On 10.10.2006, at 16:41, Klaus D. Witzel wrote:

> Three things:
>
> 1) found a todo list in CompilerTODO, is this up to date
>

I should remove that class... the todo list used now is

	http://smallwiki.unibe.ch/newcompiler/


> 2) is it expected that a full decompile takes all the steps as in  
> #testDecompileBlockParam (from aCompiledMethod via  
> BytecodeDecompiler via IRDecompiler to AST-Nodes and friends)
>

Yes. But Bytecode --> IR is already done,  AST-> text, too. The only  
thing missing is IR->AST

> 3) afterComment, beforeComment and insideComment are Undeclared  
> (the dev.image comes with AST-ms.88?)
>

You need AST-md.100, it's on SqueakSource, project AST

      Marcus

> /Klaus
>
> On Mon, 09 Oct 2006 19:26:27 +0200, Mathieu wrote:
>> Klaus D. Witzel a écrit :
>>> Hi Mathieu,
>>>
>>> on Mon, 09 Oct 2006 12:43:58 +0200, you wrote:
>>>> Klaus D. Witzel a écrit :
>>>>> On Sun, 08 Oct 2006 13:59:05 +0200, stephane ducasse wrote:
>>>>> ...
>>>>>> If people wants to help we need the decompiler also working well.
>>>>>
>>>>> Can I have a specification (sufficient, please) that describes  
>>>>> what
>>>>> source construct emits what bytecode(s). I cannot promise that  
>>>>> I will
>>>>> then be able to work on the decompiler but I promise to give it  
>>>>> a try.
>>>>>
>>>>
>>>> Forgot to tell if you want to see exacly what happen
>>>
>>> Yes I want to see, but must look into the opposite direction :|  
>>> Let me
>>> ask further: which component of the framework (the slides in
>>> 11Bytecode.pdf are excellent, thank you!) is responsible for pretty
>>> print? The latter is what the browser usually requests from the  
>>> decompiler.
>>
>> Normaly it's RBFormatter who suppose to pretty print it(you can  
>> install other pretty printer like
>> gutenberg).
>> And the compilation stop at the AST representation so you don't  
>> need to compile it completly.
>>
>> see: Compiler>>format:in:notifying:decorated:
>>
>> You also have to understand the Visitor but this one it's a bit  
>> strange beceause when
>> visitMethodNode is expected you will see acceptMethodNode... (HTH)
>>
>> 	Math
>>
>>>
>>> /Klaus
>>>
>>>> you can use #>> and see the bytecode
>>>
>>> Yes, thank you :)
>>>
>>>> Object>>#haltIf:
>>>>
>>>> ->
>>>> 45 <10> pushTemp: 0
>>>> 46 <D0> send: isSymbol
>>>> 47 <AC 17> jumpFalse: 72
>>>> 49 <89> pushThisContext:
>>>> 50 <69> popIntoTemp: 1
>>>> 51 <11> pushTemp: 1
>>>> 52 <D1> send: sender
>>>> 53 <D2> send: isNil
>>>> 54 <A8 0F> jumpTrue: 71
>>>> 56 <11> pushTemp: 1
>>>> 57 <D1> send: sender
>>>> 58 <81 41> storeIntoTemp: 1
>>>> 60 <D3> send: selector
>>>> 61 <10> pushTemp: 0
>>>> 62 <B6> send: =
>>>> ...
>>>>
>>>>>> Mathieu also worked on fixing the way comments are attached in  
>>>>>> the
>>>>>> tree (before this was kind of random).
>>>>>
>>>>> Can't wait to see this in action :)
>>>>>
>>>>> /Klaus
>>>>>
>>>>>> Stef
>>>>>
>>>>>
>>>>>
>>>>
>>>>
>>>>
>>>
>>>
>>>
>>>
>>
>>
>>
>
>
>