Nick Burrett nick at sqrt.co.uk
Sun Dec 26 08:56:50 PST 2004

Peter Naulls said:
> In message <3035. at>
>           "Nick Burrett" <nick at sqrt.co.uk> wrote:
>> > The reason I ask is that I'm trying to use -nostdinc, and it's
>> defeated
>> > by the above flag in GCC's internal specs.  I can work around this,
>> but
>> > it does cause a little bit of hassle.
>> It's rare to use this, unless you're trying to target something that
>> does
>> not require a run-time library.  Does -nostdinc not remove the path to
>> the
>> unixlib system headers ?  Have you tried using '-isystem' like we do
>> when
>> compiling UnixLib ?
> I am making substantial changes to the orangisation of Unixlib files due
> to glibc file imports.  glibc has a number of instances of filenames
> with the same name, so the exact way files are included is much more
> important than it was previously.

This might horribly clash with the changes I'm making for the GCC-4.0
port.  Surely these filenames of the same name cannot exist within the
'clib' directory ?  I would have thought that such files were private to
the implementation of certain functions and should therefore exist within
the directory of the source file.  In which case we use '-I .' and make
sure that we use:

 #include "foo.h"
instead of
 #include <foo.h>

> I do make use of -isystem (although for precise reasons I forget, since
> I added those a while ago).  -nostdinc removes /home/riscos/env/include
> but not /home/riscos/env/include/unixlib/

Probably because the -icrossdirafter is specified in CPP_SPEC in
riscos-aof.h, so would be like you adding -nostdinc and then a
-icrossdirafter on the same command line.  It was a trade off.


More information about the gcc mailing list