There was a bug in ANGLE's glUniformMatrix implementation as
recently as version 963 that is fixed now. The bug was that a call
to UniformMatrix4f with a count > 1 for a variable declared
with, e.g., "uniform mat4 my_matrix;" would incorrectly cause
GL_INVALID_OPERATION to be raised. Excess data is supposed to be
ignored for arrays.
Is there now a conformance test for this? Since the problem has
been fixed, I'm guessing there is.