BUG: Strange phenomenon with Constant's about firedrake HOT 4 CLOSED

This could be related to #3079. If you run firedrake-clean between running with the first and second cases do you still see the wrong result?

from firedrake.

This could be related to #3079. If you run firedrake-clean between running with the first and second cases do you still see the wrong result?

Unfortunately, I still get the same results even after firedrake-clean.

from firedrake.

Instantiating one more Constant means that the Constants in G have a different count from uh, otherwise the first Constant(1.0) has the same count as uh.

from firedrake.

I've identified the issue. With a mixed system like this we split the form and compile the pieces separately. This means that TSFC and assemble are seeing different forms so we pick out different Constants.

There is a hackathon/conference this week so depending on how it goes someone in the team will have either a lot of time to look at this or no time at all. If not fixed this week then I will prioritise it for next week.

from firedrake.