OK, thanks. I've added a C++ section to the wiki, in which I have put the
5 points below: http://cocci.ekstranet.diku.dk/wiki/doku.php?id=start
I will also add the various references that have been contributed.
Unfortunately none of the 5 problems struck me as being completely trivial
to handle. I will look at them more at some point.
I ended up with the following regular expression:
(letter | '$') (letter | digit | '$' | '~') *
("::" (letter | '$' | '~') (letter | digit | '$' | '~') *
('<' (letter | '$' | '~') (letter | digit | '$' | '~') * '>') ?) *
This only allows the < and > at specific places. It will be in the next
release, with the commandline argument -c++
julia
On Wed, 22 Dec 2010, Michael T wrote:
> Julia Lawall <julia <at> diku.dk> writes:
> > On Fri, 17 Dec 2010, Michael T wrote:
> > > Thanks! I will test this next week if that is OK. I think that the
> > > rules on
> > > the use of [<>] should be enough to keep those cases separate.
> [...]
> > OK, if you feel inspired, you can probable play around with the regular
> > expression in lexer.mll. In the patch you can see the two lines that are
> > relevant; you shouldn't have to change anything else.
>
> I did try playing around with the regex (please find a slightly modified
> version
> of your patch below) and got a result that did a lot better for me without
> changing the output of "make test" as compared to an unpatched build,
> timestamps
> excepted of course. (You might want to double check that of course if you do
> apply the patch.)
>
> It also predictably brought out a number of other C++ parsing problems. I
> will
> summarise them briefly and can also bake up some C++ source that illustrates
> their usage if you like. I can also put them on a wiki as you suggested if
> you
> can point to a place. So:
>
> 1) In C++, destructors (and constructors for that matter) don't take a return
> type (they basically return void, but some clever person decided that it
> should
> be left empty). Example of a destructor definition for class MyClass:
>
> MyClass::~MyClass() { ... }
>
> 2) When objects are constructed they can take a parameter list. Example of a
> local (object) variable of type MyClass:
>
> MyClass myinstance(param1, param2);
>
> 3) In C++, objects are created and destroyed on the heap using the operators
> new
> and delete instead of the functions malloc() and free(). The last example
> using
> dynamic creation and destruction:
>
> MyClass *pmyinstance = new MyClass(param1, param2);
> delete pmyinstance;
>
> 4) The nastiest so far. Constructor definitions can take initialisation lists
> which give the default values of the class members. If the members are
> themselves classes then the initialisation list is a parameter list as in 2).
> Example of a contructor for MyClass:
>
> MyClass::MyClass(type1 param1, type2 param2) : member1(value1),
> member2(value2),
> member3(param3-1, param3-2)
> { ... }
>
> 5) C++ has references as well as pointers. Think of them as almost the same
> thing but written differently ('&' instead of '*'). Example in a function
> declaration:
>
> int &myfunc(int ¶m);
>
> Hoping that I haven't made you regret looking at this in the first place (as I
> mentioned before, I think this is a case of "the more the merrier", not "all
> or
> nothing"), here is the patch.
>
> Regards,
>
> Michael
>
> --- a/parsing_c/lexer_c.mll
> +++ b/parsing_c/lexer_c.mll
> @@ -229,6 +229,7 @@ let ulong = (UnSigned,CLong)
>
>
> (*****************************************************************************)
> let letter = ['A'-'Z' 'a'-'z' '_']
> +let extended_letter = ['A'-'Z' 'a'-'z' '_' ':' '<' '>' '~'] (* for c++ *)
> let digit = ['0'-'9']
>
> (* not used for the moment *)
> @@ -643,7 +644,7 @@ rule token = parse
> * truncate to 31 when compare and truncate to 6 and even lowerise
> * in the external linkage phase
> *)
> - | letter (letter | digit) *
> + | ( letter (extended_letter | digit) * ':' '~' ?) ? letter (letter |
> digit) *
> { let info = tokinfo lexbuf in
> let s = tok lexbuf in
> Common.profile_code "C parsing.lex_ident" (fun () ->
>
>
> _______________________________________________
> Cocci mailing list
> [email protected]
> http://lists.diku.dk/mailman/listinfo/cocci
> (Web access from inside DIKUs LAN only)
>
_______________________________________________
Cocci mailing list
[email protected]
http://lists.diku.dk/mailman/listinfo/cocci
(Web access from inside DIKUs LAN only)
