sole parent of a working version. Adjust defaultExtent to accomodate the additional button.

Motivated by the exercise of preparing to copy Chronology-Core versions from one repository into equivalent Chronology-Core.UTC versions in another repository (possibly trunk). In this scenario, it is helpful to be able to reparent a newly loaded Chronology-Core version from one repository to be the child of the last saved Chronology-Core.UTC version in another. The '.UTC' suffix in this case is intended to branch those versions from the trunk update stream, allowing later merge to trunk with branch history preserved.

by David T. Lewis (lewis@mail.msen.com) at November 06, 2018 03:52 AM

ThreadedFFIPlugin:
Fix a bug passing floats on the stack on X64. The original code passed stacked floats as doubles.
Mark ffiPush*Float:in: as <inline: #always> since they are inined and this eliminates unused functions.
Add some commentary to stack alignment methods to aid locating associated defines.

by Eliot Miranda (eliot.miranda@gmail.com) at November 05, 2018 06:32 PM

Print the matrix so one can see the rows and columns

November 04, 2018 03:41 PM

Replace #asOop with #identityHash

November 03, 2018 08:36 AM

Moves deprecated methods from 53Deprecated to 60Deprecated:

Deprecates two drag and drop menu methods. One broke with the naming convention for menu item selectors and the other is unnecessary as the superclass already implements the same behavior.

Deprecates two DummyUIManager methods which are not provided anymore by UIManagers

by Patrick Rein (patrick.rein@gmail.com) at November 01, 2018 03:37 PM

Replaces a deprecated menu item selector with one which sticks to the pattern for menu item selectors.

by Patrick Rein (patrick.rein@gmail.com) at November 01, 2018 02:57 PM

Migrates methods to use the menu method selector which sticks to the existing selector pattern

by Patrick Rein (patrick.rein@gmail.com) at November 01, 2018 02:51 PM

Deprecates two drag and drop menu methods. One broke with the naming convention for menu item selectors and the other is unnecessary as the superclass already implements the same behavior.

by Patrick Rein (patrick.rein@gmail.com) at November 01, 2018 02:49 PM

Changes the drag and drop menu boxes to reflect the migration started long ago from one drag and drop setting to two settings, one for drag and one for drop. Solves the DockingBar context menu confusion.

by Patrick Rein (patrick.rein@gmail.com) at November 01, 2018 02:44 PM

Deprecates two DummyUIManager methods which are not provided anymore by UIManagers

by Patrick Rein (patrick.rein@gmail.com) at November 01, 2018 02:32 PM

Adds a test checking for classes handling ProvideAnswerNotification although the UIManager should be responsible.

by Patrick Rein (patrick.rein@gmail.com) at November 01, 2018 02:20 PM

Updates the DummyUIManager to answer the ProvideAnswerNotification

by Patrick Rein (patrick.rein@gmail.com) at November 01, 2018 01:57 PM

Removes ProvideAnswerNotification code from several Morphs as it is now handled in the corresponding UIManager

by Patrick Rein (patrick.rein@gmail.com) at November 01, 2018 01:51 PM

Refactors the MorphicUIManager to capture and use the ProvideAnswerNotification

by Patrick Rein (patrick.rein@gmail.com) at November 01, 2018 01:48 PM

Refactors the MVCUIManager to capture and use the ProvideAnswerNotification

by Patrick Rein (patrick.rein@gmail.com) at November 01, 2018 01:48 PM

Updates the comments of the UIManager methods and adds a utility method to make use of the ProvideAnswerNotification mechanism.

by Patrick Rein (patrick.rein@gmail.com) at November 01, 2018 01:44 PM

Adds tests for the valueSupplyingAnswer: protocol i n preparation of the new ProvideAnswerNotification handling.

by Patrick Rein (patrick.rein@gmail.com) at November 01, 2018 01:39 PM

Sort by name within bins when printing the results of a SpaceTally.

by Eliot Miranda (eliot.miranda@gmail.com) at October 31, 2018 10:38 PM

Add (how) "To Build A Similar Virtual Machine" information to the VM General tab in the SystemReporter.

by Eliot Miranda (eliot.miranda@gmail.com) at October 31, 2018 09:42 PM

Deprecation update.

by Chris Muller (ma.chris.m@gmail.com) at October 31, 2018 08:42 PM

Deprecation update.

by Chris Muller (ma.chris.m@gmail.com) at October 31, 2018 08:42 PM

Use the integerAt[Put] primitive to implement signedByteAt:[put:]

by Eliot Miranda (eliot.miranda@gmail.com) at October 31, 2018 05:44 PM

Remove ByteArray>>signedByteAt:[put:]. Both are provided in the base in both Squeak and Pharo, and the overrides here are cimbersome.

by Eliot Miranda (eliot.miranda@gmail.com) at October 31, 2018 05:41 PM

Pragmatise the plugins.

by Eliot Miranda (eliot.miranda@gmail.com) at October 31, 2018 04:53 PM

Added

traceRM:levels:
traceRM:levels:withContext:


SRE Execution Tracer. Object>>traceRM:levels: is like Transcript>>show: with the addition of the oop of the writing object and a dump of the stack to a specified depth.

October 31, 2018 04:41 PM

a lot of changes

October 30, 2018 11:37 AM

Simplify stackLightArrayValue:

by Eliot Miranda (eliot.miranda@gmail.com) at October 30, 2018 12:00 AM

fix SDConcurrentDictionary for Squeak 5.2

October 29, 2018 05:21 AM

Worlds package from FONC's tinllizie

October 28, 2018 01:00 PM

empty log message

October 28, 2018 01:00 PM

packages for soon delete

October 28, 2018 01:00 PM

empty log message

October 28, 2018 01:00 PM

empty log message

October 28, 2018 12:59 PM

empty log message

October 28, 2018 12:59 PM

empty log message

October 28, 2018 12:59 PM

empty log message

October 28, 2018 12:58 PM

empty log message

October 28, 2018 12:58 PM

empty log message

October 28, 2018 12:57 PM

empty log message

October 28, 2018 12:57 PM

empty log message

October 28, 2018 12:57 PM

empty log message

October 28, 2018 12:56 PM

empty log message

October 28, 2018 12:55 PM

empty log message

October 28, 2018 12:55 PM

empty log message

October 28, 2018 12:54 PM

empty log message

October 28, 2018 12:54 PM

empty log message

October 28, 2018 12:53 PM

osc support (not complete yet)

October 28, 2018 12:53 PM

empty log message

October 28, 2018 12:51 PM

Start time for a Date created from a DateAndTime should preserve local time offset

by Dave Lewis (lewis@mail.msen.com) at October 23, 2018 02:15 AM

init commit 2

by Max Muchnik (usermax83@gmail.com) at October 21, 2018 03:57 PM

init commit

by Max Muchnik (usermax83@gmail.com) at October 21, 2018 03:48 PM

Specify stations

by Spencer Ward (spencerkward@gmail.com) at October 19, 2018 07:58 AM

UnixOSProcessPluginSimulator

Provides a minimal implementation to allow UnixOSProcessPlugin>>primitiveGetCurrentWorkingDirectory to be used in the VM simulator.

Also:

- add error checking and fail if object allocation fails in #primitiveGetCurrentWorkingDirectory.
- Use the new C method naming (getcwd:_:)

by Alistair Grant (akgrant0710@gmail.com) at October 19, 2018 05:25 AM

Remove null implementations of initialiseModule and/or shutdownModule. The loader/unloader invokes these only if they exist, consequently null versions are simply a waste of time and space.

by Eliot Miranda (eliot.miranda@gmail.com) at October 18, 2018 11:22 PM

Remove null implementations of initialiseModule and/or shutdownModule. The loader/unloader invokes these only if they exist, consequently null versions are simply a waste of time and space.

by Eliot Miranda (eliot.miranda@gmail.com) at October 18, 2018 11:21 PM

Remove null implementations of initialiseModule and/or shutdownModule. The loader/unloader invokes these only if they exist, consequently null versions are simply a waste of time and space.

by Eliot Miranda (eliot.miranda@gmail.com) at October 18, 2018 11:21 PM

Find station code from name

by Spencer Ward (spencerkward@gmail.com) at October 18, 2018 06:28 PM

Search finds cheapest ticket on each day.

by Spencer Ward (spencerkward@gmail.com) at October 18, 2018 03:30 PM

Search over long period

by Spencer Ward (spencerkward@gmail.com) at October 18, 2018 12:50 PM

Ticket object introduced

by Spencer Ward (spencerkward@gmail.com) at October 18, 2018 12:24 PM

Another #allItalics fix.

October 18, 2018 11:20 AM

Calculate cheapest ticket.

by Spencer Ward (spencerkward@gmail.com) at October 18, 2018 11:11 AM

Added to more in place replacements for the missing

Text>>allItalics

method.

text addAttribute: TextEmphasis italic from: 1 to: text size "allItalics"

Later if Squeak has #allItalics this should be reverted.

October 18, 2018 11:03 AM

Introduce SearchResult object to wrap journies.

by Spencer Ward (spencerkward@gmail.com) at October 18, 2018 10:19 AM

Trainline search returns Journey objects

by Spencer Ward (spencerkward@gmail.com) at October 18, 2018 09:26 AM

update to stable v3

October 18, 2018 08:56 AM

Minor fix to MDCacheClientBase>>#get: (use #ifNotNil: instead of #ifNotNilDo:) to make all tests green again

October 18, 2018 08:54 AM

Release on SM

October 17, 2018 07:33 PM

Creation as Monticello Package

October 17, 2018 07:30 PM

Patches for Squeak 5.2.

by Chris Muller (ma.chris.m@gmail.com) at October 17, 2018 06:14 PM