Planet Squeak Code

Eliminate info duplicaton in new interpreter version name.

by Eliot Miranda (eliot.miranda@gmail.com) at November 17, 2019 06:27 PM

Use substrings and lines idiom instead of findTokens:

Note: findTokens: feeded with a CharacterSet (CharacterSet separators or CharacterSet crlf) would be equally efficient as above specialized selectors, just a bit longer to write.

by Nicolas Cellier (nicolas.cellier.aka.nice@gmail.com) at November 17, 2019 03:42 PM

Replace (findTokens: Character separators) with substrings which is both a shorter and faster idiom.

| tmp |
tmp := String allSubInstances.
{
[tmp do: [:e | e substrings]] bench.
[tmp do: [:e | e findTokens: Character separators]] bench.
[tmp do: [:e | e findTokens: CharacterSet separators]] bench.
}.
#(
'1.17 per second. 852 milliseconds per run. 16.22996 % GC time.'
'0.285 per second. 3.51 seconds per run. 5.30746 % GC time.'
'1.18 per second. 849 milliseconds per run. 20.13357 % GC time.')

by Nicolas Cellier (nicolas.cellier.aka.nice@gmail.com) at November 17, 2019 02:49 PM

Test a few CharacterSet union:/intersection: edge cases

by Nicolas Cellier (nicolas.cellier.aka.nice@gmail.com) at November 17, 2019 11:47 AM

Fix union: and intersection: of CharacterSet

We must avoid using naive loops for huge sets that canBeEnumerated not.

An enumerationCost scale is introduced for that purpose.

The idea is that intersection: must be performed by the cheapest set, while union: must be performed by the most expensive. The most expensive eventually knows how to handle loops lazily.

by Nicolas Cellier (nicolas.cellier.aka.nice@gmail.com) at November 17, 2019 11:46 AM

Let ImageSnapshot>>reboot set up the display window in prior to resuming the VM in the new snapshot image. The fullscreen flag is handled properly, but window size does not currently work on X11 because ioSetDisplayMode is not implemented in the X11 driver (window resize requests currently can originate only from the external X11 server).

by Dave Lewis (lewis@mail.msen.com) at November 16, 2019 02:46 PM

more WasmForensics refactoring

November 16, 2019 06:14 AM

Oops.

by Eliot Miranda (eliot.miranda@gmail.com) at November 16, 2019 05:52 AM

Name: System-mt.1119
Author: mt
Time: 15 November 2019, 2:31:30.94533 pm
UUID: bd66f2d9-be6a-9744-840d-9654e2d5565d
Ancestors: System-dtl.1118, System-ct.1097

Merges System-ct.1097, which adds means to open tools for class (or system categories), method versions, and packages.

by Dave Lewis (lewis@mail.msen.com) at November 16, 2019 01:17 AM

And, goddamnit, have a Newspeak VM reveal the same.

by Eliot Miranda (eliot.miranda@gmail.com) at November 16, 2019 12:59 AM

ImageChecker: ImageChecker must not specify excluding unmarked objects, since objects are unmarked on load. Have its main routine respond to a -version argument.

Slang: Call an Open Smalltalk VM an Open Smalltalk VM.

by Eliot Miranda (eliot.miranda@gmail.com) at November 16, 2019 12:58 AM

Name: System-dtl.1118
Author: dtl
Time: 14 November 2019, 11:44:09.783497 am
UUID: e1e88bc2-8860-4665-9ac4-af0e01dc258c
Ancestors: System-mt.1117

When reporting VM parameters, enable formatting in the case of a parameter value that is represented as a Float rather than an Integer. This may happen if a VM parameter is defined in units of milliseconds but internally measured to microsecond precision by the VM, in which case the value may be returned as a Float in order to retain full precision. SmalltalkImage>>formatVMParameter: handles the formatting for integer and non-integer values.

Includes recommendations from Marcel and Chris, original System-dtl.1118 submission is moved to treated inbox.

by Dave Lewis (lewis@mail.msen.com) at November 16, 2019 12:56 AM

Slang: avoid generating gcc3x-validImage.c. Including sqGnu.h only benefits something that has an interpret method.

by Eliot Miranda (eliot.miranda@gmail.com) at November 15, 2019 11:45 PM

Add an ImageLeakChecker that is only able to load an image (currently only Spur formats have been generated and tested). Do this with horrible Slang hackery to discard lots of stuff. Generate validImage.c in a stack vm directory. (N.B. this hasn't been run in the simulator yet; it'll take a little bit of work to get this to work, but in teh simulator it's easy to use the inbuilt facilities in a VM via setCheckForLeaks:)

As part of this effort clean up several methods scattered around the system so that they either are marked with <api> when they're exported, or they are always inlined, or use symbols for returnTypeC:, or have the correct return type, or no longer the sole method in a category.

Slang:
Get VMMaker to respect doInlining if imolemented by an interpreter. It used to only pay attention to a Cogit.

Enlarge and rationalize the kinds of "quick" mthods that will be inlined when doInlinign answers asSpecifiedAndQuick. Include any and all that just answer self or a constant (i.e. include those that take arguments). Include any and all that perform only a signle assignment, either answering the value or answering self.

Fix inlining of literal blocks vs inlining of two element statement lists created by the inliner. Do so by introducing a subclass of TStmtListNode called TLiteralBlockNode which has one method, isLiteralBlock and is used by BlockNode>>asTranslatorNodeIn:. Use asRootTranslatorNodeIn: to create top-level method statement lists. Better comment collectInlineList:.

Allow comma on strings to translate into ANSI C automatic literal string concatenation.

Next step is to update the stack makefiles to build validImage executables.

by Eliot Miranda (eliot.miranda@gmail.com) at November 15, 2019 11:37 PM

Fix a slip in x86-64 disassembly decoration. Misplaced closing bracket.

by Eliot Miranda (eliot.miranda@gmail.com) at November 15, 2019 11:08 PM

Refresh Version Inspector after reparent was pressed.

by David T. Lewis (lewis@mail.msen.com) at November 15, 2019 08:18 PM

Add "debug accept action" into debug menu of PluggableTextMorph. Furthermore, refactor all "debug action" code in this package by using ToolSet>>#debugBlock:runUntilMethod:withTitle:.

Complements System-ct.1121.

November 15, 2019 06:55 PM

Add convenience method for debugging a block at a method.

November 15, 2019 06:47 PM

Update SqueakProjectHelp squeakResourcesOnline

* Remove links that are offline (the merger might check whether they are really permanently offline)
* Reformat the page to a modern look

Depends on Collections-ct.861.

November 15, 2019 06:34 PM

Migrates VersionsBrowser>>versionsHelpString into a HelpTopic class

Depends on Collections-ct.861.

November 15, 2019 06:19 PM

Adds fileIn/Out support for TextIndent.

November 15, 2019 06:18 PM

VersionsBrowser>>#selectMethod:: Compare not only timeStamp, but also source, as timeStamp is only minute-precise.

November 15, 2019 06:00 PM

Refactor ProcessBrowser nil checks again and move them into #stackListKey:from:. Also, use #browseMethod: in #browseContext to ensure the right version is browsed

Thanks to Marcel for the review!

November 15, 2019 04:30 PM

Fixes tree-widget tests. They got broken due to the latest updates in tree widgets, which now automatically select the selection path in the beginning without the model having to say so. That's comparable to fetching the initial list/roots etc.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at November 15, 2019 03:34 PM

This update may be slow due to the package postscript processing.

Apply Kernel-mt.1278, which mergesKernel-mt.1266.

From Kernel-nice.1270, apply the BoxedFloat64>>truncated change to Float for V3.

Also update Behavior>>becomeCompact to ensure that a newly added compact class does not occupy a previously used slot in the compact classes table. This is a requirement for segment loading, as documented in the becomeCompact method comment.

Reference Levente's suggestion that ByteArray adoptInstance: 'aSting' could work in a V3 image if class ByteArray is added to the compact classes array:
http://lists.squeakfoundation.org/pipermail/squeak-dev/2019-October/204924.html

In the package postscript, make class ByteArray compact. This will allow the trunk version of StandardFileStream>>binary to work on V3, which in turn will eliminate the need for a separate V3 version of the Files package in the V3 update stream.

Name: Kernel-mt.1278
Author: mt
Time: 13 November 2019, 10:32:26.550043 am
UUID: 1aae36c0-4723-934a-a101-43de1d7435aa
Ancestors: Kernel-nice.1277, Kernel-mt.1266

Merges Kernel-mt.1266, which makes #messagesSequence compatible with SistaV1.

Name: Kernel-nice.1270
Author: nice
Time: 1 October 2019, 9:22:39.460936 pm
UUID: 05ae7bbb-2517-4c6e-a9c0-db89b42742e5
Ancestors: Kernel-mt.1269

In SmallFloat64>>#truncated fallback, don't bother with Infinities and NaN, since all SmallFloat64 are finite, nor with Float fractionPart, since all non-integer Float value fits in SmallInteger in a 64bits image, and thus are already handled by the primitive.

It's legitimate to inject such implementation specific knowledge into this implementation specific class.

Also revise BoxedFloat64>>#truncated, mostly with comments. Dividing by SmallInteger maxVal + 1 / 2 would lead to an infinite loop in a 64bits image, fortunately this branch is unreachable then. Each time I read it, I have the shadow of a doubt ;)

by Dave Lewis (lewis@mail.msen.com) at November 14, 2019 01:07 AM

Password was very specific for ServerDirectory.
Re-factor Password into a more generic password holder (usuable in other classes).
Create ServerPassword as a sub-class of Password with necessary specilizations moved down to it.
Make ServcieDirectory point to ServerPassword.

This allows us to use Password generally to store a password in memory (encoded) during a Squeak session, which is cleared out when saving. The intended use case is to not store passwords in code or in the file system - instead prompt the user at first use and cache it for the rest of the life of the session.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at November 13, 2019 12:58 PM

Move noAuthorizationMethod to the class side.

Without this change the #socks5 method is broken.
All the other default settings that #socks5 uses are on the class side too and its the only place that uses it.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at November 13, 2019 12:58 PM

Merges inbox contributions.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at November 13, 2019 12:57 PM

Increase the accuracy of magnifier a little bit. Not sure whether we want to add some "low performance" check here. In that case, I would opt for "500" or "1000" as stepTime.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at November 13, 2019 12:49 PM

Simplify help display of FreeCell (Etoys has also depended on ToolBuilder before)

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at November 13, 2019 11:19 AM

Add Form>>#scaledToWidth:, an analogon for #scaledToHeight:

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at November 13, 2019 11:18 AM

FreeCell>>#help: Use #editWithLabel: instead of reinventing the wheel ...

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at November 13, 2019 11:18 AM

Respect submorphs in MovingEyeMorph.>>#color:

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at November 13, 2019 11:18 AM

Replace miscellaneous deprecated calls

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at November 13, 2019 11:18 AM

Refactor SpectrumAnalyzerMorph: Extract #fftSize: to enable foreign calls

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at November 13, 2019 11:18 AM

Recategorize WatchMorph constructor

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at November 13, 2019 11:17 AM

Fixes a drawing bug in SyntaxMorph

Before this commit, you could not even run [SyntaxMorph test] due to this slip ...

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at November 13, 2019 11:17 AM

Fixes small layout error in tiles due to missspelling

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at November 13, 2019 11:17 AM

Allow for embedding SyntaxMorphs into test tiles.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at November 13, 2019 11:15 AM

Merge! Merge! Merge! Various fixes in Etoys-related places.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at November 13, 2019 11:15 AM

Two minor layout fixes.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at November 13, 2019 11:09 AM

Complements Morphic-mt.1586.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at November 13, 2019 10:53 AM

Fixes convenience messages to put a morph in a scroll pane. No need for the special case of a two-way scroll pane anymore.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at November 13, 2019 10:52 AM

Make #messagesSequence compatible with SistaV1 and CompiledBlcoks again *and* keep the meaning of #messagesSequence as restored in the previous commit by eem.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at November 13, 2019 09:32 AM

Merges Kernel-mt.1266, which makes #messagesSequence compatible with SistaV1.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at November 13, 2019 09:32 AM

empty log message

by Nicolas Monasterio (nicoinfo.edu@gmail.com) at November 13, 2019 02:19 AM

empty log message

by Nicolas Monasterio (nicoinfo.edu@gmail.com) at November 13, 2019 01:55 AM

empty log message

by Nicolas Monasterio (nicoinfo.edu@gmail.com) at November 13, 2019 01:43 AM

empty log message

by Nicolas Monasterio (nicoinfo.edu@gmail.com) at November 13, 2019 01:36 AM

empty log message

by Graham (gkgoat6700@gmail.com) at November 12, 2019 07:45 PM

empty log message

by Graham (gkgoat6700@gmail.com) at November 11, 2019 09:45 PM

empty log message

by Graham (gkgoat6700@gmail.com) at November 11, 2019 09:44 PM

empty log message

by Graham (gkgoat6700@gmail.com) at November 11, 2019 09:27 PM

empty log message

by Graham (gkgoat6700@gmail.com) at November 11, 2019 09:26 PM

empty log message

by Graham (gkgoat6700@gmail.com) at November 11, 2019 09:01 PM

empty log message

by Graham (gkgoat6700@gmail.com) at November 11, 2019 08:36 PM

empty log message

by Graham (gkgoat6700@gmail.com) at November 11, 2019 08:17 PM

empty log message

by Graham (gkgoat6700@gmail.com) at November 11, 2019 07:34 PM

empty log message

by Graham (gkgoat6700@gmail.com) at November 11, 2019 07:29 PM

empty log message

by Graham (gkgoat6700@gmail.com) at November 11, 2019 07:24 PM

empty log message

by Graham (gkgoat6700@gmail.com) at November 11, 2019 07:20 PM

empty log message

by Graham (gkgoat6700@gmail.com) at November 11, 2019 02:37 PM

empty log message

by Graham (gkgoat6700@gmail.com) at November 11, 2019 02:37 PM

empty log message

by Graham (gkgoat6700@gmail.com) at November 11, 2019 02:36 PM

empty log message

by Thiago da Silva Lino (thiagolino@yahoo.com.br) at November 07, 2019 09:02 AM

Quick fix for versions browser.

Note that I think that the interaction that results in selection changes should be implemented in the widgets, not the models.

Is [shift] pressed?
Is it a drag?
Is it a click on an already selected item?

All these things belong into widget code to keep model code simple. For example, I think that we should also move the logic behind range selection from MessageTrace into the AlternatePluggableListMorphOfMany. So other models (or tools) could benefit from it.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at November 07, 2019 08:21 AM

empty log message

November 06, 2019 10:51 PM

Adds missing method to make #test01ImplementationHooks pass again.

We might want to re-design this at some point. There are now several cases in the image where class-side caches need to be invalidated on UI-theme application. Maybe change the test...

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at November 06, 2019 05:52 PM

empty log message

November 06, 2019 12:08 PM

Fixes a regression in PluggableListMorph >> #getVisibleList. Also restore code formatting in that method.

Note that "3@3" referred to #hMargin in LazyListMorph, which is not required since the entire row is defined by its selection rectangle, not the beginning of the text. Still, "0@1" is required because #bottomLeft is (always) one pixel to far down.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at November 06, 2019 08:07 AM

Let OrderedDictionary print its elements in its own sequence.

by Chris Muller (ma.chris.m@gmail.com) at November 05, 2019 11:07 PM

Move noAuthorizationMethod to the class side.

Without this change the #socks5 method is broken.
All the other default settings that #socks5 uses are on the class side too and its the only place that uses it.

November 05, 2019 01:28 PM

Let the printString of an OrderedDictionary reveal its key sequence.

by Chris Muller (ma.chris.m@gmail.com) at November 05, 2019 01:25 AM

Use pushRemappableOop: to protect all positive64BitIntegerFor: calls in primitiveVMParameter. A large integer is instantiated in the VM even if normalized to small integer in the image, so protection is needed.

by David T. Lewis (lewis@mail.msen.com) at November 05, 2019 12:46 AM

VMMaker 4.16.8
Correct a long-standing issue in VM parameter reporting for values specified as milliseconds in the interface, but measured as microseconds internally. Adhere to the documented interface and answer the values in units of milliseconds, using Float values to supply higher precision when available from the VM. For example, a duration measured as 123456 microseconds is reported as a VM parameter value of 123.456 milliseconds. The change affects VM parameters 8, 10, 36, and 38.

by David T. Lewis (lewis@mail.msen.com) at November 05, 2019 12:45 AM

Proposal to add GC pressure to #bench method.

October 30, 2019 09:25 AM

*at:put: should always return the stored value.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at October 30, 2019 08:16 AM

Make it possible to recover from (deprecation) warnings in important drawing methods.

See http://forum.world.st/The-Trunk-System-mt-1093-mcz-tp5103937p5106420.html

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at October 29, 2019 02:09 PM

Fixes (a regression in) list selection in change-list tool. It is not the job of the model to implement toggle in a multi-selection setting. The model only has to react to index selector and selection-list selectors.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at October 29, 2019 11:39 AM

Fix issue 436 again

if msecs is usqInt, then the subtraction msecs - ioMsecs() is performed as unsigned 32bits, and can lead to high value instead of slightly negative value, which remain positive when converted to 64 bit signed long deltaMsecs...

That makes a near 48 days delay before waking up!

The expression must be evaluated with 64bits signed long arithmetic, what we obtain by declaring msecs as sqLong...

When we write C code, we must think in C language, not upper level language with exact integer arithmetic (no wrap, no overflow), that's a potential danger of Slang...

by Nicolas Cellier (nicolas.cellier.aka.nice@gmail.com) at October 28, 2019 06:14 PM

Provide error handling for #adoptInstance in StandardFileStream>>binary to provide pre-Spur behavior when running on a V3 image. Prior to Spur, class ByteArray cannot adopt an instance of ByteString, so use #asByteArray instead.

Resolves the only remaining difference for package Files when running on either Spur or V3 images. The packages with significant Spur/V3 differences (excluding unit tests) are now Collections, Compiler, Kernel, and System. See www.squeaksource.com/TrunkUpdateStreamV3 for verification.

by David T. Lewis (lewis@mail.msen.com) at October 28, 2019 12:50 PM

Fix the problem with small delays in old images, which occasionally break OpenSmalltalk CI as reported in issue 436

https://github.com/OpenSmalltalk/opensmalltalk-vm/issues/436

The problem has been observed on a macos32x86/squeak.cog.v3 VM + Squeak-4.6 image still using millisecond clock to handle delays (and thus #primitiveSignalAtMilliseconds).
It happens with small delays that may be seen as slightly negative inside the primitive, and erroneously confused with a clock roll-over.

This should not happen in newer image that use microsecond clock, though I didn't perform any related analysis.
The problem can easily be triggered by this snippet (save you work first!):

s := Semaphore new.
Delay primSignal: s atMilliseconds: Time primMillisecondClock - 2.
s wait.

Now, the detailed explanations:

---------------------------------------------------

At image side, the absolute time msecs at which to wake up is usually formed by adding:

msecs := absolute_millisecond_clock1 + relative_delay_duration.

In #primitiveSignalAtMilliseconds, some time has elapsed since, and we subtract millisecond clock to retrieve the remaining duration to wait:

deltaMsecs := msecs - absolute_millisecond_clock2.

If ever the millisecond clock roll-over in between the two calls clock1 and clock2,
then we will have clock1 close to clock2 + MillisecondClockMask, and:

deltaMsecs := relative_delay_duration - already_elapsed_time + MillisecondClockMask

that is a huge delay, 1<<29 milliseconds, about 6 days!

A negative deltaMsecs means an already expired delay, it cannot happen because of rollover
(unless the VM hangs more than 3 days, case which we do not have to handle: the delay is then expired because we cannot handle such huge delays - see below).

Technically, the usage of #positive32BitValueOf: would allow passing msecs upTo: 1<<32-1, that is a huge delay (at least 1<<32 - (1<<29), about 48 days).

But how to distinguish a huge relative_delay_duration from a rollover?
We can't really. We might check if clock2 is small, and declare it's a rollover if under a threshold, else a big delay.
But there will always be edge cases when such oracle would give the wrong answer.
We could try and pass 3 arguments instead of 2, semaphore, clock1 , relativeDelay.
But are we going to need it? Not really.
With a clock that roll-over like this, the only sane thing is to cap the delay at midway:

limit := MillisecondClockMask >> 1.

Then reduce a positive deltaMsecs with centered modulo:

deltaMsecs > limit ifTrue: [deltaMsecs := deltaMsecs - MillisecondClockMask].

Only a strictly positive deltaMsecs need then to be honored as the next wake up time (nextWakeupUsecs).

---------------------------------------------------

But wait, this is only one face of the problem!

Due to current implementation of ioRelinquishProcessorForMicroseconds(),
the microsecond increment (deltaMsecs * 1000) must fit in a long.
If ever it overflows, and wrap over negated values, then bad consequences will happen via aioSleepForUsecs() and aioPoll().
This is including unresponsive runaway images spitting out errno 22; select: invalid argument in the console, as observed in bug report.
On 32bits unix VM, that further limit the maximum delay to about 2147 seconds!

IMO, it's urgent to protect ioRelinquishProcessorForMicroseconds() from such overflow,
But that will deserve another commit (it's platforms source code, not Smalltalk code).

by Nicolas Cellier (nicolas.cellier.aka.nice@gmail.com) at October 25, 2019 10:12 PM

Refactor Utilities' CommonRequestStrings: Instead of an obsolescent string, store the code itself.

This allows to benefit from senders and other refactoring tools. Also, it's impossible to compile invalid code. As you can see, I commented one line out as it contains a selector that is no longer present in modern Squeak.

This change should not affect overall performance, as #commonRequestStrings keeps caching the result.

Please note that when merging this commit into Trunk, a conflict with System-ct.1103 will be to solved manually.

October 25, 2019 04:50 PM

Proposal: Support for a permanent, image-defined host window title

Do "DisplayScreen defaultHostWindowTitle: 'Image sweet Image'", save your image and reopen it

October 25, 2019 03:53 PM

Removes Blocks check from MethodFinder

Due to several implementators of #value(:)?(value:)* and #cull(:)?(cull:)* we have got nowadays, this is an unnecessary restriction. Furthermore, it impedes maintenance. For more information, read http://forum.world.st/MethodFinder-Blocks-td5105421.html

October 25, 2019 11:00 AM

Provide error handling for #adoptInstance in StandardFileStream>>binary to provide pre-Spur behavior when running on a V3 image. Prior to Spur, class ByteArray cannot adopt an instance of ByteString, so use #asByteArray instead.

Resolves the only remaining difference for package Files when running on either Spur or V3 images. The packages with significant Spur/V3 differences (excluding unit tests) are now Collections, Compiler, Kernel, and System. See www.squeaksource.com/TrunkUpdateStreamV3 for verification.

by David T. Lewis (lewis@mail.msen.com) at October 25, 2019 02:15 AM

Adds things for column-specific list filtering:

- backgroundColor for LazyListMorph
- optional filter-term indication for LazyListMorph
- column highlights for multi-column lists
- Character tab to cycle between columns

Note that I opted to not use the [space] key for column toggling because (a) we have to widget-focus cycling via [tab] at the moment and (b) the space character might be a valuable filter term.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at October 23, 2019 03:31 PM

Complements ToolBuilder-Kernel-mt.134.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at October 23, 2019 01:41 PM

Complements ToolBuilder-Kernel-mt.134.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at October 23, 2019 01:41 PM

Support the functional layout properties for LazyListMorph in tool-builder specs. Chose "alignment" instead of "positioning" because specs are not Morphic-specific. The terms "item" and "column" reflect item- or column-specifc settings, which can be confusing because our current lists do only support uniform (item) rows.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at October 23, 2019 01:40 PM

tryng to solve loop and block

October 23, 2019 12:43 PM

Complements Morphic-mt.1580.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at October 23, 2019 10:50 AM

Deprecates MulticolumnLazyListMorph because PluggableMultiColumnListMorph uses the regular LazyListMorph by now.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at October 23, 2019 10:50 AM

empty log message

by Graham (gkgoat6700@gmail.com) at October 21, 2019 09:39 AM

empty log message

by Graham (gkgoat6700@gmail.com) at October 20, 2019 09:04 PM

empty log message

by Graham (gkgoat6700@gmail.com) at October 20, 2019 09:04 PM

empty log message

by Graham (gkgoat6700@gmail.com) at October 20, 2019 08:47 PM

empty log message

by Graham (gkgoat6700@gmail.com) at October 20, 2019 08:47 PM

empty log message

by Graham (gkgoat6700@gmail.com) at October 20, 2019 07:35 PM

empty log message

by Graham (gkgoat6700@gmail.com) at October 20, 2019 07:17 PM

empty log message

by Graham (gkgoat6700@gmail.com) at October 20, 2019 02:43 PM

empty log message

by Graham (gkgoat6700@gmail.com) at October 20, 2019 02:05 PM

empty log message

by Graham (gkgoat6700@gmail.com) at October 20, 2019 02:03 PM

empty log message

by Graham (gkgoat6700@gmail.com) at October 20, 2019 02:03 PM

empty log message

by Graham (gkgoat6700@gmail.com) at October 20, 2019 01:40 PM

empty log message

by Graham (gkgoat6700@gmail.com) at October 20, 2019 01:32 PM

When rebooting the VM into a new ImageSnapshot, honor the resuming flag to prevent re-evaluating in the resumed image.

by Dave Lewis (lewis@mail.msen.com) at October 20, 2019 12:45 PM

empty log message

by Graham (gkgoat6700@gmail.com) at October 20, 2019 12:17 PM

empty log message

by Graham (gkgoat6700@gmail.com) at October 20, 2019 11:42 AM

empty log message

by Graham (gkgoat6700@gmail.com) at October 20, 2019 11:41 AM

empty log message

by Graham (gkgoat6700@gmail.com) at October 19, 2019 09:26 PM

empty log message

by Graham (gkgoat6700@gmail.com) at October 19, 2019 09:26 PM

empty log message

by Graham (gkgoat6700@gmail.com) at October 19, 2019 09:26 PM

Add missing methods so that SlangBrowser works again on classes and methods that are not part of the VM. Methods that can be translated to C (or inlined C) are displayed, others result in assertion errors or invalid C code.

For example, Scanner>>scanFieldNames: and many other methods in class Scanner are translated to C or inlined C, but Scanner>>scanAllTokenPositionsInto: cannot be translated and fails with an assertion error.

by David T. Lewis (lewis@mail.msen.com) at October 19, 2019 12:15 AM

Utilities>>setterSelectorfor: was deprecated

by Karl Ramberg (karlramberg@gmail.com) at October 18, 2019 07:59 PM

Use Processor userSchedulingPriority rather than priority 40.

by John-Reed Maffeo (jrmaffeo@gmail.com) at October 17, 2019 02:37 PM

Change JsonTests>>testStreaming to compare to JsonObject new, rather than just testing isDictionary.

testStreaming should not only check that the result is a dictionary, but that it is empty.

by John-Reed Maffeo (jrmaffeo@gmail.com) at October 17, 2019 02:37 PM

Saved from Windows 7 computer. Changes made to support cross platform for Win and Mac. Added Json and OSProcess as required packages.

by John-Reed Maffeo (jrmaffeo@gmail.com) at October 17, 2019 02:17 PM

Using AlignmentMorph as container

by Thiago da Silva Lino (thiagolino@yahoo.com.br) at October 16, 2019 03:12 PM

First commit of new Quinto game!

by Patrick R. (patrick.rein@gmail.com) at October 16, 2019 12:46 PM

Make the lazy list morph also lazy in terms of the icon extent. Only probe the extent of the first icon and then use that for the entire list. This matches the use of a single font to determine the height of the entire list. See #maxHeight.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at October 16, 2019 11:21 AM

Minor fix because #select: keeps the original kind, which is a set, which cannot be sorted in-place.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at October 16, 2019 11:09 AM

Reverts fix for test TestAnchorTest >> #testResizeAnchoredMorph because it breaks at least two important things. Also document those things.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at October 16, 2019 11:08 AM

Name: System-mt.1115
Author: mt
Time: 11 October 2019, 6:31:06.485593 pm
UUID: 414c0966-8964-aa49-9eb0-a44dd2542d3e
Ancestors: System-mt.1114

Simplify feature-based class search. For example, given the term "LazyMor", find both LazyListMorph and MulticolumnLazyListMorph.

by Dave Lewis (lewis@mail.msen.com) at October 15, 2019 11:44 PM

empty log message

by Graham (gkgoat6700@gmail.com) at October 15, 2019 09:14 PM

empty log message

by Graham (gkgoat6700@gmail.com) at October 15, 2019 09:12 PM

empty log message

by Graham (gkgoat6700@gmail.com) at October 15, 2019 09:12 PM

empty log message

by Graham (gkgoat6700@gmail.com) at October 15, 2019 09:11 PM

empty log message

by Graham (gkgoat6700@gmail.com) at October 15, 2019 09:11 PM

Tests conversion between ParseNodes and tiles

October 15, 2019 08:00 PM

Refines conversion of MessageNodes into test tiles: Don't fill empty cases with "nil"

October 15, 2019 07:56 PM

empty log message

by Graham (gkgoat6700@gmail.com) at October 15, 2019 07:17 PM

init

by Graham (gkgoat6700@gmail.com) at October 15, 2019 06:41 PM

empty log message

by Graham (gkgoat6700@gmail.com) at October 15, 2019 06:40 PM

empty log message

by Graham (gkgoat6700@gmail.com) at October 15, 2019 06:23 PM

empty log message

by Graham (gkgoat6700@gmail.com) at October 15, 2019 06:22 PM

empty log message

by Graham (gkgoat6700@gmail.com) at October 15, 2019 06:22 PM

Changes "revert to tile version" behavior in script editor: Instead of discarding the changed code, always use the new approach and convert the decompiled code back to tiles.

October 15, 2019 12:58 PM

Allow for embedding SyntaxMorphs into test tiles.

October 15, 2019 12:46 PM

Adds basic support for reconverting ParseNodes into EToys tiles

Try out:

p := World assuredPlayer.
e := (Player >> #liftAllPens) decompile asScriptEditorFor: p.
e openInHand.

October 15, 2019 12:42 PM

Fixes small layout error in tiles due to missspelling

October 15, 2019 12:17 PM

Fixes a drawing bug in SyntaxMorph

Before this commit, you could not even run [SyntaxMorph test] due to this slip ...

October 15, 2019 12:12 PM

Better error message in pluggable lists for misbehaving models. :-)

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at October 15, 2019 11:58 AM

Models must never provide "nil" as the list to the view.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at October 15, 2019 11:57 AM

Fixes that annoying huge stack of progress bars that pops up when loading/refreshing code in Monticello.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at October 15, 2019 11:40 AM

More robustness: (1) Catch both Error and Halt in Morphic drawing and (2) Reset drawing errors in morphs when restoring the Display.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at October 15, 2019 11:30 AM

Fixes a bug with list-selection restoring after changing UI themes, for example. Tweaks appearance of list filters for demo/hi-dpi mode. Removes the use of #bigDisplay property. RealEstateManage handles window resizing just fine.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at October 15, 2019 08:35 AM

Caches ScaleFactor. Removes the use of #bigDisplay property because it does not work at the moment. Everything relates to the default font size. Further #bigDisplay scaling makes things look awkward.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at October 15, 2019 08:32 AM

Adds a missing accessor for multi-column list configuration.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at October 15, 2019 07:42 AM

Fixes two minor issues after the recent LazyListMorph updates which affect the #cellInset property.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at October 15, 2019 07:32 AM

Clean up app registry after debugger tests.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at October 15, 2019 07:11 AM

Fixes Morphic dependency for text anchors.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at October 15, 2019 07:07 AM

empty log message

by Graham (gkgoat6700@gmail.com) at October 14, 2019 08:54 PM

empty log message

by Graham (gkgoat6700@gmail.com) at October 14, 2019 08:43 PM

empty log message

by Graham (gkgoat6700@gmail.com) at October 14, 2019 08:42 PM

Name: System-mt.1114
Author: mt
Time: 9 October 2019, 3:31:40.772803 pm
UUID: 8d7cf94a-6d5a-6d4c-aef8-e48a1bd36ea8
Ancestors: System-eem.1113

For exception handling and debugging, untangle some dependencies between System and Kernel.

by Dave Lewis (lewis@mail.msen.com) at October 10, 2019 11:20 AM

Name: Kernel-mt.1275
Author: mt
Time: 10 October 2019, 11:13:31.338276 am
UUID: 8558b4a5-037d-4e61-a937-0f552f86fbfb
Ancestors: Kernel-mt.1274

Shout styling should not depend (implicitely) on Morphic.

by Dave Lewis (lewis@mail.msen.com) at October 10, 2019 11:16 AM

Name: Kernel-mt.1274
Author: mt
Time: 9 October 2019, 3:31:08.214803 pm
UUID: 72b966cb-5325-794c-b6a3-068303a6b153
Ancestors: Kernel-eem.1273

For exception handling and debugging, untangle some dependencies between System and Kernel.

by Dave Lewis (lewis@mail.msen.com) at October 10, 2019 11:16 AM

Name: Compiler-mt.412
Author: mt
Time: 9 October 2019, 9:47:40.360433 am
UUID: e033a572-485d-41e4-9fcc-5cc55d49a893
Ancestors: Compiler-eem.411

Reformat the sources from my recent changes in the Compiler package in the style of rectangular blocks because it is the default style in that package.

Thanks to Eliot for the pointer.

by Dave Lewis (lewis@mail.msen.com) at October 10, 2019 11:13 AM

Shout styling should not depend (implicitely) on Morphic.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at October 10, 2019 09:15 AM

Shout styling should not depend (implicitely) on Morphic.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at October 10, 2019 09:14 AM

Shout styling should not depend (implicitely) on Morphic.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at October 10, 2019 09:14 AM

Shout styling should not depend (implicitely) on Morphic.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at October 10, 2019 09:13 AM

Adds a test for that (fixed) resize-anchor-should-recompose bug.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at October 10, 2019 09:01 AM

Adds "Treated Inbox" to the list of well-known repositories.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at October 10, 2019 08:00 AM

Cogit:
Fix MoveAwR and MoveRAw for FPReg and SPReg on X64 by avoiding the swap of FPReg/SPReg with RAX prior to the move; i.e. RAX will get smashed. Document this in three relevant places. Note that this situation arises because CFramePointer and CStackPointer are global variables, not private variables in the interpreter's collection, and hence not accessible via VarBaseReg(RBX). Sere we to arrange CFramePointer and CStackPointer were accessible via VarBaseReg(RBX) trampolines would be shorter and perhaps faster. We should do this asap.

In the simulator change the regime for CFramePointer & CStackPointer from interpreter local to in-memory to get the simulator to generate code using the same scheme.

Have manageFrom:to: inlined.

Fix some minor bugs with assembly decoration.

Simulator: don;'t wipe out system attribute 2 in openOn:extraMemory: allowing setting of arguments before openOn:extraMemory:.

Plugins: fix a typo.

by Eliot Miranda (eliot.miranda@gmail.com) at October 10, 2019 03:08 AM

Kent Beckian purposefully plenk'd pretty printing.

by Chris Muller (ma.chris.m@gmail.com) at October 09, 2019 10:00 PM

After reparenting a WorkingCopy, note it dirty.

by Chris Muller (ma.chris.m@gmail.com) at October 09, 2019 09:51 PM

Relieve the user from having to move or delete the unneeded progress bar when debugging individual tests from the Test Runner.

by Chris Muller (ma.chris.m@gmail.com) at October 09, 2019 09:50 PM

This version is the current working copy of the OTAR application. I set up project entries in Squeak Map and Squeak Source. In the process, I deleted a project with "OTAR" in the name which I had no use for.

by John-Reed Maffeo (jrmaffeo@gmail.com) at October 09, 2019 07:32 PM

empty log message

by Graham (gkgoat6700@gmail.com) at October 09, 2019 06:36 PM

empty log message

by Graham (gkgoat6700@gmail.com) at October 09, 2019 06:17 PM

Fixes text-composition bug for morphs embedded in text morphs through the text-anchor property.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at October 09, 2019 03:55 PM

I forgot a flag in Morphic-mt.1563.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at October 09, 2019 03:45 PM

Complements Morphic-mt.1563.

Adds at least a comment to ImmAbstractPlatform to get started.

by Marcel Taeumel (marcel.taeumel@hpi.uni-potsdam.de) at October 09, 2019 03:41 PM

Name: System-eem.1113
Author: eem
Time: 8 October 2019, 1:09:10.347077 pm
UUID: 8706b6c9-6d43-4c5d-97c2-aa69d0ceba2d
Ancestors: System-mt.1112

Implement WeakRegistry class>>postRecompileAction to update WeakRegistry finalizers post recompile to avoid references to stale methods.

by Dave Lewis (lewis@mail.msen.com) at October 09, 2019 01:06 AM

Name: Kernel-eem.1273
Author: eem
Time: 8 October 2019, 12:59:35.689325 pm
UUID: 50bfe346-15d7-4aa9-ba69-0c21d3d8d011
Ancestors: Kernel-mt.1272

Add a method that documents the postRecompileAction convention for rebuilding state affected by a Compiler recompileAll.

by Dave Lewis (lewis@mail.msen.com) at October 09, 2019 01:04 AM

Name: Compiler-eem.411
Author: eem
Time: 8 October 2019, 1:01:29.356552 pm
UUID: fb8ba22e-7115-49b7-8a23-52ae58c168cf
Ancestors: Compiler-mt.410

Implement the postRecompileAction convention for rebuilding state affected by a Compiler recompileAll.
Eliminate such a dependency in StdLiterals itsdelf by using a Symbol for performing literalEqual:. Remove Decompiler initialize from the comment; this method doesn;t do anything affected by VariableNode class>>#initialize anymore.

by Dave Lewis (lewis@mail.msen.com) at October 09, 2019 01:02 AM

Name: System-mt.1112
Author: mt
Time: 7 October 2019, 10:07:46.787231 am
UUID: 3dbbf205-1ee0-3841-ae6b-2d926a4a2055
Ancestors: System-pre.1111

Fixes and clean-up in the debugger infrastructure:

- Adds several tests in DebuggerTests
- Replaces ErrorRecursion and ErrorRecursionGard with a process-local flag to avoid false recursion detection
- Simplifies interrupt-request processing in Project >> #interruptName:preemptedProcess:
- Simplifies "Debug it" invocation. See TextEditor >> #debugIt and Process' class-side messages.
- Move log-file writing for low space to SmalltalkImage >> #lowSpaceWatcher.
- Adds Debugger >> #close, similar to Project >> #close.
- Moves #cmdDotEnabled preference check to StandardToolSet
- Adds a convenient way to debug exceptions without suspending the current process.

by Dave Lewis (lewis@mail.msen.com) at October 09, 2019 12:51 AM

Name: Kernel-mt.1272
Author: mt
Time: 7 October 2019, 10:09:56.205231 am
UUID: 8fe3f896-0806-fe46-8e86-68f3c0f98d4d
Ancestors: Kernel-ul.1271

Complements System-mt.1112, which fixes and clean-up in the debugger infrastructure.

by Dave Lewis (lewis@mail.msen.com) at October 09, 2019 12:46 AM

empty log message

by Graham (gkgoat6700@gmail.com) at October 08, 2019 08:30 PM

empty log message

by Graham (gkgoat6700@gmail.com) at October 08, 2019 08:26 PM

empty log message

by Graham (gkgoat6700@gmail.com) at October 08, 2019 08:13 PM

empty log message

by Graham (gkgoat6700@gmail.com) at October 06, 2019 08:12 PM

empty log message

by Graham (gkgoat6700@gmail.com) at October 06, 2019 06:19 PM

empty log message

by Graham (gkgoat6700@gmail.com) at October 06, 2019 05:36 PM

empty log message

by Graham (gkgoat6700@gmail.com) at October 06, 2019 03:06 PM

empty log message

by Graham (gkgoat6700@gmail.com) at October 06, 2019 03:05 PM

empty log message

by Graham (gkgoat6700@gmail.com) at October 05, 2019 08:16 PM

empty log message

by Graham (gkgoat6700@gmail.com) at October 05, 2019 11:28 AM

empty log message

by Graham (gkgoat6700@gmail.com) at October 05, 2019 11:23 AM

class containers

by Graham (gkgoat6700@gmail.com) at October 05, 2019 10:35 AM

empty log message

by Graham (gkgoat6700@gmail.com) at October 04, 2019 06:59 PM

empty log message

by Graham (gkgoat6700@gmail.com) at October 03, 2019 09:47 AM

empty log message

by Graham (gkgoat6700@gmail.com) at October 03, 2019 09:46 AM

empty log message

by Graham (gkgoat6700@gmail.com) at October 03, 2019 09:44 AM

empty log message

by Graham (gkgoat6700@gmail.com) at October 02, 2019 08:21 PM

empty log message

by Graham (gkgoat6700@gmail.com) at October 02, 2019 06:15 PM

empty log message

by Graham (gkgoat6700@gmail.com) at October 02, 2019 06:04 PM

empty log message

by Graham (gkgoat6700@gmail.com) at October 02, 2019 05:32 PM