gst-plugins-bad: dvbsrc: delay locking till actually needed

[Reynaldo H. Verdejo Pinochet]

Hi

On 09/04/2014 08:34 PM, Jan Schmidt wrote:
> Hiya,
> 
> This commit doesn't make sense to me. There's no point locking around
> the do_tune() to prevent the properties being changed at the wrong
> moment if there's no locking in the set_property call.
> [..]

The point is avoiding clashes in between gst_dvbsrc_create()
and _tune().

The current logic is that the device can't be tuned while
reading. This might not be strictly required but it's the
way it's been done right ATM[*]. Now, set_property()
wasn't acquiring the lock for any of the other standard's
tune-related props. It was me the one adding the locking
at set_property for the ISDB-T tuning params when adding
support for this standard but it was inconsistent with the
rest of the set_property code.

After this commit the ISDB-T behavior just matches the rest
while making sure the above mentioned design "limitation"[*]
is respected. It's arguably not perfect but at least it's
consistent.

Your observation did leave me thinking tough, on whether
the lock should be adquired at _tune() rather than at
_set_fe_params(). It does seem a bit racy right now, will
take another look at this tomorrow.

Thanks for taking a look,


-- 
Reynaldo H. Verdejo Pinochet
Open Source Group
Samsung Research America / Silicon Valley