driver: fix deserialisation of JointFeedbackEx

[gavanderhoorn]

See comment on c9c62dd for more info.

Summary: always deserialise every submsg, but only keep those that contain valid information.

I've verified this on our SIA20F, which has two motion groups: robot itself + a linear rail.

Instead of 'upgrading' to unloadFront(..) (as suggested by @andreaskoepf), I decided to make this a minimal change and kept unload(..).

[gavanderhoorn]

This is a fix for #91.

Partially based on the work by @jettan and @alemme (via @andreaskoepf).

[gavanderhoorn]

@shaun-edwards: would you have time for a review? If not, I'll ask someone else.

[gavanderhoorn]

@alemme, @andreaskoepf: would you perhaps see a chance to test this on your SDA? You would only need to change this file, right?

[alemme]

Hi everyone. I successfully tested your changes on our sda10d system.

[gavanderhoorn]

Thanks @alemme. Appreciated.

[gavanderhoorn]

@sinaiaranda-CIDESI: the assertion you're running into is most likely the same as in #91. This PR should fix that one.

There are some more issues with your setup and motoman_driver in kinetic-devel though, so things won't directly work after this is merged, but it's a step.

I don't completely understand how, when using #179 things don't work for you though. But please open a separate issue and we can continue there.

[gavanderhoorn]

Thanks @shaun-edwards for the review. And thanks @alemme as well.

[SinaiAranda]

@gavanderhoorn The problem has been resolved with this modification of PR.

[gavanderhoorn]

I'm slightly confused: #91 is the issue that reported the problem, not a PR. And the commit you link should be the "hacky" way this was fixed, with the PR we're discussing here being a more appropriate way.

If I may suggest: use the fix in this PR or just use current kinetic-devel, which contains the fix.

Fixes for #226 should fix the remaining issues and make things operational again (note: those haven't been merged yet).

[SinaiAranda]

@gavanderhoorn Excellent, have do not problem.