ecl_run depends on Resdata to be able to make summary objects

[berland]

This line exposes the Eclipse100 job to bugs in resdata for MPI runs.

ert/src/ert/shared/share/ert/forward-models/res/script/ecl_run.py

Line 389 in dd3a01d

ecl_sum = Summary(case)

For the user, a crash in this code looks like Eclipse100 itself has crashed, while it hasn't.

[berland]

The crash can be reproduced with

summary.x XXX_PRED-0.UNSMRY

on the finished file produced in the asset. This implies a follow-up bug in resdata as this summary file is not readable by subsequent forward models.

[eivindjahren]

We looked at this together, and it seems like resdata crashes on the first report step. One unusual thing about the summary file is that the seqhdr is negative and ministeps don't start from 0.

[berland]

Underlying bug in resdata: equinor/resdata#945

[berland]

We looked at this together, and it seems like resdata crashes on the first report step. One unusual thing about the summary file is that the seqhdr is negative and ministeps don't start from 0.

Checked a smry file from Eclipse 2021, and it has the same properties with respect to seqhdr and ministeps but there is no crash.