None timeline load

[ThomasWilshaw]

I made a start at #23 and I've managed to (I think) sensibly handle none Timeline schemas. I started looking at suppporting just clips and had a couple of questions:

1. Is this the right sort of approach?
2. For some reason app.cpp:912 was causing the selected_object state to be cleared leading to crashes. Commenting it out fixes the issue but I can't figure out why it's happening in the first place.

This is very much still a WIP but I wanted to check it was heading in the right direction before I go any further.

[jminor]

Yes, this strategy seems like a good start.

You could go one step further and keep root around for later use replacing the top level appState.timeline pointer.

Once loaded, then the call to DrawTimeline inside MainGui could be wrapped in a DrawRoot function that switches on the root object's type to decide how to render it.

[jminor]

Side note for future reference: after the work in this PR is complete, we can implement #72 by making a double-click on a nested composable switch the root to that object, allowing you to drill into the sub-composition. Popping back out of the nested composition would restore root to the previous object.

[ThomasWilshaw]

I'm having a little trouble with this. I've replaced appState.timeline with otio::SerializableObject::Retainer<otio::SerializableObjectWithMetadata> root in app.h and that is generally working fine. However if I try to cast that (dynamic_cast<otio::Timeline*>(appState.root)) I get the following error:

the operand of a pointer dynamic_cast must be a pointer to a complete class type

I think this is probably because I don't fully inderstand how Retainers work but I wondered if you had any suggestions?

[ThomasWilshaw]

I've updated things to store the root OTIO object rather than assuming a timeline and it more or less seems to work. I'm still having an issue where object->to_json_string(&error_status) in SelectObject() is corrupting the root object somehow.

[jminor]

Hmm... I tried it out, and I attempted to diagnose the to_json_string issue - it feels like the returned string is getting deallocated. I can print it within the function, but it gets lost later.

I'm suspicious of the switch from
otio::SerializableObject::Retainer<otio::Timeline> timeline to
otio::SerializableObjectWithMetadata* root
since maybe the root object is getting deallocated. I was able to use otio::SerializableObject::Retainer< otio::SerializableObjectWithMetadata> root by adding some take_value() calls around, but it all falls apart on load.

I'll have to return to this when I have my C++ brain turned back on. I've been writing Swift and Python lately, so the rules of engagement are fuzzy in my head right now. 😳 The needed info is likely in here: https://opentimelineio.readthedocs.io/en/stable/cxx/cxx.html#memory-management