get_header_tar.c: makedev [PATCH]

[Peter S. Mazinger]

On Wed, 28 Sep 2005, Rob Landley wrote:

> On Wednesday 28 September 2005 17:14, Shaun Jackman wrote:
> > get_header_tar.c uses the macro makedev, which is not defined by any
> > standard header. In practice, it's defined either by sys/types.h,
> > sys/sysmacros.h, or sys/mkdev.h. glibc defines makedev in
> > sys/sysmacros.h which is included by sys/types.h. To support newlib,
> > include sys/sysmacros.h directly. I armoured the include in an ifdef
> > to allow for a system that defines makedev in sys/types.h but does not
> > provide sys/sysmacros.h.
> 
> If sys/sysmacros.h is the correct #include then we should include it 
> unconditionally.  (If uClibc needs to be fixed, we can fix it.)
> 
> If sys/types.h is the correct #include, then fix newlib.
> 
> How do we go about determining this?

glibc-2.3.5:
defined in sys/sysmacros.h only for __GLIBC_HAVE_LONG_LONG
included by sys/types.h only for __USE_BSD

uClibc-svn:
defined in sys/sysmacros.h unconditionally
included by sys/types.h only for __USE_BSD

dietlibc-cvs:
defined in sys/stat.h unconditionally
sys/types.h does not include sys/stat.h or sys/sysmacros.h
sys/sysmacros.h includes sys/stat.h

klibc-1.1:
defined in sys/sysmacros.h unconditionally
sys/types.h includes sys/sysmacros.h unconditionally
(these 2 headers suggest, that types.h should always be included before 
sysmacros.h)

newlib (combined cvs sources w/ binutils and others, also 1.13.0):
makedev not present at all

Python's pyconfig.h suggests that some could have makedev defined in 
mkdev.h though (but those could have sys/types.h including sys/mkdev.h)

the header rpc/types.h (glibc, uClibc and dietlibc) does:
#ifndef makedev
#include <sys/types.h>
#endif

Well, dietlibc is not quite consequent.

Peter

-- 
Peter S. Mazinger <ps dot m at gmx dot net>           ID: 0xA5F059F2
Key fingerprint = 92A4 31E1 56BC 3D5A 2D08  BB6E C389 975E A5F0 59F2