FreeTDS Development Group

[Frediano Ziglio <freddyz77 AT tin.it>]

Il lun, 2002-11-04 alle 21:05, Castellano, Nicholas ha scritto:
> I have just completed several passes over the FreeTDS code base, making much
> more widespread use of "const" to indicate data items which should be
> considered read-only.
> 
> This was useful because it helped located one serious bug
> (tds_money_to_string was permanently altering its money input parameter) and
> several smaller problems.
> 
> I have restricted myself to changing only "internal" interfaces to use
> const.  I would like to continue this effort and include exposed portions of
> our API.  For example, I would like to change
>       RETCODE dbcmd(DBPROCESS *dbproc, char *cmdstring);
> to
>       RETCODE dbcmd(DBPROCESS *dbproc, const char *cmdstring);
> because dbcmd does not (and should not) alter the cmdstring argument.
> 
> There is no real technical reason not to do this.  However, there is one
> aesthetic argument against it:  Sybase does not use "const" in their
> prototypes for these functions.
> 
> The real advantage is that it might help spot more bugs in FreeTDS and in
> user code calling it, with the help of compiler warnings.
> 
> Does anyone object to this change?
> 

No, placing constant here is not a problem nor for caller (char* can
always be converted to const char* and without explicit cast of
warnings) nor for library implementation/compatibility (C has no name
mangling like C++, so internal function name is the same)

> Cheers,
> --nick
> 
freddy77