Initialize variables in servers/physics

[qarmin]

Part of #43636

[pouleyKetchoupp]

Thank you! I've double checked initialized values and left some comments about missing ones.

[pouleyKetchoupp]

Empty constructors could be removed.

[qarmin]

I was only able to move empty constructors to .h files, because probably due virtual destructors, build fails when completely removing default constructor.

[akien-mga]

This might be worth reviewing @pouleyKetchoupp @lawnjelly as the previous code seems a bit weird. We set callbacks in the bvh manager before setting only the pair_callback to nullptr and leaving unpair_callback uninitialized.

[pouleyKetchoupp]

Actually this is fine but the code is hard to read :)

It's just that there's _pair_callback (with underscore) passed to the bvh and pair_callback (without) which is initialized to nullptr.

unpair_callback was uninitialized though, so it's all good now.

[akien-mga]

Seems to have a fail condition if it's <= 0.0, maybe the default should be 0.0 as uninitialized state?

servers/physics_3d/joints/jacobian_entry_3d_sw.h
73:             m_Adiag = massInvA + m_0MinvJt.dot(m_aJ) + massInvB + m_1MinvJt.dot(m_bJ);
75:             ERR_FAIL_COND(m_Adiag <= real_t(0.0));
89:             m_Adiag = m_0MinvJt.dot(m_aJ) + m_1MinvJt.dot(m_bJ);
91:             ERR_FAIL_COND(m_Adiag <= real_t(0.0));
104:            m_Adiag = m_0MinvJt.dot(m_aJ) + m_1MinvJt.dot(m_bJ);
106:            ERR_FAIL_COND(m_Adiag <= real_t(0.0));
121:            m_Adiag = massInvA + m_0MinvJt.dot(m_aJ);
123:            ERR_FAIL_COND(m_Adiag <= real_t(0.0));
126:    real_t getDiagonal() const { return m_Adiag; }
166:    real_t m_Adiag;

CC @pouleyKetchoupp

[pouleyKetchoupp]

Yeah it would be good to check uninitialized cases, but we would also need a check in getDiagonal() to catch these cases (debug only to avoid potential performance hits in release), something like:

real_t getDiagonal() const
{
#ifdef DEBUG_ENABLED
    ERR_FAIL_COND_V(m_Adiag <= real_t(0.0), m_Adiag);
#endif
    return m_Adiag;
}

[akien-mga]

Might be worth doing in a separate PR not to grow the scope of this one too much I guess.

[akien-mga]

Reviewed everything, looks great aside from the few comments I left. Should be mergeable once resolved.

[akien-mga]

Looks great!