[libcamera-devel] [PATCH v7 04/13] Add Python bindings

[Kieran Bingham]

Quoting Tomi Valkeinen (2022-05-05 15:03:59)
> On 05/05/2022 16:51, Kieran Bingham wrote:
> 
> >> +       pyCamera
> >> +               .def_property_readonly("id", &Camera::id)
> >> +               .def("acquire", &Camera::acquire)
> >> +               .def("release", &Camera::release)
> >> +               .def("start", [](Camera &self) {
> >> +                       self.requestCompleted.connect(handleRequestCompleted);
> > 
> > What happens if someone calls start() multiple times? Is that going to
> > mean the signal / slot will be connected multiple times?
> > 
> > That could be problematic ... as I think it means it could then call the
> > request completed handler multiple times - but it may also already be
> > trapped by the recent work we did around there.
> > 
> > So for now I think this is fine - but may be something to test or
> > consider later.
> 
> Good point. I'll look at it.
> 
> >> diff --git a/subprojects/pybind11.wrap b/subprojects/pybind11.wrap
> >> new file mode 100644
> >> index 00000000..2413e9ca
> >> --- /dev/null
> >> +++ b/subprojects/pybind11.wrap
> >> @@ -0,0 +1,8 @@
> >> +[wrap-git]
> >> +url = https://github.com/pybind/pybind11.git
> >> +revision = 82734801f23314b4c34d70a79509e060a2648e04
> > 
> > Great - this is now pointing directly to the pybind sources so I think
> > that's much cleaner:
> > 
> > It's hard to do a dedicated review on so much that I honestly don't
> > fully comprehend yet - but I think getting this in and working/building
> > on top is better for everyone, and as far as I know - it's already quite
> > well tested by RPi (albeit, an older version, so I hope David can rebase
> > and test sometime soon).
> 
> Yes, I think David rebasing their work and testing it is an important 
> step before we can consider merging this.
> 
> Hint, hint, David! ;)
> 
> Note that the API has changed a bit. The fb mmap is different, and the 
> ColorSpace enums are nested. I think those are the only API changes.
> 
> Another thing, Python bindings are currently enabled by default if the 
> dependencies are there. I was wondering if the bindings should be 
> disabled by default until we're more confident that they build fine.

I'm fine with them building when dependencies are met. If that catches
failures early - that's better in my opinion.

--
Kieran


> 
>   Tomi