*xtra variables not saved in photos?

[mathren]

Describe the bug

According to $MESA_DIR/star_data/public/star_data_step_work.inc (lines ~947 onwards in r24.03.1) it should be possible to use ixtra, lxtra, and xtra arrays for applications in the run_star_extras.f90 which have a state which should work across restarts and retries. Specifically there is a comment saying:

         ! more extras -- for run_star_extras applications with state.
         ! if your extra state info is all held in the following variables,
         ! restarts and retries will work without doing anything special.

I've been trying to use these to save checkpoints (mod file, profile, and photo) if and only if a certain condition (e.g., O depletion) has been reached, and flip lxtra(1) to .true. when I save it. The idea is that once lxtra(1) .eqv. .true. , MESA will not over-write these checkpoints anymore. I've added in my extras_startup lines like the following:

         if (restart .eqv. .false.) then
            print *, "NOT RESTARTING: SET TO FALSE"
            s% lxtra(1)            = .false.
         else
            print *, "RESTART: READ s%lxtra_old(1)"
            s% lxtra(1)            =  s% lxtra_old(1)
         end if
         print *, s%lxtra(1)

However, this does not seem to work as expected: regardless of from where I restart (from a photo) lxtra(1) is always initialize to .false. (even if I expected lxtra_old(1) to be .true. after the checkpoint), and therefore if I restart from after the desired checkpoint, the first timestep will always overwrite my checkpoint.

I suppose this is because the information is not saved in the photos. This is maybe a niche need, but at the very least I find the comments above confusing.

To Reproduce

The attached zip file contains a simple minimal work example where lxtra(1) is flipped at model number 5 and that saves a photo for each timestep. I would expect that restarting from photos x001, x002, x003, x004 lxtra(1) would be set .false. in extras_startup, while restarting from photos x005 onwards it would be set to .true. Instead it appears to always be set to .false.

MWE.zip

I could find another way to "protect" saved checkpoints, e.g. checking if the file exists -- but ideally I would like to have MESA overwrite or not the checkpoints saved depending on whether I restart from before or after said checkpoint.

[fxt44]

I suppose this is because the information is not saved in the photos.

in r24.03.1 the arrays ixtra, lxtra, and xtra are written to a photo file at lines 144-146 in $MESA_DIR/star/private/photo_out.f90. they are read from a photo file at lines 240-245 in $MESA_DIR/star/private/photo_in.f90.

a question to answer then is if these arrays are being re-initialized after a photo file is read.

[evbauer]

I think the _old variable is misleading here. These variables have a different purpose of saving the starting state of the model in case we need to fall back to that in the course of evolution.

The lxtra variable itself should have a value that persists through restarts, so you shouldn't actually need to do any assignment upon restart at all. Does that resolve your issue?

[mathren]

I have fallen in this trap indeed and over-complicated things! It seems that removing completely that if from extras_start_step works as expected.

So old refers to at the beginning of the step, not to the previous timestep? This is a bit confusing, maybe a renaming of the variables to *_init_step or something like that would have prevented me from confusing myself -- but totally my bad.
Another option would be to rephrase the comments/documentation to make this clearer. Happy to do a PR for that if you think this would be useful to others as well.