Re: [libdc] Fwd: libdc1394 and ladybug3

[Stefan R. <st...@s5...>]

On Jul 17 Damien Douxchamps wrote:
> Hi David, Josep,
> 
> On Tue, 2013-06-18 at 22:58 -0700, David Moore wrote:
> > 
> > dc1394/juju/capture.c:
> > 
> >     create.speed = SCODE_400;
> 
> Is there a reason for this to be hardcoded? It was in Christian
> Hoegsberg's original juju patch so I guess it must have been a quick
> hack at the time.

The existence of struct fw_cdev_create_iso_context.speed is a little
oddity, due to this struct also being used to create isochronous
transmission contexts (besides reception contexts like libdc1394 uses).

> The last report from Josep even indicate that this setting has little
> effect (the camera operates well at 800Mbps). Josep: could you confirm
> that nothing happens when you change the line to SCODE_100, 200, 800? Or
> even omit the line entirely?

The kernel ABI documentation claims that .speed is ignored if it is an IR
context, and without having re-checked the implementation just now, I am
fairly confident that this is true.

So, we should be fine if the line is omitted.  Some ABI users prefer to
initialize unused ioctl argument members though, to suppress noise from
debugging utilities.  (The alternative to teach the debugging aids about
all the oddities of the myriads of ioctls isn't very attractive either.)
If the latter is referred, maybe write it somehow like so:

	create.speed = 0; /* ignored */

> > To me that looks like it's still hardcoded to operate at 400Mbps.

This particular ioctl only influences the local receiving hardware (which
does not need to know the wire speed), but not the remote transmitting
hardware (which has to be configured for the desired speed of course,
separately from local context creation).
-- 
Stefan Richter
-=====-===-= -=== =----
http://arcgraph.de/sr/