T265: A safe way to hang the camera on macOS, development branch, #5213 applied

[neilyoung]

---

Required Info
Camera Model	T265
Firmware Version	0.2.0.879
Operating System & Version	macOS Catalina
Kernel Version (Linux Only)
Platform
SDK Version	2.31.0, development branch, #5213 applied
Language
Segment	others

Issue Description

Probably related to the problems reported here #5394

Receipt to hang a T265 with the a.m. firmware and SDK under macOS Catalina:

• Take pose example
• Alter these lines

    // Add pose stream
    cfg.enable_stream(RS2_STREAM_POSE, RS2_FORMAT_6DOF);

like so:

    // Add pose stream
    cfg.enable_stream(RS2_STREAM_POSE, RS2_FORMAT_6DOF);
    auto pose_sensor = cfg.resolve(pipe).get_device().first<rs2::pose_sensor>();
    // Options
    pose_sensor.set_option(RS2_OPTION_ENABLE_POSE_JUMPING, 1);
    pose_sensor.set_option(RS2_OPTION_ENABLE_MAPPING, 1);
    pose_sensor.set_option(RS2_OPTION_ENABLE_RELOCALIZATION, 1);
    pose_sensor.set_option(RS2_OPTION_ENABLE_MAP_PRESERVATION, 1);

It should not work anymore.

EDIT: If I omit the set_option calls, the script runs, but you MUST NOT call it too quickly after having terminated it. The promised result of a quick re-open after close is this:

~/Documents/librealsense/build/examples/pose $ ./rs-pose 
RealSense error calling rs2_config_resolve(config:0x7fbf82500c90, pipe:0x7fbf825009d0):
    No device connected

I think it all boils down to some regression in #5213, since I don't remember to have seen this behaviour last week.

[neilyoung]

And would somebody please comment on the internal defaults of these cfg settings?

RS2_OPTION_ENABLE_POSE_JUMPING
RS2_OPTION_ENABLE_MAPPING
RS2_OPTION_ENABLE_RELOCALIZATION
RS2_OPTION_ENABLE_MAP_PRESERVATION

My guess is: true, true, true, false. Is that correct?

EDIT: Yepp... get_option confirms that:

RS2_OPTION_ENABLE_POSE_JUMPING 1.000000
RS2_OPTION_ENABLE_MAPPING 1.000000
RS2_OPTION_ENABLE_RELOCALIZATION 1.000000
RS2_OPTION_ENABLE_MAP_PRESERVATION 0.000000

[neilyoung]

The astonishing fact: The python binding is absolutely not affected by this bug (?) I can set_options and the script runs w/o issues. It is just the CPP thing...

EDIT: No, forget about this. The same bitchy behaviour :) I'm confused already, having seen too many fails with this...

[bfulkers-i]

Thanks for the report @neilyoung, I am able to reproduce this, we will look into it and provide a fix.

[neilyoung]

@bfulkers-i Great. Thanks for the follow up.

[neilyoung]

Fixed with #5394