Re: - c-end-of-defun - !

[Alan Mackenzie <acm@muc.de>]

'evening, Alin and Andreas!

Thank you both very much for reporting this bug!  (Andreas: the bug you
reported 2 hours after Alin is the same bug.)

On Wed, Apr 18, 2007 at 03:17:00PM +0200, A Soare wrote:
> 
> int
> function (void)
> {
>     int b;
>     
> }

> struct ONE function (void)
> {
>     int a;
>     
> }

> struct
> TWO function (void)
> {
>     int a;
>     
> }

> 
> Here are 3 valid functions written in C.
> Copy them into a buffer in C-MODE, and eval (end-of defun) in the first one and in the second.

> In the second case => FAIL.
 
> That is because of this code in c-end-of-defun:
> 
>     ;; Do we need to move forward from the brace to the semicolon?
>     (when (eq arg 0)
>       (if (c-in-function-trailer-p)	; after "}" of struct/enum, etc.
> 	  (c-syntactic-re-search-forward ";"))
 
> (c-in-function-trailer-p) in our case MUST return NULL. Not the (point).
 
This is indeed the case.  
 
> Is really need to call (c-syntactic-re-search-forward ";") here?
 
Yes.  c-end-of-defun is being fooled by the "struct" in the functions'
return types into thinking the function is actually a struct declaration
like this:

struct foo {
    int bar ;
    int baz ;
} blorg ;

A struct also counts as a defun, and it ends at the first semicolon
following the brace (as contrasted with a function, which ends at the
brace).  This is what the (c-syntactic-re-search-forward ";") is for.

The solution is to analyse the defuns' headers more thoroughly.  It
shouldn't be too hard, and shouldn't take too long.

> Alin Soare.

-- 
Alan Mackenzie (Ittersbach, Germany)