Bug in ncmpii_sanity_check, version 1.8.1

[Sjaardema, Gregory D]

Not sure if you are still fixing bugs in 1.8.1 and I verified that it doesn’t exist in 1.9.0.pre1, so feel free to ignore…

In the sanity test starting at line 166, there is the following block:

   for (i=firstDim; i<(*varp)->ndims; i++) {
#ifdef RELAX_COORD_BOUND
        if (start[i] < 0 || start[i] > (*varp)->shape[i]) {
            DEBUG_ASSIGN_ERROR(err, NC_EINVALCOORDS)
            goto fn_exit;
        }
        if (start[i] == (*varp)->numrecs) {
            if (api == API_VAR1) {
                /* for var1 APIs, count[0] is considered of 1 */
                DEBUG_ASSIGN_ERROR(err, NC_EINVALCOORDS)
                goto fn_exit;
            }
            else if (count != NULL && count[i] > 0) {
                DEBUG_ASSIGN_ERROR(err, NC_EINVALCOORDS)
                goto fn_exit;
            }
        }
#else
        if (start[i] < 0 || start[i] >= (*varp)->shape[i]) {
            DEBUG_ASSIGN_ERROR(err, NC_EINVALCOORDS)
            goto fn_exit;
        }
#endif
    }

The “(*varp)->numrecs” in the if statement at line 172 should be “(*varp)->shape[i]”  I think this is a cut-paste bug where two instances were changed, but the third was missed.

..Greg
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/parallel-netcdf/attachments/20171130/cba85cbd/attachment.html>