[IRC] Add a Link for the mandatory FixedConstraint

[fredroy]

First reason to do that was to avoid crashing shamelessly if one did not put a FixedConstraint in the same node...

[alxbilger]

I like the idea of the PR, but I don't like to create the fixed constraint automatically. Also, to note that the PR is breaking

[alxbilger]

what about initialization of the FixedProjectiveConstraint? Even if you call it, FixedProjectiveConstraint::init can also fail, and there is no automatic mechanism to prevent invalidity of the scene. So, I am in favor of triggering an error without backup.

[fredroy]

Problem is, if you dont create a fixedconstraint (i.e is null), as the IRC code is... let's say, not really solid 🤔, it will crash later on (not test on pointer here and there etc)
The creation here of FixedConstraint is not really intended to serve as a backup actually ; more as a way to not crash by itself 😅 Thats why there is a msg_error (and not a warning).
Or instead of crashing (if we dont do anything) we can do a fatal() or something like that... a graceful exit

And as for the breaking, I agree and I assume that I wont do deprecation and stuff 🙊

[hugtalbot]

I agree with both of you : @alxbilger for the analysis and @fredroy about the IRC component

Hard to decide which would be the better option.
I would vote for the simplest one (thinking about our future us debugging) : just an error, even if this leads to crashes (why not the fatal if it is slightly nicer)

[damienmarchal]

To me the component should be emit an error and be set in invalid state.
Then the component state should be checked if it is valid.

[fredroy]

To me the component should be emit an error and be set in invalid state. Then the component state should be checked if it is valid.

This is what I did few lines after (commented) but only a few functions in IRC does the componentState testing.
And I dont really want to update all the functions to add the test 🥸