On Wed May 2 2012 23:39:15 Sakari Ailus wrote:
> Hi Hans,
>
> On Wed, May 02, 2012 at 10:45:22PM +0200, Hans Verkuil wrote:
> > On Wed May 2 2012 21:13:47 Sakari Ailus wrote:
> > > Replace enums in IOCTL structs by __u32. The size of enums is variable and
> > > thus problematic. Compatibility structs having exactly the same as
> > > original
> > > definition are provided for compatibility with old binaries with the
> > > required conversion code.
> >
> > Does someone actually have hard proof that there really is a problem? You
> > know,
> > demonstrate it with actual example code?
> >
> > It's pretty horrible that you have to do all those conversions and that code
> > will be with us for years to come.
> >
> > For most (if not all!) architectures sizeof(enum) == sizeof(u32), so there
> > is
> > no need for any compat code for those.
>
> Cases I know where this can go wrong are, but there may well be others:
>
> - ppc64: int is 64 bits there, and thus also enums,
Are you really, really certain that's the case? If I look at
arch/powerpc/include/asm/types.h it includes either asm-generic/int-l64.h
or asm-generic/int-ll64.h and both of those headers define u32 as unsigned int.
Also, if sizeof(int) != 4, then how would you define u32?
Ask a ppc64 kernel maintainer what sizeof(int) and sizeof(enum) are in the
kernel
before we start doing lots of work for no reason.
Looking at arch/*/include/asm/types.h it seems all architectures define
sizeof(int)
as 4.
What sizeof(long) is will actually differ between architectures, but char, short
and int seem to be fixed everywhere.
Regards,
Hans
>
> - Enums are quite a different concept in C++ than in C --- the compiler may
> make assumpton based on the value range of the enums --- videodev2.h should
> be included with extern "C" in that case, though,
>
> - C does not specify which integer type enums actually use; this is what GCC
> manual says about it:
>
>
> <URL:http://www.gnu.org/software/gnu-c-manual/gnu-c-manual.html#Enumerations>
>
> So a compiler other than GCC should use 16-bit enums and conform to C
> while breaking V4L2. This might be a theoretical issue, though.
>
> More discussion took place in this thread:
>
> <URL:http://www.spinics.net/lists/linux-media/msg46167.html>
>
> Regards,
>
>
--
To unsubscribe from this list: send the line "unsubscribe linux-media" in
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html