Modified provenance code to allocate the minimal space
needed for _NCProperties attribute in file. Basically
required using malloc in the provenance code and in ncdump.
Otherwise should cause no externally visible effects.
Also removed the ENABLE_FILEINFO from configure.ac since
the provenance code is no longer optional.
This modifies the previous change to be more pedantically correct. It should always be an NC_EINVALCOORDS error if start exceeds fdims[2]; however, if start equals fdims[2], then it is only an error if count is non-zero.
The following code is in nc4hdf.c, function `nc4_put_vara`.
```
/* Check dimension bounds. Remember that unlimited dimnsions can
* put data beyond their current length. */
for (d2 = 0; d2 < var->ndims; d2++)
{
dim = var->dim[d2];
assert(dim && dim->dimid == var->dimids[d2]);
if (!dim->unlimited)
{
if (start[d2] >= (hssize_t)fdims[d2])
BAIL_QUIET(NC_EINVALCOORDS);
if (start[d2] + count[d2] > fdims[d2])
BAIL_QUIET(NC_EEDGE);
}
}
```
There is an issue when the process with the highest rank has zero items to output. As an example, if I have 4 mpi processes which are each writing the following amount of data:
* rank 0: 0 items
* rank 1: 2548 items
* rank 2: 4352 items
* rank 3: 0 items.
I will define the variable to have a length of 6900 items (0 + 2548 + 4352 + 0). When I am outputting data to the variable, each rank will call nc_put_vara_longlong with the following start and count values:
* rank 0: start = 0, count = 0
* rank 1: start = 0, count = 2548
* rank 2: start = 2548, count = 4352
* rank 3: start = 6900, count = 0.
In each case, the `start` for rank N is equal to `start` for rank N-1 + `count` for rank N-1. This all works ok until the highest rank is writing 0 items. In that case, the `start` value for that rank is equal to the total size of the variable and the check in the code fragment shown above fails since `start[] == fdims[]`.
This could be fixed in the application code by checking whether the `count` is zero and if so, then set `start` to 0 also, but I think that is a kluge that should not be required.
Note that this test appears three times in this file. In one case, the check for non-zero count already exists, but not in the other two. This pull request adds the check to the other two tests.
The H5Aexists hdf5 function does the same function as the manually coded loop with much less code and fewer function calls.
Also, the H5Aopen_idx and H5Aget_num_attrs functions are deprecated.
If H5Aopen_idx on line 1964 fails, then attid will be < 0. The BAIL will goto exit at line 1989 and then the test of "if (attid ...)" at line 1995 will pass (attd != 0) and then call H5Aclose(attid) with a negative attid. Similar issue for spaceid.
Result of function if probably the same since there is a failure somewhere, but more difficult to track down if looks like failure is happening in the wrong place.
not being computed. Fix in nc4file.c.
Not sure how this ever worked for any variable.
What is also weird is that the dim hash is
apparently being computed.
re: github netcdf-c issue #271
This occurs for several reasons, including:
1. using H5Aopen_name instead of H5Aexists to test if attribute exists.
2. using H5Eset_auto instead of H5Eset_auto2.
There are probably others that will have to be extinguished as encountered.
p.s Hope I did not overdo this and kill too much.
The hash field for phony dimensions was not being set
(in nc4hdf.c). Also added test case (nc_test4/?).
Note that I searched for other similar failures and
did not find any, but I may have missed them.
If in parallel mode and the H5Fcreate fails, then the h5->comm field may/will point to MPI_COMM_NULL instead of to a valid communicator. If that is passed to MPI_Comm_free, then the code will core dump down in the MPI_Comm_free function and not return a valid return status to the caller routine. If communicator is checked, then execution proceeds back up the call tree and caller can report a better error message about the failed file create.
Charlie Zender noted that we forgot to define what happens for
various netcdf API attribute operations, notably nc_inq_att()
and nc_get_att().
So, I added a list of legal and illegal api calls for the provenance
attributes in docs/attribute_conventions.md.
I also added more test cases to ncdump/tst_fileinfo.c to verify
and fixed resultant errors.
This consists of a persistent attribute named
_NCProperties plus two computed attributes
_IsNetcdf4 and _SuperblockVersion.
See the 'Provenance Attributes' section
of docs/attribute_conventions.md for details.
The var struct has a 'dim' field which was not being used
Instead, the dimids field would always search for the dim
with the matching dimid. For db with large numbers of dims,
this could be a significant time sync.
Modified code to always set var-dim[i] when var->dimids[i] was
set (if the dim existed at that point). Then use the var->dim
field instead of var->dimids and search whenever requested.
All var->dim accesses are protected by asserts that verify
non-null and that the var->dim[]->dimid == var->dimids[].
In non-classic netcdf-4 models, it is allowable to have
large numbers of dims and vars. In many operations, the
entire list of dims or vars is searched for a dim/var matching
a specific name which results in *lots* of strncmp or strcmp
calls.
If we add a hash field to the var and dim structs similar to what
has already been done for the netcdf-3 formats, then we can hash the
name being searched for and numerically compare that value with
the var/dim hash value. If they match, then do a more expensive
strncmp call to ensure that the names truly match.
