Inconsistency in filter-related info in nc-config vs. netcdf_meta.h

[czender]

Today's main snapshot includes some new info in nc-config which is good. However, it seems that nc-config and netcdf_meta.h are not consistent in what/how they report about available filters. Namely, netcdf_meta.h lacks any tokens that declare the availability of the Bzip2 and Blosc filters. netcdf_meta.h does contain a useful (meaning client applications like NCO can easily make use of it in C-code) Zstandard token:

#define NC_HAS_PAR_FILTERS   1 /* Parallel I/O with filter support. */
#define NC_HAS_NCZARR        1 /*!< Parallel I/O with filter support. */
#define NC_HAS_MULTIFILTERS  1 /*!< Nczarr support. */
#define NC_HAS_LOGGING       1 /*!< Logging support. */
#define NC_HAS_QUANTIZE      1 /*!< Quantization support. */
#define NC_HAS_ZSTD          1 /*!< Zstd support. */

(Also note the comment for NC_HAS_NCZARR is, erroneously, a copy of the NC_HAS_PAR_FILTERS comment). Whereas nc-config reports

--has-nczarr       -> yes
 --has-zstd         -> yes
 --has-benchmarks   -> no
 --has-multifilters -> yes
 --has-stdfilters   -> deflate bz2 blosc zstd
 --has-quantize     -> yes

So Zstandard is reported by itself (without additional parsing) by both nc-config and netcdf_meta.h. The presence of Bzip2 and Blosc, on the other hand, can only be learned by parsing the list output of --has-stdfilters from nc-config. If the new method of specifying available standard filters was intentionally to list them in nc-config but not say anything in netcdf_meta.h then, well, fine. At least it is intentional. But it the absence of Bzip2 and Blosc from netcdf_meta.h is unintentional then please fix it so client applications can make use of NC_HAS_<filtername> tokens for all standard filters.

[WardF]

Good catch; I'll get this oversight corrected!