@RussTreadon-NOAA is it easy to pull this into your test case and re-run? If not, I can test it

Thank you, @CoryMartin-NOAA , for the quick fix. I'll pull this into my Hercules working copy, recompile gsi.x, and rerun the failed gdas_anal.

@RussTreadon-NOAA my new friend first suggested action='read' but then in VSCode suggested status='old', I thought status='old' was better, but if you prefer action='read' we can do that instead

Hercules test

Build gsi.x from bugfix/blacklist. Rerun 20211221 00Z gdas_anal from g-w CI case C96C48_hybatmDA. gsi.x ran to completion without error. However, EIB's MSU staged GSI fix directory had already been populated with the correct rejectlist_global.txt.

As a test add export BLACKLIST=/path/to/non/existent/file to config.anal and rerun the 20211221 00Z gdas_anal. This time gsi.x aborted because file reject_global.txt did not exist

76: forrtl: severe (29): file not found, unit 19, file /work2/noaa/stmp/role-da/HERCULES/RUNDIRS/C96C48_hybatmDA/gdas.2021122100/anal.3616677/blacklist
76: Image              PC                Routine            Line        Source
76: gsi.x              000000000221D935  Unknown               Unknown  Unknown
76: gsi.x              0000000000D50224  blacklist_mp_blac         105  blacklist.f90
76: gsi.x              0000000000961D73  read_prepbufr_            610  read_prepbufr.f90
76: gsi.x              000000000094EB5C  read_obsmod_mp_re        1474  read_obs.F90
76: gsi.x              0000000000861BB2  observermod_mp_se         329  observer.F90
76: gsi.x              000000000105B4F3  glbsoi_                   221  glbsoi.f90
76: gsi.x              000000000068D267  gsisub_                   200  gsisub.F90
76: gsi.x              000000000041519D  gsimod_mp_gsimain        2451  gsimod.F90
76: gsi.x              00000000004150DF  MAIN__                    633  gsimain.f90

This is the correct behavior because the file is opened with status='old'. The old status means the file exists. gsi.x did not create a zero length rejectlist_global.txt.

The failure without creating a size 0 file is the desired behavior.

@RussTreadon-NOAA are we ready to merge this? or do we need to do other tests (the regression tests)?

@CoryMartin-NOAA , @ShunLiu-NOAA , @hu5970 : this is a quick PR to address an issue discovered via g-w CI testing.

The g-w team has suspended all CI testing until the g-w GSI hash has been updated with this change.

If you are OK with this change, I recommend that we merge ASAP.

Merging now