Re: white space after star unexpected

John E Hein wrote:
> Moore, Robert wrote at 09:34 -0700 on Jun 19, 2008:
>  > I understand quite well.
>  > 
>  > Indent needs to be smarter.
>
> I don't think anyone would disagree (well, some would).  But it has
> never been a full c parser (let alone c++).  No one has ever put forth
> the patches to make that work.  indent-ng would be an interesting SoC.
>
>
> _______________________________________________
> bug-indent mailing list
> [email protected]
> http://lists.gnu.org/mailman/listinfo/bug-indent
>
>
>   
Well! This has become the liveliest debate on indent for quite some time!
As Robert has said, Indent needs to be smarter. However there are some 
obstacles to achieving that:
1) The parsing job for a smart indent is lots harder than for a C 
compiler as it has to

    * understand C and dialects thereof, as well as C++ (according to
      lots many respondents)
    * understand and manage white-space and various styles of
      commenting, all of which a C compiler can just throw away.
    * understand the preprocessor syntax and process it while still
      keeping the original source file content intact.
    * understand the mixes of all these items together. (eg what about
      indenting inside #ifdef ... #else .. #endif)

2) The current code of indent is not at all well structured for 
achieving this. The original developer said at the time that it was just 
a quick hack, and it has been hacked ever since by hanging more and more 
extra bits off it, most of which were done without fully understanding 
the existing code.

I would love to have the time to work on an indent-ng, but with having 
to earn a living in a very demanding job, maintain my home and give my 
family the time they deserve I have no time over.


On the subject of the original mail that triggered all this, * is used 
in C to declare a pointer, as in the prototype that was complained 
about. It is also used as a dereferencing unary operator (e.g. *pi = 10 
+ *pj;) and a multiplication binary operator (e.g. i = 10 * x;). Indent 
is not smart enough to know the context unless it knows that the word 
before the * is a type - which it has to be told as it doesn't 
pre-process the code.

Regarding the request to change the behaviour, many users don't see that 
space as "unexpected". I have see all of the following as preferred 
styles for prototypes:

void func(type * p);
void func(type* p);
void func (type *p);
void func (type*); /* etc */

I personally don't consider any of them worth a flame war and I don't 
think any of them worth my effort to extend indent to support them - but 
that;'s maybe because my favourite IS supported :^). Anyone who does 
care enough is welcome to submit an update and I will integrate it, but 
the update must satisfy the regression test, have a new test added to 
the regression test suite to prove the correctness of the update and the 
documentation must be extended.