Just downloaded AST->88 NewCompiler->117 from SqueakMap

ClosureCompiler evaluate: '##' and co leads to Syntax errors.

#1 #$ #) etc are also rejected.

BUT ClosureCompiler evaluate: '#:'. answer an empty string HUH?
MIGHT BE A BUG

ClosureCompiler evaluate: '#::'.
#:SmaCCParserError: Token not expected ->:

'# a' is understood as #a (like VW).
'#x:y:z' is understood as #x:y: z (send message z) like VW.

Of course it accepts more than 2 binary letters to form a binary 
symbol/message (already an old compiler feature, but not a st-80 nor VW 
one):
ClosureCompiler evaluate: '#+++'.
  #'+++'

Interesting thing: ClosureCompiler does understand #|| and #--
This is not in the original st-80 but is a logical feature since also 
accept || and -- as regular binary messages.

So finally, only a bug now.
And Symbol storeString still not in sync with Parser.
This will cause decompiled code to be different from source code
#'x:y:z' does print #x:y:z that is interpreted (#x:y:) z BUG.
(see Scanner isLiteralSymbol: which was the real focus of my initial 
post before i ran into weird old compiler behaviour).


stephane ducasse a écrit :
> did you try with the new compiler?
> 
> On 23 mai 07, at 01:47, nicolas cellier wrote:
> 
>> Hello list,
>>
>> Today's quizz: guess the results of these evaluations and how it prints
>>
>> Easy to begin with:
>>     Compiler evaluate: '#++'.
>> Yes Scanner class>>#isLiteralSymbol: not really consistent with what 
>> the parser is ready to understand. It even has some unreachable code 
>> in it.
>>
>> Now what if:
>>     Compiler evaluate: '##'.
>> Enjoy the variations, not all correct:
>>     Compiler evaluate: '#'.
>>     Compiler evaluate: '# "please" # "print" # "me"'.
>>     Compiler evaluate: '##()'.
>>     Compiler evaluate: '##)'.
>>     Compiler evaluate: '##('.
>>     Compiler evaluate: '##(##)'.
>> Never heard about explicit number literal?
>>     Compiler evaluate: '#1'.
>> Maybe the negative ones would work?
>>     Compiler evaluate: '#-1'.
>> So try and explain this one:
>>     Compiler evaluate: '#--1'.
>>
>> Hmm, better correcting the parser than trying to write its EBNF rules!
>>
>> Nicolas
>>
>>
>>
> 
> 
>