Check for SDF files containing sibling elements with the same name (Forward port of #3016)

[j-rivero]

Forward port of #3016.

Gazebo11 supports SDF 1.7 which is different of #3016 since it only support SDF 1.6. I had to comment and reverse the expectations in one of the tests since it is affected by gazebosim/sdformat#586 from libsdformat9.

[j-rivero]

CI has some problems:

• abichecker is complaining about rendering - LensFlare.hh, unrelated to this PR but needs checking
• Windows have problems with workspace files corruption

[j-rivero]

abichecker is complaining about rendering - LensFlare.hh, unrelated to this PR but needs checking

I've looked into it. #3007 seems the one triggering the problem but that PR is adding the symbol and not removing it, so it should be a bug in our abi checker code or in the underlying abi-compliance-checker tool. Not affecting this PR.

[jennuine]

This looks good to me but isn't Windows complaining from unsetenv not being defined?

C:\Jenkins\workspace\gazebo-ci-pr_any-windows7-amd64\ws\gazebo\test\integration\sdf_errors.cc(117): error C3861: 'unsetenv': identifier not found

Also, can the new environment variable GAZEBO11_BACKWARDS_COMPAT_WARNINGS_ERRORS be documented somewhere?

[j-rivero]

This looks good to me but isn't Windows complaining from unsetenv not being defined?

Ouch, right. Should be fixed in 4db4697

Also, can the new environment variable GAZEBO11_BACKWARDS_COMPAT_WARNINGS_ERRORS be documented somewhere?

Forget to port the change in Changelog, ready in 0987978

Gazebo11 supports SDF 1.7 which is different of #3016 since it only support SDF 1.6. I had to comment and reverse the expectations in one of the tests since it is affected by osrf/sdformat#586 from libsdformat9.

I've implemented the logic to handle this case in 0219c47 since I'm not sure if the method description in sdformat fit well about checking SDF versions requirements. We can follow the discussion in the sdformat issue but this implementation should not hurt in anyway I think.

[jennuine]

LGTM. CI failures don't appear to be related to this PR but could use a second pair of eyes

[j-rivero]

LGTM. CI failures don't appear to be related to this PR but could use a second pair of eyes

Looking into them:

The only ones I've detected that are somehow 'new' (not in the daily build or not in other builds) are:

• INTEGRATION_transport.test_ran unrelated to SDF management
• UNIT_SonarSensor_TEST.test_ran: segfaulting, there is a warning that applying negative mass can end up in a segfault :(

[j-rivero]

Brew CI seems generally broken right now. Going to merge to have the release out since previous builds were fine as far as I can tell.