On Thursday, October 16, 2008, 11:41:07 PM, Bill Meier wrote:
> There are about 300+ non-generated epan/dissector source files which
> include stdio.h and/or stdlib.h
> I've successfully done both a Linux and a Windows build with these
> includes removed from epan/dissector/...
Under which compiler/glibc?
I recently had to add in <stdio.h> to a bunch of (local-to-me) source
files because I upgraded to the latest Fedora 9 with g++ 4.3.0 and
that seems to believe that NULL is only defined once <stdio.h> is
*explicitly* pulled in.
There were other similar issues with <string.h> and <stdlib.h> and
possibly a couple of others with various functions too.
(My hastily acquired conclusion is that this compiler/libc combination
appears to have some sort of anti-side-effect system so headers which
are included by other headers don't export their incidental symbols
unless the main file includes them explicitly too. I think I approve
of this in principal, if not the retroactive consequences, but one has
to be careful.)
> So, I plan to commit these changes tomorrow (Fri) unless someone
> indicates a reason that I shouldn't.
Random *local* include files I would whole heartedly agree, but not
system ones unless you have tested on every single current and
historical system variant.
John
--
Dead stars still burn
