[antlr-interest] gUnit and Error Recovery

[Leon Su]

Hi Louis,

gUnit indeed requires ANTLR error messages to be directed to stderr  
ouput which is by default set up in the emitErrorMessage called by  
displayRecognitionError.
Then, gUnit determines  whether a test succeeds or not based on the  
output from stderr.
Thanks for your investigation and please let me know if you have any  
question.

Leon

On Aug 8, 2008, at 4:12 AM, Louis Rose wrote:

> Hi Leon,
>
> Many thanks for your help.
>
> After a little more investigating, I now believe this is an issue  
> with error reporting.
>
> The grammar that I'm using generates a parser which extend a base  
> class, BaseParser. The BaseParser class overrides the  
> displayRecognitionError method, and redirects the output (to an  
> Eclipse console).
>
> Does gUnit require errors to be reported to System.err? If so, I'll  
> find a way to change BaseParser to allow this.
>
> Cheers,
> Louis.
>
> Leon Su wrote:
>> Hi Louis,
>> I just quickly tested the grammar fragment you provided:
>> remove (cls)* from pkg, and add WS:( ' ' | '\t' | '\r' | '\n' )+  
>> {$channel = HIDDEN;};
>> tested by gUnit1.0.2+ANTLR3.1b2 and gUnit-1.0.1+ANTLR3.0.1
>> and both got the failed 2nd test as we expected ( and received the  
>> error msg from ANTLR: line 0:-1 missing RBRACKET at '<EOF>' )
>> If you're using gunit-1.0.1 currently, don't forget to download  
>> the updated gunit-1.0.1.jar Version2 on the wiki page.
>> Please let me know if you any problem.
>> Leon
>> On Aug 7, 2008, at 10:25 AM, Louis Rose wrote:
>>> Hi all,
>>>
>>> I'm currently considering migrating a large number of unit tests  
>>> to gUnit. I have a query about the way gUnit works with ANTLR's  
>>> error recovery capabilities.
>>>
>>> Here's a fragment from an ANTLR grammar that I wish to test:
>>>
>>> pkg: NAME^ LBRACKET! (cls)* RBRACKET!;
>>>
>>> LBRACKET     : '{';
>>> RBRACKET     : '}';
>>> NAME         : ID_START_LETTER (ID_LETTER)*;
>>>
>>> fragment
>>> ID_START_LETTER
>>>     :    '_'
>>>     |    'a'..'z'
>>>     |    'A'..'Z'
>>>     |    '@'
>>>     ;
>>>
>>> fragment
>>> ID_LETTER
>>>     :    ID_START_LETTER
>>>     |    '0'..'9'
>>>     |    '-'
>>>     ;
>>>
>>>
>>> I have the following gUnit tests:
>>>
>>> pkg:
>>>     "Families {}" -> "Families"
>>>     "Families {"    FAIL
>>>
>>> I would expect the second test to verify that a terminating right  
>>> bracket is required. (i.e. parsing would fail). In fact, the  
>>> parser successfully recognises the input from the second test:
>>>
>>> 1 failures found:
>>> test2 (pkg, line6) -
>>> expected: FAIL
>>> actual: OK
>>>
>>> Tests run: 2, Failures: 1
>>>
>>>
>>> Is this working as intended? How can I test that all uses of pkg  
>>> should be terminated with a closing bracket?
>>>
>>> Many thanks,
>>> Louis Rose.
>>>