[PATCH] libcamera: delayed_controls: Inherit from Object class

[Stanislaw Gruszka]

Hi Laurent,

On Mon, Feb 24, 2025 at 11:24:54PM +0200, Laurent Pinchart wrote:
> Hi Stanislaw,
> 
> On Mon, Feb 24, 2025 at 11:16:45AM +0100, Stanislaw Gruszka wrote:
> > On Mon, Feb 24, 2025 at 03:19:34AM +0200, Laurent Pinchart wrote:
> > > A second use-after-free bug related to signals staying connected after
> > > the receiver DelayedControls instance gets deleted has been found, this
> > > time in the simple pipeline handler. Fix the issue once and for all by
> > > making the DelayedControls class inherit from Object. This will
> > > disconnect signals automatically upon deletion of the receiver.
> > > 
> > > Signed-off-by: Laurent Pinchart <laurent.pinchart at ideasonboard.com>
> > > ---
> > > Stan, would you be able to test this with the simple pipeline handler ?
> > > It should work both with and without your series that deals with the
> > > frame start signal, and should fix the crash that Kieran has reported.
> > 
> > It does not make assertion
> > "state_ == ProxyRunning" failed in processStatsThread()
> > gone for me. 
> 
> Right, that one is addressed by Milan's "PATCH v2 0/5] Fix occasional
> software ISP assertion error on stop" series.
> 
> This patch should fix a different crash, but now that I wrote this,
> Kieran may not have reported it with the simple pipeline handler. He has
> reported it for the rkisp1 though.
> 
> Would you be able to test this patch on top of Milan's series, without
> your changes ? I'd like to make sure it introduces no regression. We can
> then apply your series on top.

Tested on master with Milan's 
[PATCH v3 0/6] Fix occasional software ISP assertion error on stop

"state_ == ProxyRunning" assertion is gone and I did not encounter
any regression with simple pipeline. Things looks good.

Tested-by: Stanislaw Gruszka <stanislaw.gruszka at linux.intel.com>

Regards
Stanislaw