[FFmpeg-devel] [PATCH] define _BSD_SOURCE for bktr.c

Måns Rullgård mans
Fri Dec 26 19:58:55 CET 2008


Diego Biurrun <diego at biurrun.de> writes:

> On Mon, Dec 15, 2008 at 11:02:55PM +0000, Jacob Meuser wrote:
>> On Mon, Dec 15, 2008 at 11:45:10PM +0100, Diego Biurrun wrote:
>> > On Mon, Dec 15, 2008 at 04:53:50PM +0000, Jacob Meuser wrote:
>> > > On Mon, Dec 15, 2008 at 11:32:03AM -0000, M?ns Rullg?rd wrote:
>> > > > 
>> > > > Diego Biurrun wrote:
>> > > > > On Mon, Dec 15, 2008 at 02:50:48AM +0000, Jacob Meuser wrote:
>> > > > >> On Sun, Dec 14, 2008 at 02:10:03PM +0100, Diego Biurrun wrote:
>> > > > >> >
>> > > > >> > The patch has 17 different hunks, hunks 5-7, 13, 14 are
>> > > > >> > applied, hunk 8 has been explained, the rest is
>> > > > >> > mysterious.
>> > > > >>
>> > > > >> (2) line 934 - use the hardware arch instead of machine arch.  this
>> > > > >> was sent to me from another developer.
>> > > > >
>> > > > > Why don't you just pass --arch=WHATEVER to configure?
>> > > > 
>> > > > If OpenBSD has a different mechanism for detecting the architecture,
>> > > > I have nothing against supporting this in a clean way.  I am, however,
>> > > > slightly confused by the hardware vs. machine distinction?  I thought
>> > > > those words were more or less synonymous.
>> > > 
>> > > there are, in OpenBSD, a few differences.  I don't use any of those
>> > > machines, but I believe, e.g. zaurus vs arm is an example, as
>> > > well as macppc vs powerppc.
>> > 
>> > Could you point us at some examples or documentation?
>> 
>> http://www.openbsd.org/cgi-bin/man.cgi?query=arch
>> http://www.openbsd.org/cgi-bin/man.cgi?query=uname
>
> Executive summary: 'uname' outputs the host architecture, 'arch'
> outputs the target architecture.

What you just said makes no sense.  Those commands always report
characteristics of the machine on which they run; cross-compiling is
irrelevant here.

I think the answer is that the BSD uname reports the system type
rather than the CPU architecture, e.g. pc instead of x86.  The
standard is sufficiently vague ("the name of the hardware type") that
an implementation can get away with this, annoying though it may be.

GNU coreutils includes an "arch" command documented as being
equivalent to "uname -m".

Do we know of a system with an "arch" command doing something else
entirely?  If not, we could simply try both.

-- 
M?ns Rullg?rd
mans at mansr.com




More information about the ffmpeg-devel mailing list