I'm doing a longitudinal study and try to test for measurement invariance. For this, I took Little's (2013) approach and added the constraints simultaneously to both groups and measurement times.
Consequently, for strict factorial invariance I constrained the intercepts of all observed variables, but there I faced the problem, that I could not estimate the latent factor means. For group 1, the estimates for both time points are constrained to 0; I could only estimate the latent mean differences for group 2. However, this model seems too restrictive since it doesn't allow the mean at time 2 for group 1 to vary.