[v4.0.x] Using larger timestep with own tracer model (`nn_dttrc > 1`) leads to `lib_mpp` deadlock

I’m using NEMO with a custom MY_TRC model.
If I try setting nn_dttrc to something larger than 1, the model halts in lib_mpp, in the routine mpp_report: the routine is not coded for the case ncom_dttrc /= 1
Is the problem only related to the creation of the report ? In that case, I would desactivate the report, I can live without it. But I fear there’s more to it than that ?
Thank you,

I removed everything inside the mpp_report routine.
Nemo now crashes in trc_sub, when it attemps to fetch from allocatable variable WSLPI when it is not allocated. I used ln_traldf_iso=true (iso-neutral diffusion).
Now I use ln_traldf_hor=true, the same error occurs, now for the variable HMLD.

HMLD is allocated during the first time-step, in subroutine zdf_mxl, called by zdf_phy, itself called by step.
However trcsub is called by trc_init, itself called by nemo_init, before the first time step.

I wonder if I’m doing something wrong, or if this particular combination of choices has not been foreseen?

By commenting the lines related to HMLD in trcsub, I got Nemo to run multiple timesteps.
Contrary to what’s indicated, I don’t think TOP necessarily uses HMLD ; in my MY_TRC model, I don’t use it. I hope it’s not used elsewhere in the TRC routines

In general, I have the impression that the ability of using nn_dttrc > 1 has not been very much tested with respect to different choices for the physics in Nemo 4.0 (e.g. which direction of diffusion, see above)