It seems to work but...

Ralf Corsepius corsepiu at faw.uni-ulm.de
Fri Jun 20 02:14:17 UTC 2003


Am Don, 2003-06-19 um 18.59 schrieb John Trostel:
> It's probably a comment I shouldn't worry about, but _why_ am I seeing
> the following statement about a nonexistant directory?
Because you are using gcc -v ;-)

> (Other than that, the app seems to build just dandy)
> 
> ignoring nonexistent directory "/opt/rtems/m68k-rtems/sys-include"
> #include "..." search starts here:
> #include <...> search starts here:
>  /opt/rtems/lib/gcc-lib/m68k-rtems/3.2.3/include
>  /opt/rtems/m68k-rtems/include
> End of search list.
>  /opt/rtems/lib/gcc-lib/m68k-rtems/3.2.3/../../../../m68k-rtems/bin/as
> -mc68020 -o f.o /tmp/ccnXBIMa.s

The sys-include directory referenced in the call to gcc above is
supposed to be meaningless for newlib-based gccs (like rtems
cross-gccs), because newlib based gccs do not apply sys-include.

I.e. using sys-include in the default include path of gccs which
actually doen't use sys-include is not entirely correct, and can be
considered a bug in gcc, IMO.

For rtems-gcc, having sys-include in the default include path definitely
is not correct, but will also not do much harm in most cases (It can be
harmful if using automount or similar)

The work-around to get rid of the warning would be to create an empty 
/opt/rtems/m68k-rtems/sys-include
(I think I'll be changing RTEMS gcc rpm.specs to create the sys-include
directory by default)

An actual fix would be to patch gcc to not using sys-include for
rtems-gccs.

Ralf





More information about the users mailing list