[SR-Dev] git:sip-router: mod if: more prototypes and defines

[Andrei Pelinescu-Onciul]

A few corrections (inline).

On Nov 20, 2008 at 18:43, Andrei Pelinescu-Onciul <andrei at iptel.org> wrote:
> On Nov 20, 2008 at 18:02, Henning Westerholt <henning.westerholt at 1und1.de> wrote:
> > On Thursday 20 November 2008, Andrei Pelinescu-Onciul wrote:
> > > [..]
> > > Example for a variable number of parameters function:
> > >
> > > static cmd_export_t cmds[]={
> > >     {"print", print_f_0, 0, 0, REQUEST_ROUTE},   // overload test
> > >     {"print", print_f_1, 1, print_fixup_f_1, REQUEST_ROUTE},
> > >     {"print", print_f_2, 2, print_fixup_f_2, REQUEST_ROUTE},
> > >     {"print", (cmd_function)print_f_var, VAR_PARAM_NO, 0,  REQUEST_ROUTE},
> > >      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
> > >     {0, 0, 0, 0, 0}
> > > };
> > 
> > Hi Andrei,
> > 
> > thank you for providing this. I found working with a bunch of function 
> > parametern is much more natural then providing different "avp_foo" variables 
> > that are used in other modules as a workaround. Where one should check for a 
> > valid number of parameters  when this VAR_PARAM_NO is used? In the fixup 
> > function, or the function implementation?
> 
> In the implementation is easier.
> 
> In the fixup is possible, but would require using something like:
>  action_u_t* pno = (char*)param - &(((action_u_t*)0)->u.string) -
>                    (param_no+1)*siezof(action_u_t);
                      ^^^^^^^^^
                       it should be only param_no (because the param. no
                       starts from 1 and not from 0)

> parameter_no=pno->u.number.
> 
> (for static int print_fixup_f_2(void **param, int param_no)).
> 
> Another nicer way to do it is: 
> 
> pno=fixup_get_param(param, param_no, -1);
                      ^^^^^^^^^^^^^^^^^^^^
                      that's wrong it should be 
                      fixup_get_param(param, param_no, 0)

> parameter_no=pno->u.number;
> 
> (all this takes into accounts that the fixup functions get a pointer
>  to an action_u_t u.string member and that before the parameters
>  the number of parameters is stored, slightly different than kamailio)
> 
> We could make a macro for it to hide the ugliness and param. arrays
> implementation details.

There is already a function that does this:

fixup_get_param_count(void** cur_param, int cur_param_no)

returns the number of parameters on success or -1 on error
(but it can't fail so -1 is for future use).

Andrei