Re: [Help-smalltalk] _gst_invalidate_method_cache and _gst_reset_inline_caches

[Holger Hans Peter Freyther]

On Sat, Dec 21, 2013 at 09:38:26PM +0100, Paolo Bonzini wrote:

> I think the species of MethodDictionary should be LookupTable?

It is more complicated than this. E.g. with the >>#remove:


 remove: anAssociation [
        ...
           copy := self copy.
           result := copy dangerouslyRemove: anAssociation.
           self become: copy.
           Behavior flushCache].
        ...
    ]


So >>#copy should return a MethodDictionary. I considered just
changing >>#copyEmpty: but >>#shallowCopy is already using this.

What about introducing a >>#fastSelect: inside VisualGST to
avoid the flush? In the case of Doit/DoIt in VisualGST and Behavior..
what if we introduce a method that just creates a CompiledMethod
but does not add it to a method dictionary?

holger



Example issue while running the test code:

Swazoo.SwazooCompilerTest>>#testEvaluate did not understand 
#dangerouslyRemoveKey:
TestVerboseLog>>logError: (SUnit.star#VFS.ZipFile/SUnit.st:609)
Swazoo.SwazooCompilerTest(TestCase)>>logError: 
(SUnit.star#VFS.ZipFile/SUnit.st:877)
[] in TestResult>>runCase: (SUnit.star#VFS.ZipFile/SUnit.st:443)
MessageNotUnderstood(Exception)>>activateHandler: (ExcHandling.st:515)
MessageNotUnderstood(Exception)>>signal (ExcHandling.st:254)
LookupTable(Object)>>doesNotUnderstand: #dangerouslyRemoveKey: 
(SysExcept.st:1408)
[] in MethodDictionary>>removeKey:ifAbsent: (MethodDict.st:108)
[] in Semaphore>>critical: (Semaphore.st:80)
BlockClosure>>ensure: (BlkClosure.st:271)
Semaphore>>critical: (Semaphore.st:60)
MethodDictionary>>removeKey:ifAbsent: (MethodDict.st:111)
Metaclass(Behavior)>>removeSelector:ifAbsent: (Behavior.st:261)
Behavior class(Behavior)>>evalString:to: (Behavior.st:432)
Behavior class(Behavior)>>evaluate:to: (Behavior.st:503)
SpEnvironment class>>evaluate:receiver:in: (Sport.star#VFS.ZipFile/sport.st:385)