[libcamera-devel] libcamera Android Enhancements

[Kieran Bingham]

Quoting Jacopo Mondi (2022-10-28 09:25:12)
> Hi Nicholas
> 
> On Thu, Oct 27, 2022 at 05:41:25PM -0500, Nicholas Roth via libcamera-devel wrote:
> > Fourth round of changes-- made suggested changes.
> 
> I'll skip v4 and get to v5 but I would like you to
> 1) do not send new versions in reply to the previous ones, it's hard
> to follow and might even confuse patchwork

Patchwork looks like it coped ok - but please always add a cover letter
with a title for the series and a version. It's small, but it helps
patchwork have a title and version for the series:

 - https://patchwork.libcamera.org/project/libcamera/list/?series=3579&state=*
 - https://patchwork.libcamera.org/project/libcamera/list/?series=3581&state=*

(Your patches are at
https://patchwork.libcamera.org/project/libcamera/list/?submitter=97)

> 2) When you receive a tag like Reviewed-by or Tested-by or whatever
> please collect it as part of your next version. Copy the tag and paste
> it below your Signed-off-by: in your commit
> 
> >
> > Let me know what needs to happen to get this merged!
> 
> I think there are a few points that needs to be addressed and on which
> you probably want more input from Laurent and Kieran
> 
> 1) The durations patch: it seems you have explored various
>    alternatives. I let Laurent Kieran and Naush comment if they're
>    happy with the final result or not

This one makes me uncomfortable (that the problem exists). This is
defintely one I want feedback from Laurent on. He's on vacation this
week, and part of next - so that will take a bit of time at least.



> 2) Still not sure why you can't use the sensor blankings to initialize
>    FrameDurationLimits but I will comment on the patch
 

This one I think can be done right - but also I think we need to adjust
how the controls are handled/registered by the RKISP and IPU3 so I don't
mind if this is looked at again at that point. This patch isn't "Worse"
than the existing situation - so I'd propose accepting the latest
version to simplify things, and handle the detail later. Specifically,
we'll need to actaully implement handling of the FrameDurationLimits
being set by the application anyway.

> 3) Disabling isolation: this is tough. I think we really don't want to
>    give such an easy way to run un-signed binaries in non-isolated
>    mode. However you have a valid use case. I'll defer this to Kieran
>    and Laurent which are more aware of the implications.

And this one is problematic. We need something better than a switch
indeed.

But lets get this series reduced for the ones that can be merged
already. The patches with tags from Jacopo already are likely suitable
for merge, (with an ack on the RPi parts).

--
Kieran


> 
> >
> > Thanks,
> > -Nicholas
> >
> >