spgeom.repeat on particular tshs gives wrong results / crashes, at least when used twice

I ran into a bug with an electrode that I wanted to .repeat instead of .tile. It resulted in a crash.

When doing Helec.repeat(3, 1).repeat(3, 0), the first repeat will always work well (as in, not crash), but the second one will usually, but not always, crash:

In [1]: import sisl as si

In [2]: H=si.get_sile("elec.TSHS").read_hamiltonian()

In [3]: H.repeat(3, 0).repeat(3, 1)
---------------------------------------------------------------------------
IndexError                                Traceback (most recent call last)
<ipython-input-3-645fb0da8346> in <module>()
----> 1 H.repeat(3, 0).repeat(3, 1)

/home/jolub/git/sisl/sisl/sparse_geometry.py in repeat(self, reps, axis)
   1922 
   1923         # Now we should fill the data
-> 1924         isc = geom.o2isc(col)
   1925         # resulting orbital in the new geometry (without wrapping
   1926         # for correct supercell, that will happen below)

/home/jolub/git/sisl/sisl/geometry.py in o2isc(self, io)
   3089         """
   3090         idx = _a.asarrayi(io) // self.no
-> 3091         return self.sc.sc_off[idx, :]
   3092 
   3093     def o2sc(self, o):

IndexError: index -7976227 is out of bounds for axis 0 with size 225

The index which is said to be out of bounds (and clearly is) is always a new number. I guess it doesn’t crash if this number happens to be in bounds. The tshs can be gotten here (time- and download limited link).

Tested on current master.

[zerothi]

I think I have narrowed down the problem, now I just need to solve it 😉

[zerothi]

@jonaslb Great!