Re: Fix, GDL2, EOQualifier (getKey()): args

[Tim McIntosh]

On Apr 23, 2008, at 6:34 AM, David Ayers wrote:
> > this is addressing changes in EOQualifier getKey(), made on March 30.
> >
> > > 2008-03-30  Matt Rice  <ratmice@gmail.com>
> > >
> > >   (getKey): Don't use a pointer to a va_list. See
> > >   http://gcc.gnu.org/bugzilla/show_bug.cgi?id=14557
> >
> > This doesn't work without a pointer to args (I am using Mac OS 10.4  
> > with
> > Intel CPU). It will just stick with the first argument, so  
> > obviously the
> > pointer incremented in the use of va_arg() will not return.
> > Quote from the discussion on bugzilla:
> >
> > > I don't see anything wrong with using "va_list *". The standard even
> > says:
> > >    It is permitted to create a pointer to a va_list and pass that
> > >    pointer to another function, in which case the original function
> > >    may make further use of the original list after the other
> > >    function returns.
> > >
> > > About passing by value it says:
> > >
> > >    The object ap may be passed as an argument to another function;
> > >    if that function invokes the va_arg macro with parameter ap, the
> > >    value of ap in the calling function is indeterminate and shall be
> > >    passed to the va_end macro prior to any further reference to ap
>
> Thanks for the report and the patch!  The change was provoked by  
> issues
> compiling on x86_64.  I don't suppose you can test it there?  What  
> would
> be great is if we could extract the relevant code into a simple test
> case for gcc to insure that our approach is portable.

What I gather from the above is:

1. After passing ap to another function by value, it is not valid to  
use ap in the caller without re-initializing it.  Currently  
_qualifierWithArgs() is violating this rule by passing 'args' in two  
sequential calls to getKey().

2. It is valid to do this if ap is instead passed by pointer (as in  
the original code and proposed changes).

3. Taking the address of a va_list that has been passed by value is  
unportable (see GCC bug comment #14).  The original code and the  
proposed changes do this.  It looks like this could be fixed by  
changing _qualifierWithArgs() to take 'va_list *' instead of 'va_list'  
and eliminating (or similarly changing the signature of) - 
qualifierWithQualifierFormat:varargList:.

-Tim