- Dec 01, 2022
-
-
* Also unify handling of boolean long options to pio_write..
-
-
-
-
-
-
-
-
* This enables nicer failure modes for long file names and better diagnostics.
-
* This matches the array size type used in cdi.
-
* This reduces code duplication in differently decomposed test-programs and gives better change locality later on.
-
* Using CFLAGS for this was incorrect and resulted in unnecessarily complex and divergent Makefile recipes.
-
-
* This will later on allow the introduction of differently packed grid array data without affecting the code for the scalar part.
-
-
-
-
* This will permit multiple parallel modes concurrently.
-
-
* Having one storage variable for namespace id limited CDI-PIO to one namespace.
-
-
-
-
* This prepares for derived resource class packing, i.e. extends the set of operations that can be applied to in fact different objects.
-
-
-
-
* The changes in this commit delay calling of inq[XY](Vals|Bounds)Ptr until the absolute last time possible in the affected functions.
-
-
* This way the availability of mask, [xy]-{bounds|values} etc. can be queried without adding a method each time.
-
-
-
-
* This implies NetCDF data mode is only entered once, unless missing values are written later.
-
-
* Otherwise the compiler is not allowed to fold the storage of stdname, longname and units into one.
-
* Localize independent uses of variable timevalue in the respective blocks. * Merge calls to cdf_get_vara_double for time bounds. * Move invariant data to text segment.
-
-
-
-