Strip out calls to nullify_format_instance

[ndevenish]

We had problems in the past with the behaviour of HDF5 under forking/parallel reads e.g. #35 #144 01463cc 2550f7c b19510e 36f63d6 and also #1539.

While testing a fix to centralise this check I stripped out the behaviour of nullify_format_instance and found that... I couldn't reproduce any of the problems. After some digging, it looks like in HDF5 1.10.5 this problem was fixed by:

Added a new option to enable/disable using pread/pwrite instead of
read/write in the sec2, log, and core VFDs.
This option is enabled by default when pread/pwrite are detected.

Which does reading/writing in a stateless way. So, all of these workaround hacks have become redundant.

Possibly.

All the tests I've run appear to pass - this doesn't remove the need for nullify_format_instance completely because there seem to be some vestigial problem with Lazy* formats cctbx/dxtbx#245. Does anyone have any memories/documented cases where this broke before to run tests through also? Otherwise, I'd suggest that we put it in and see if any problems come up related to it.

[codecov]

Codecov Report

Merging #1542 (f20ca62) into master (fbf1168) will increase coverage by 0.02%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master    #1542      +/-   ##
==========================================
+ Coverage   66.52%   66.55%   +0.02%     
==========================================
  Files         615      615              
  Lines       69152    69136      -16     
  Branches     9553     9546       -7     
==========================================
+ Hits        46005    46012       +7     
+ Misses      21211    21191      -20     
+ Partials     1936     1933       -3     

[phyy-nx]

Did you test it under MPI conditions? test_stills_process.py has an mpi test in it.

[phyy-nx]

conda install -c conda-forge mpi4py openmpi
mpirun hostname # test mpi is working, should see hostname for each available rank
libtbx.pytest -n 3 test_stills_process.py --regression # all three tests should run

I did the above on master and on the nullify_pid branch and the tests pass. I presume it's ok then.

[ndevenish]

I've merged this now, with the understanding that we should back it out quickly if it is causing any apparent problems.