Am Freitag 05 März 2010 17:35:07 schrieb Bastien Nocera:
> On Sun, 2010-02-21 at 16:45 +0100, Hendrik Sattler wrote:
> <snip>
>
> > > > Additionally to setting self->fd here, you should also modify
> > > > obex_transport_handle_input() to no choose the special USB path when
> > > > self->fd is valid.
> > >
> > > Except that it doesn't work because it doesn't actually select on
> > > self->fd, but on self->fd+1.
> > >
> > > ret = select((int)highestfd+1, &fdset, NULL, NULL, &time);
> > >
> > > I'm pretty sure that's not correct for our case.
> >
> > The select call is correct. The change in obex_transport_handle_input()
> > can be an additional patch.
>
> You seem to have committed this patch despite my comments, which breaks
> osso-gwobex's USB support, as well as obex_test's USB support when built
> with libusb1.
You didn't react on my comment that your comment was plain wrong. We don't do
call select() on fd+1. However, it may be that the POLLIN fd is not usuable
with select()?
The following code adds it to the FD_SET:
FD_ZERO(&fdset);
if (self->fd != INVALID_SOCKET) {
FD_SET(self->fd, &fdset);
if (self->fd > highestfd)
highestfd = self->fd;
}
And the +1 in the select() call is correct but does not affect the self->fd.
In what way does it break for you? (try looking with strace)
> The select call is correct when the ->fd is the output file descriptor,
> which it isn't for USB.
You mean the output file descriptor is for _reading_? The above select() call
watches for readable data, not writable, so POLLIN should be the right fd for
that, isn't it?
> So either the select call needs to be modified, or the patch needs to be
> reverted.
If the select() on the POLLIN fd does really not work for you, then we'll
revert it. It would be nice to see the strace'd select() call first.
HS
PS: sorry, I don't have a USB OBEX device.
------------------------------------------------------------------------------
Download Intel® Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
_______________________________________________
Openobex-users mailing list
Openobex-users@lists.sourceforge.net
http://lists.sourceforge.net/lists/listinfo/openobex-users
