Joerg Henrichs

PR #1852 fixes the problem in this case, additionally #1855 also fixes the root cause of this crash if an array is used with and without indices.

This is ready for a first review, but ideally would be done after merging #1831 (which will likely cause some clashes). Note that the link checking will fail because the...

I believe the link check is caused by the stand-alone directory missing, which this PR will add. And I don't understand the coverage issue at all (I believe I might...

OK, I hope I've fixed all incorrect references to NetCDF. The link-check errors are expected since they point to paths that do not yet exist on master. The coverage issue...

I have started to implement the following option: ``` :param bool array_shape_accesses_are_read: if True (default), an access \ to an array using `size`, `lbound`, or `ubound` are considered read \...

In #1840 a patch is shown that removes a crash if a whole-array access is used. This seems to be a useful fix that should be integrated here as well.

Yes, I think that's good enough. More detailed options can easily be added if required, afaik we have all actual use cases covered.

Gee, that really surprised me, I thought we did run compile tests regularly. Having a quick look shows that `gocean/eg1` has no support for compilation: ``` make compile ... No...

I am working on this now. I will comment out the calls for which we have no source code (but leave them as comments in case someone feels like completing...

In #1751 a transformation is introduced that removes induction variables. Ideally this transformation should be executed before calling the dependency analysis, e.g. to remove: ``` do i=1, n im =...