fixes degeneracies when doing atan2(0,0)

[vrabaud]

This fixes the unit test in urdfdom at ros/urdfdom#49

[vrabaud]

BTW, this is a bad enough bug that a new release should be made, or at least, the patch should be pushed to Ubuntu.

[isucan]

+1 for Vincent's comment.
@esteve, @wjwwood Do you know who is in charge of the ubuntu release now?

[wjwwood]

@j-rivero do you know?

[j-rivero]

I am probably the responsible for submitting the updates. If we plan to submit a patch to an already release version (trusty, utopic and vivid) then we need to send a SRU (Stable Release Update) bug, something like we have done in the past for sdformat.

So I would need:

• A standalone test case that demonstrate the error with a different result between the unpatched version and the patched one.
• A short analysis about possible regressions introduced by this patch (check out [Regression Potential] in the sdformat example)
• (Optional) a new release containing this patch, so I fix the development version. I can do it by manually porting the patch, but is less elegant.

Thanks.

[vrabaud]

So:

• test is here: https://github.com/ros/urdfdom/blob/master/urdf_parser/test/urdf_unit_test.cpp
• well, none: math is simple, just a specific degeneracy use case is handled. No ABI/API change. No recompilation (pure header implementation).
• I can do a release if needed but more generally, we would need a more recent version for Wily or something.

[j-rivero]

Thanks very much Vincent. Here we go: https://bugs.launchpad.net/ubuntu/+source/urdfdom-headers/+bug/1471047.

A new released version to be uploaded to Willy would be very welcome.

[vrabaud]

yay, thx ! @isucan , can we have a new release for Willy ?