Re: [evolution-patches] Cleanups for evolution/calendar

[Kjartan Maraas <kmaraas broadpark no>]

tor, 17,.08.2006 kl. 11.52 +0530, skrev Srinivasa Ragavan:
> Hi Kjartan,
> 
> On Wed, 2006-08-16 at 20:35 +0200, Kjartan Maraas wrote:
> > ons, 16,.08.2006 kl. 13.01 -0400, skrev Joe Shaw:
> > > Hi,
> > > 
> > > On Wed, 2006-08-16 at 18:22 +0200, Kjartan Maraas wrote:
> > > > > > > On Wed, 2006-08-16 at 16:53 +0200, Kjartan Maraas wrote:
> > > > > > > > -               d (printf("%s:%d (list_changed_cb) - Removing Calendar %s\n", __FILE__, __LINE__, l->data));            
> > > > > > > > +               d (printf("%p:%d (list_changed_cb) - Removing Calendar %s\n", __FILE__, __LINE__, l->data));            
> > > >
> > > > I'm a bit baffled by the use of %d for all these structs and pointers.
> > > > If the debug output is supposed to show strings to the user they should
> > > > all be %s and casted to char * I guess?
> > > 
> > > %d is only used for the second value, __LINE__, which is a special
> > > preprocessor macro set by the compiler.  My understanding is that
> > > __LINE__ is just an integer value.  Hence the use of %d.
> > > 
> > > Likewise, __FILE__ is set in the same way, but it is a string constant.
> > > Thus, %s.
> > > 
> > > The warning is actually for "%s" but references the 4th argument (the
> > > format string first, __FILE__ second, __LINE__ third, l->data fourth).
> > > The issue is that l->data is defined as a gpointer, but %s implies char
> > > *.
> > > 
> > > l is an element in a GList or GSList and it was designed to be as
> > > generic as possible.  That's why the data value is defines as a
> > > gpointer: you can store a pointer to a string, pointer to a structure,
> > > even an integer casted to a pointer in there.  But the compiler doesn't
> > > know exactly what it is and can't automatically cast from a gpointer to
> > > the appropriate type.
> > 
> > Yeah, I follow that.
> > 
> > > Therefore, it's important for the developer to know what type l->data
> > > really is.  I presume it's a string since the %s was already in the
> > > format specifier, so just casting l->data to char * is probably the
> > > right thing.  Otherwise someone probably would have noticed garbage
> > > being printed on the console by now.
> > > 
> > Looking through the rest of the file almost all debug printouts use %d
> > and try to print out pointers to structs, strings and so on. I'm not
> > sure what info is really wanted there so maybe the people who added the
> > debug statements in the alarm code could chime in and reveal what was
> > intended here? :-)
> > 
> It is a string and casting to (char *) should just solve the problem.
> You are right, the %d has to be converted to %p. It was added to debug
> the multiple corruptions happening in alarm code. 
> 
Still a bit confused here. Do you want the relevant %d changed to %s and
the argument casted to char * or should I just change the format
specifier to %p?

Cheers
Kjartan