Re: CVS 8/2 compile error

[Dan Williams <dcbw redhat com>]

On Fri, 2005-02-11 at 09:24 -0500, Colin Walters wrote:
> On Tue, 2005-02-08 at 13:34 +0100, Tom Parker wrote:
> >Getting errors on NetworkManagerDevice.c, because I don't appear to have 
> >a "pci/types.h". No idea where the heck this would be (on a Debian 
> >system), but then worked out that we only needed a few types from it and 
> >so wrote a patch.
> >
> >Index: src/NetworkManagerDevice.c
> >===================================================================
> >RCS file: /cvs/gnome/NetworkManager/src/NetworkManagerDevice.c,v
> >retrieving revision 1.91
> >diff -u -r1.91 NetworkManagerDevice.c
> >--- src/NetworkManagerDevice.c  7 Feb 2005 23:04:05 -0000       1.91
> >+++ src/NetworkManagerDevice.c  8 Feb 2005 12:32:11 -0000
> >@@ -3600,7 +3600,10 @@
> >  /**************************************/
> >  /*    Ethtool capability detection    */
> >  /**************************************/
> >-#include <pci/types.h>
> >+typedef unsigned long long u64;
> >+typedef unsigned int u32;
> >+typedef unsigned short u16;
> >+typedef unsigned char u8;
> >  #include <linux/sockios.h>
> >  #include <linux/ethtool.h>
> 
> Hmm.  Why not port the code to the glib typedefs (e.g. guint32) instead?
> While those typedefs are correct (AFAIK) on all the important platforms
> that Linux runs on today, there's no guarantee that in the future e.g.
> "int" will be 32 bits.  On such a platform the code will probably start
> failing mysteriously.

_Our_ code doesn't use those, but /usr/include/linux/mii.h
and /usr/include/linux/ethtool.h do use them and that's why we appear to
need these typedefs.  Obviously we can't change kernel headers from
NetworkManager...

I'll yell at jgarzik since both those files are his :)

Dan