[bind10-dev] coding guidelines

[Francis Dupont]

> > (FD)=> the concern I have about " " is it doesn't say to ignore the -I
> > paths, it just changes the order to search is done. More, this order
> > change is compiler dependent, e.g., gcc and msvc don't use the same!
> > And the C11 standard says if the " " specific search fails the include
> > must be interpreted as with < >, so there is no way to "protect" a
> > " " include.
> 
> Well, it's not only about order. The < > ones do _not_ include the local
> directory (or, are not required to). And the " " ones are required to search
> them first.

=> this is what I believed but I checked in the standard (C11 N1570) and
it doesn't say that (it just says "implementation-defined manner").
Of course I shan't object if you assume this anyway...

> So, I'm not saying the behaviour is that " " searches only the local
> directory, or whatever.

=> but IMHO it is unfortunate it is not true as it would make " "
far more interesting to use.

> I'm saying that with the usage I described, you should be safe.

=> my concern is about the "should" (I'd have prefered a stronger term).

> > => anyway there should be something in the guidelines and this must be
> > enforced at the first occasion. For instance I am porting bind10 to WIN32
> > (cf trac2351) and I can change in
> > src/lib/exceptions/exceptions.cc
> 
> Yes, there should. I asked to add a topic for today call to agree on it.
> Currently, there's no guideline and the code reflects that.

=> this is a point where we can't more agree isn't it (:-)?

> There's the light at the end of the Windows.

=> I am not at all a Windows addict but IMHO you can't pretend for
a reasonable market without Windows support (i.e., it was just not
a priority).

Regards

Francis Dupont <fdupont at isc.org>