Re: [Swig-devel] Parsing comments in c/c++ sources

[Oliver B. <oli...@go...>]

On 30.07.2012 21:37, William S Fulton wrote:
> On 29/07/12 10:12, дмитрий кабак wrote:
>> Hello.
>>
>> I'm a GSoC participant working on doxygen comments parsing and
>> translating. And currently I have some problems in extracting comments
>> from the source files in parser.y. The problem is that parser fails when
>> there is no node to attach comment to.
>> I'm defining comment rules like that:
>>
>> interface      : interface declaration {
>>
>>                      /* add declaration to end of linked list (the
>> declaration isn't always a single declaration, sometimes it is a linked
>> list itself) */
>>
>>                      appendChild($1,$2);
>>
>>                      $$ = $1;
>>
>>                  }
>>
>>                  | interface doxygen_comment declaration {
>>
>>                      set_comment($3, $2);
>>
>>                      appendChild($1, $3);
>>
>>                      $$ = $1;
>>
>>                  }
>>
>>                  | interface declaration doxygen_post_comment {
>>
>>                      set_comment($2, $3);
>>
>>                      appendChild($1, $2);
>>
>>                      $$ = $1;
>>
>>                  }
>>
>> And that was giving "Syntax error in input" for example for empty file
>> with no node to attach doxygen_comment to. Then I modified that:
>>
>> interface      : ...
>>
>>                  | interface doxygen_comment {
>>
>>                      $$ = $1;
>>
>>                  }
>>
>> But that produces lots of reduce/reduce warnings and breaks the whole
>> comments parsing. And I don't understand why is it so, doxygen_comment
>> and doxygen_post_comment are defined as a simple DOXYGENSTRINGs tokens,
>> and there should be no problems.
>> As I'm not very good at bison, I'm asking for your help. I've tried lots
>> of lots of variations of rules, but nothing seems to work.
>> --
> Hi Dmitry
>
> I'm not sure there is much bison expertise here, I wish there was!
> Perhaps the bison mailing lists could help out more? Did you find out
> from the bison documentation what sort of things cause reduce/reduce
> warnings? Probably best to go back to basics.
>
> William
>
Actually I have quite a lot of experience in that...
but I am totally running out of time currently.

Those warnings can usually be resolved (easily) by splitting (too 
convenient) expressions into more atomic ones...
Though it takes time to analyze which I don't have right now...
Any other helpers out there?

Oliver