Problem in restart after merging the parallel code

[giovannipizzi]

From Arash:
With the benzene example, if I run a wannierisation, and then do a second run setting "restart = wannierise” (or indeed plot) I get an error

forrtl: severe (24): end-of-file during read, unit 11, file /export111/work/aam/w90-examples/example12-val/benzene.chk

Or:
example01, example02 etc: instead of the end-of-file error on reading .chk files, I sometimes get the following error on the first cycle of the restart:

Cycle:      1
 wann_main: ZHEEV in internal_new_u_and_m failed, info=            3
            trying Schur decomposition instead
 wann_main: SCHUR failed, info=            4
 Exiting.......
 wann_main: problem computing schur form 1

[jryates]

I think this is one problematic part of the code. In the comms_gatherv there is a reference to m_matrix which should be m_matrix_1b. This fixes the lead restart on my machine.
However, this is nothing to do with the gamma-point code, so the original issue must be something different. I can't reproduce any problems with the benzene example (various compilers, valgrind etc).

do nn = 1, nntot
m_matrix_1b_loc=m_matrix_loc(:,:,nn,:)
call comms_gatherv(m_matrix_1b_loc,num_wannnum_wanncounts(my_node_id),&
m_matrix,num_wannnum_wanncounts,num_wannnum_wanndispls)
call comms_bcast(m_matrix_1b(1,1,1),num_wannnum_wannnum_kpts)
m_matrix(:,:,nn,:)=m_matrix_1b(:,:,:)
end do!nn

[giovannipizzi]

Not fixed yet... (see https://travis-ci.org/wannier-developers/wannier90/jobs/311536683)

[jryates]

I think the test is wrong. The problem was not with reading the chk file, it was with writing it. So the test has a wrong chk file.

[giovannipizzi]

Interesting, thanks! I'll see if I can fix this with a new chk file.

[giovannipizzi]

It was even more simple than that... there was no reference file! I committed and created a PR so we have also this test case in, feel free to merge if it's OK. Thanks!