Re: Ada in gcc-3.4.3 on x86_64

[Robert Dewar <dewar at adacore dot com>]

Dean Kolosiek wrote:
> I followed the build instructions at http://gcc.gnu.org/install/build.html 
> which does not mention make -C gcc gnatlib_and_tools
>
>
> That worked, and I reproduced a problem I saw in 3.3.3. When calling 
> a C function from Ada there is no code generated to set %al to the number 
> of xmm registers being passed. Variadic C functions use %al to compute a 
> jump into a table of movaps instructions to save the xmm registers. The 
> result is a jump to a random location.

In general, if the calling sequence for a variadic function is not the
same as the normal calling sequence, then there is no way in GNAT to
interface directly to such a function. From the GNAT RM:

> In C, @code{varargs} allows a function to take a variable number of
> arguments. There is no direct equivalent in this to Ada. One
> approach that can be used is to create a C wrapper for each
> different profile and then interface to this C wrapper. For
> example, to print an @code{int} value using @code{printf},
> create a C function @code{printfi} that takes two arguments, a
> pointer to a string and an int, and calls @code{printf}.
> Then in the Ada program, use pragma @code{Import} to
> interface to printfi.
>
> It may work on some platforms to directly interface to
> a @code{varargs} function by providing a specific Ada profile
> for a a particular call. However, this does not work on
> all platforms, since there is no guarantee that the
> calling sequence for a two argument normal C function
> is the same as for calling a @code{varargs} C function with
> the same two arguments.

So there is no bug here, just a case where the "may" of para
2 does not apply. So the proper approach is to create C wrappers
as noted in the first quoted para.