GNATLINKing objects that are not in the same directory

GNATLINKing objects that are not in the same directory

[Karen Carron]

I have not been able to get GNATLINK to work when the object files are
not in the same directory as the one I'm doing gnatlink from.

Here is an example of the command I'm using:

gnatlink -aO/home/xxx/dev/test ro.ali tmain.o -lF77 -lsunmath

tmain.o, which contains the main program whose source is written in
FORTRAN, is the file that is in another directory.  I'm not sure if it
makes a difference that it is the main program?  Anyway, I've tried
the -L, -l, and -I options also, at various places on the command line
(before ro.ali, btw. ro.ali and tmain.o, btw. tmain.o and -lF77, and
at the end).  I've also tried putting the path into a file and using
that with the 4 switches (-aO, -L, -l, -I) at all those places on the
command line.  I've also tried eliminating that switch, and setting
the environment variable ADA_OBJECTS_PATH to the search path.  None of
these things have worked.  I get the following error in all cases:

gcc: tmain.o: No such file or directory.

Of course, when I move the file into the directory I'm executing the
gnatlink command from, it works fine. I would like to figure out how
to do this because there is a much bigger system I need to work with. 
I have an example of how this is done in a make file, but I'd rather
use a script for this project.  Any ideas are welcome!

Thanks!

Re: GNATLINKing objects that are not in the same directory

[Simon Wright]

kcarron@belcan.com (Karen Carron) writes:

> I have not been able to get GNATLINK to work when the object files are
> not in the same directory as the one I'm doing gnatlink from.
> 
> Here is an example of the command I'm using:
> 
> gnatlink -aO/home/xxx/dev/test ro.ali tmain.o -lF77 -lsunmath
> 
> tmain.o, which contains the main program whose source is written in
> FORTRAN, is the file that is in another directory.  I'm not sure if it
> makes a difference that it is the main program?  Anyway, I've tried
> the -L, -l, and -I options also, at various places on the command line
> (before ro.ali, btw. ro.ali and tmain.o, btw. tmain.o and -lF77, and
> at the end).  I've also tried putting the path into a file and using
> that with the 4 switches (-aO, -L, -l, -I) at all those places on the
> command line.  I've also tried eliminating that switch, and setting
> the environment variable ADA_OBJECTS_PATH to the search path.  None of
> these things have worked.  I get the following error in all cases:
> 
> gcc: tmain.o: No such file or directory.

I suspect that when -aO says "When searching for library and object
files, look in directory dir" it's talking about what gnatmake does
(the flag is documented under gnatmake, not gnatlink -- 3.14a1) when
working out the closure.

tmain.o is not an Ada object, so it's of no interest to gnatlink
(which is concerned about elaboration). Of course, you have to have it
for ld when gnatlink calls it, cos it has your main() in it ..

You are expecting your executable to be named ro, right? so what is
wrong with

  gnatlink -aO/home/xxx/dev/test ro.ali /some/directory/tmain.o -lF77 -lsunmath

-S