GNAT & POSIX

GNAT & POSIX

[mcriley on BIX]

I've got version 2.03 of GNAT for DOS and I can't get GNATBL
to link my test program.

To test out linking C code with GNAT, I wrote a package spec
that declares the C routines, then used pragma Import to bring
them in.  Then I have a main program that calls these routines.

When using GNATBL to link I get an abundance of 'Undefined
references', _not_ referring to the C routines I'm linking, those
are resoved quite nicely, but of a whole slew of POSIX threads
functions.  These are all coming out of Interfaces.C.POSIX_<stuff>.

In no way, shape, or form, am I knowingly interacting with POSIX,
and I'd rather keep it that way if possible!  It seems that these
functions are brought in as a side-effect of including a
pragma Import(C, ...).

I've been wrestling with this for several days and any assistance
would be most welcome.  Thanks.

Marc A. Criley

Re: GNAT & POSIX

[Kenneth Almquist]

mcriley@BIX.com wrote:
> When using GNATBL to link I get an abundance of 'Undefined
> references', _not_ referring to the C routines I'm linking, those
> are resoved quite nicely, but of a whole slew of POSIX threads
> functions.  These are all coming out of Interfaces.C.POSIX_<stuff>.

This sounds like you are using Ada tasks on a platform that doesn't
support them.  The run time system for GNAT (named GNARL) includes a
bunch of tasking code which is loaded if your program uses any of the
tasking features of Ada (including the "delay" statement).  Unless
you are running the OS/2 version of GNAT, the tasking code uses POSIX
threads, and will generate "Undefined references" on platforms which
don't support POSIX threads.
				Kenneth Almquist