This file contains a high-level description of this package's evolution. Releases are in reverse chronological order (most recent first). Note that, as of netcdf 4.2, the `netcdf-c++` and `netcdf-fortran` libraries have been separated into their own libraries.
* [Bug Fix] Corrected an issue when subsetting a netcdf3 file via `nccopy -v/-V`. See [Github #425](https://github.com/Unidata/netcdf-c/issues/425) and [Github #463](https://github.com/Unidata/netcdf-c/issues/463) for more information.
* [Bug Fix] Corrected `--has-dap` and `--has-dap4` output for cmake-based builds. See [GitHub #473](https://github.com/Unidata/netcdf-c/pull/473) for more information.
* [Bug Fix] Corrected an issue where `NC_64BIT_DATA` files were being read incorrectly by ncdump, despite the data having been written correctly. See [GitHub #457](https://github.com/Unidata/netcdf-c/issues/457) for more information.
* [Bug Fix] Addressed an issue with how cmake was implementing large file support on 32-bit systems. See [GitHub #385](https://github.com/Unidata/netcdf-c/issues/385) for more information.
* [Bug Fix] Addressed an issue where ncgen would not respect keyword case. See [GitHub #310](https://github.com/Unidata/netcdf-c/issues/310) for more information.
* [Enhancement] The remote testing server can now be specified with the `--with-testserver` option to ./configure.
* [Enhancement] Modified netCDF4 to use ASCII for NC_CHAR. See [Github Pull request #316](https://github.com/Unidata/netcdf-c/pull/316) for more information.
* [Bug Fix] Corrected an error with how dimsizes might be read. See [Github #410](https://github.com/unidata/netcdf-c/issues/410) for more information.
* [Bug Fix] Corrected an issue where 'make check' would fail if 'make' or 'make all' had not run first. See [Github #339](https://github.com/Unidata/netcdf-c/issues/339) for more information.
* [Bug Fix] Corrected an issue on Windows with Large file tests. See [Github #385](https://github.com/Unidata/netcdf-c/issues/385]) for more information.
* [Bug Fix] Corrected an issue with diskless file access, see [Pull Request #400](https://github.com/Unidata/netcdf-c/issues/400) and [Pull Request #403](https://github.com/Unidata/netcdf-c/issues/403) for more information.
* [Refactor] the oc2 library is no longer independent of the main netcdf-c library. For example, it now uses ncuri, nclist, and ncbytes instead of its homegrown equivalents.
* [Bug Fix] `NC_EGLOBAL` is now properly returned when attempting to set a global `_FillValue` attribute. See [GitHub #388](https://github.com/Unidata/netcdf-c/issues/388) and [GitHub #389](https://github.com/Unidata/netcdf-c/issues/389) for more information.
* [Bug Fix] Corrected an issue where data loss would occur when `_FillValue` was mistakenly allowed to be redefined. See [Github #390](https://github.com/Unidata/netcdf-c/issues/390), [GitHub #387](https://github.com/Unidata/netcdf-c/pull/387) for more information.
* [Upgrade][Bug] Corrected an issue regarding how "orphaned" DAS attributes were handled. See [GitHub #376](https://github.com/Unidata/netcdf-c/pull/376) for more information.
* [Upgrade] Update utf8proc.[ch] to use the version now maintained by the Julia Language project (https://github.com/JuliaLang/utf8proc/blob/master/LICENSE.md).
* [Bug] Addressed conversion problem with Windows sscanf. This primarily affected some OPeNDAP URLs on Windows. See [GitHub #365](https://github.com/Unidata/netcdf-c/issues/365) and [GitHub #366](https://github.com/Unidata/netcdf-c/issues/366) for more information.
* [Enhancement] Added support for HDF5 collective metadata operations when available. Patch submitted by Greg Sjaardema, see [Pull request #335](https://github.com/Unidata/netcdf-c/pull/335) for more information.
* [Bug] Addressed an issue where netCDF wouldn't build on Windows systems using MSVC 2012. See [GitHub #304](https://github.com/Unidata/netcdf-c/issues/304) for more information.
* [Enhancement] Incorporated an enhancement provided by Greg Sjaardema, which may improve read/write times for some complex files. Basically, linked lists were replaced in some locations where it was safe to use an array/table. See [Pull request #328](https://github.com/Unidata/netcdf-c/pull/328) for more information.
* [Bug] Fixed an issue where `ncgen` would potentially crash or write incorrect netCDF4 binary data under very specific circumstances. This bug did *not* affect data written on 32-bit systems or by using the netCDF library; it was specific to `ncgen`. This would only happen when writing a compound data type containing an 8-byte data type followed by a 4-byte data type *and* the 4-byte data type was not properly aligned; this would *possibly* result in incorrect padding. This did not affect 32-bit systems, or data written directly by the library. See [GitHub #323](https://github.com/Unidata/netcdf-c/issues/323) for more information.
* [Documentation] Updated documentation related to netCDF variable names and DAP2 access to reflect the undefined behavior potentially observed when DAP2 reserved keywords are used as netCDF variable names. See [GitHub #308](https://github.com/Unidata/netcdf-c/issues/308) for more information.
* [Bug] Fixed an issue with `nc_inq_type()` not returning proper value in some circumstances. See [GitHub #317](https://github.com/Unidata/netcdf-c/issues/317) for more information.
* [Bug] Corrected an issue related to test failures when `--disable-utilities` or `-DENABLE_UTILITIES=OFF` are specified when building with autotools or cmake, respectively. See [GitHub #313](https://github.com/Unidata/netcdf-c/issues/313) for more information.
* [Bug][Enhancement] Corrected a behavioral issue with the `_NCProperties` attribute taking up too much space. See [GitHub #300](https://github.com/Unidata/netcdf-c/issues/300) and [GitHub #301](https://github.com/Unidata/netcdf-c/pull/301) for more information.
* [Bug] Corrected behavior for `nc-config` so that, if `nf-config` is found in system, the proper fortran-related information will be conveyed. See [GitHub #296](https://github.com/Unidata/netcdf-c/issues/296] for more information.
* [File Change] Starting with release 4.4.1, netCDF-4 files created will have superblock version 0 instead of superblock version 2, as was observed in previous netCDF versions. This is due to a workaround required to avoid backwards binary incompatibility when using libhdf5 1.10.x or greater. Superblock versions 0 and 2 appear to be forward and backward compatible. Other than a different superblock number the data should remain consistent.
* [Enhancement] Added better error reporting when ncdump/nccopy are given a bad constraint in a DAP url. See [GitHub #279](https://github.com/Unidata/netcdf-c/pull/279) for more information.
* [Bug Fix] Corrected an issue where adding a \_FillValue attribute to a variable would result in other attributes being lost. See [GitHub #239](https://github.com/Unidata/netcdf-c/issues/239) for more details.
* [Enhancement] Added provenance information to files created. This information consists of a persistent attribute named `_NCProperties` plus two computed attributes, `_IsNetcdf4` and `_SuperblockVersion`. Associated documentation was added to the file `docs/attribute_conventions.md`. See [GitHub pull request #260](https://github.com/Unidata/netcdf-c/pull/260) for more information.
* [Bug Fix] Cleaned up some dead links in the doxygen-generated documentation.
* [Bug Fix] Corrected several issues related to building under Visual Studio 2014.
* [Bug Fix] Corrected several test failures related to HDF5 `1.10.0`
* [Bug Fix] Reverted SOVersion *current* to 11 from 12; it was incorrectly incremented in netCDF-C release 4.4.1-RC1.
* [Bug Fix][Enhancement] Fixed an issue with netCDF4 files generated using version `1.10.0` of the HDF5 library. The 1.10 release potentially changed the underlying file format, introducing a backwards compatibility issue with the files generated. HDF5 provided an API for retaining the 1.8.x file format, which is now on by default. See [GitHub Issue #250](https://github.com/Unidata/netcdf-c/issues/250) for more information.
* [Bug Fix] Corrected an issue with autotools-based builds performed out-of-source-tree. See [GitHub Issue #242](https://github.com/Unidata/netcdf-c/issues/242) for more information.
* [Enhancement] Modified `nc_inq_type()` so that it would work more broadly without requiring a valid ncid. See [GitHub Issue #240](https://github.com/Unidata/netcdf-c/issues/240) for more information.
* [Enhancement] Accepted a patch code which added a hashmap lookup for rapid var and dim retrieval in nc3 files, contributed by Greg Sjaardema. See [GitHub Pull Request #238](https://github.com/Unidata/netcdf-c/pull/238) for more information.
* [Bug Fix] Accepted a contributed pull request which corrected an issue with how the cmake-generated `nc-config` file determined the location of installed files. See [GitHub Pull Request #235](https://github.com/Unidata/netcdf-c/pull/235) for more information.
* [Enhancement] Added an advanced option for CMake-based builds, `ENABLE_SHARED_LIBRARY_VERSION`. This option is `ON` by default, but if turned off, only `libnetcdf.dylib` will be generated, instead of files containing the SOVERSION in the file name. This is a requested feature most people might not care about. See [GitHub #228](https://github.com/Unidata/netcdf-c/issues/228) for more information.
* [Bug Fix] Corrected an issue with duplicated error codes defined in multiple header files. See [GitHub #213](https://github.com/Unidata/netcdf-c/issues/213) for more information.
* [Bug Fix] Addressed an issue specific to Visual Studio 2015 on Windows. On very large files, some calls to the `fstat` class of functions would fail for no apparent reason. This behavior was **not** observed under Visual Studio 2013. This has now been mitigated. See [GitHub #188](https://github.com/Unidata/netcdf-c/issues/188) for more information.
* [Enhancement] Updated `nc-config` to report whether `logging` is enabled in netcdf. Additionally, if `f03` is available in an installed netcdf-fortran library, it will now be reported as well.
* [Bug Fix] Addressed an issue where `netcdf_mem.h` was not being installed by cmake. See [GitHub #227](https://github.com/Unidata/netcdf-c/issues/227) for more information.
* [Bug Fix] Addressed an issue where `ncdump` would crash when trying to read a netcdf file containing an empty ragged `VLEN` variable in an unlimited dimension. See [GitHub #221](https://github.com/Unidata/netcdf-c/issues/221) for more information.
* Modified `CMakeLists.txt` to work with the re-organized cmake configuration used by the latest HDF5, `1.8.16`, on Windows. Before this fix, netCDF would fail to locate hdf5 1.8.16 when using cmake on Windows. See [GitHub #186](https://github.com/Unidata/netcdf-c/issues/186) for more information.
* Addressed an issue with `ncdump` when annotations were used. The indices for the last row suffered from an off-by-1 error. See [GitHub issue #181](https://github.com/Unidata/netcdf-c/issues/181) for more information.
* Addressed an issue on platforms where `char` is `unsigned` by default (such as `ARM`), as well as an issue describing regarding undefined behavior, again on `ARM`. See [GitHub issue #159](https://github.com/Unidata/netcdf-c/issues/159) for detailed information.
* Updated documentation for `nc_get_att_string()` to reflect the fact that it returns allocated memory which must be explicitly free'd using `nc_free_string()`. Reported by Constantine Khroulev, see [GitHub Issue 171](https://github.com/Unidata/netcdf-c/issues/171) for more information.
* Updated documentation to reference the new `NodeJS` interface to netcdf4, by Sven Willner. It is available from [https://www.npmjs.com/package/netcdf4](https://www.npmjs.com/package/netcdf4) or from the GitHub repository at [https://github.com/swillner/netcdf4-js](https://github.com/swillner/netcdf4-js).
* Incorporated pull request https://github.com/Unidata/netcdf-c/pull/150 from Greg Sjaardema to remove the internal hard-wired use of `NC_MAX_DIMS`, instead using a dynamic memory allocation.
* Added a fix for https://github.com/Unidata/netcdf-c/issues/149, which was reported several times in quick succession within an hour of the RC4 release.
The various extended format flags of the format `NC_FORMAT_FOO` have been refactored into the form `NC_FORMATX_FOO`. The old flags still exist but have been marked as deprecated and will be removed at some point. This was done to avoid confusion between the extended format flags and the format flags `NC_FORMAT_CLASSIC`, `NC_FORMAT_64BIT_OFFSET`, etc. The mapping of deprecated-to-new flags is as follows:
* Reduced minimum cmake version to `2.8.11` from `2.8.12`. This will allow for cmake use on a broader set of popular linux platforms without having to do a custom cmake install. See https://github.com/Unidata/netcdf-c/issues/135 for more information.
* The documentation section `The Default Chunking Scheme` has been updated with more information. This lives in the `guide.dox` file in the `docs/` directory, or can be found online in the appropriate location (typically http://www.unidata.ucar.edu/netcdf/docs/), once this release has been published.
* Addressed an inefficiency in how bytes would be swapped when converting between `LITTLE` and `BIG` ENDIANNESS. See [NCF-338](https://bugtracking.unidata.ucar.edu/browse/NCF-338) for more information.
* Addressed an issue where an interrupted read on a `POSIX` system would return an error even if errno had been properly set to `EINTR`. This issue was initially reported by David Knaak at Cray. More information may be found at [NCF-337](https://bugtracking.unidata.ucar.edu/browse/NCF-337).
* The pre-built Windows binaries are now built using `Visual Studio 2012`, instead of `Visual Studio 2010`. Source-code compilation remains function with `Visual Studio 2010`, this is just a change in the pre-built binaries.
* Added support for opening in-memory file content. See `include/netcdf_mem.h` for the procedure signature. Basically, it allows one to fill a chunk of memory with the equivalent of some netCDF file and then open it and read from it as if it were any other file. See [NCF-328](https://bugtracking.unidata.ucar.edu/browse/NCF-328) for more information.
* Addressed an issue when reading hdf4 files with explicit little-endian datatypes. This issue was [reported by Tim Burgess at GitHub](https://github.com/Unidata/netcdf-c/issues/113). See [NCF-332](https://bugtracking.unidata.ucar.edu/browse/NCF-332) for more information.
* Addressed an issue with IBM's `XL C` compiler on AIX and how it handled some calls to malloc. Also, as suggested by Wolfgang Hayek, developers using this compiler may need to pass `CPPFLAGS=-D_LINUX_SOURCE_COMPAT` to avoid some test failures.
* Addressed an issure in netcdf4 related to specifying an endianness explicitly. When specifying an endianness for `NC_FLOAT`, the value would appear to not be written to file, if checked with `ncdump -s`. The issue was more subtle; the value would be written but was not being read from file properly for non-`NC_INT`. See [GitHub Issue](https://github.com/Unidata/netcdf-c/issues/112) or [NCF-331](https://bugtracking.unidata.ucar.edu/browse/NCF-331) for more information.
* Addressed an issue in netcdf4 on Windows w/DAP related to how byte values were copied with sscanf. Issue originally reported by Ellen Johnson at Mathworks, see [NCF-330](https://bugtracking.unidata.ucar.edu/browse/NCF-330) for more information.
* Addressed in issue in netcdf4 files on Windows, built with Microsoft Visual Studio, which could result in a memory leak. See [NCF-329](https://bugtracking.unidata.ucar.edu/browse/NCF-329) for more information.
* Addressed an issue in netcdf4 files where writing unlimited dimensions that were not declared at head of the dimensions list, as reported by Ellen Johnson at Mathworks. See [NCF-326](https://bugtracking.unidata.ucar.edu/browse/NCF-326) for more information.
* Fixed a bug related to renaming the attributes of coordinate variables in a subgroup. See [NCF-325](https://bugtracking.unidata.ucar.edu/browse/NCF-325) for more information.
* Fixed bug resulting in error closing a valid netCDF-4 file with a dimension and a non-coordinate variable with the same name. [NCF-324](https://bugtracking.unidata.ucar.edu/browse/NCF-324)
* Added functionality to make it easier to build `netcdf-fortran` as part of the `netcdf-c` build for *NON-MSVC* builds. This functionality is enabled at configure time by using the following **Highly Experimental** options:
> These make targets are **only** valid after `make install` has been invoked. This cannot be enforced rigidly in the makefile for reasons we will expand on in the documentation, but in short: `make install` may require sudo, but using sudo will discard environmental variables required when attempting to build netcdf-fortran in this manner.<br><br>
> It is important to note that this is functionality is for *convenience only*. It will remain possible to build `netcdf-c` and `netcdf-fortran` manually. These make targets should hopefully suffice for the majority of our users, but for corner cases it may still be required of the user to perform a manual build. [NCF-323](https://bugtracking.unidata.ucar.edu/browse/NCF-323)
* Added a failure state if the `m4` utility is not found on non-Windows systems; previously, the build would fail when it reached the point of invoking m4.
* Added an explicit check in the build systems (autotools, cmake) for the CURL-related option `CURLOPT_CHUNK_BGN_FUNCTION`. This option was introduced in libcurl version `7.21.0`. On installations which require libcurl and have this version, `CURLOPT_CHUNK_BGN_FUNCTION` will be available. Otherwise, it will not.
* The pnetcdf support was not properly being used to provide mpi parallel io for netcdf-3 classic files. The wrong dispatch table was being used. [NCF-319](https://bugtracking.unidata.ucar.edu/browse/NCF-319)
* In nccopy utility, provided proper default for unlimited dimension in chunk-size specification instead of requiring explicit chunk size. Added associated test. [NCF-321](https://bugtracking.unidata.ucar.edu/browse/NCF-321)
* Fixed documentation typo in FILL_DOUBLE definition in classic format specification grammar. Fixed other typos and inconsistencies in Doxygen version of User Guide.
* For nccopy and ncgen, added numeric options (-3, -4, -6, -7) for output format, to provide less confusing format version specifications than the error-prone equivalent -k options (-k1, -k2, -k3, -k4). The new numeric options are compatible with NCO's mnemonic version options. The old -k numeric options will still be accepted but are deprecated, due to easy confusion between format numbers and format names. [NCF-314](https://bugtracking.unidata.ucar.edu/browse/NCF-314)
* Fixed bug in ncgen. When classic format was in force (k=1 or k=4), the "long" datatype should be treated as int32. Was returning an error. [NCF-318](https://bugtracking.unidata.ucar.edu/browse/NCF-318)
* Fixed bug where if the netCDF-C library is built with the HDF5 library but without the HDF4 library and one attempts to open an HDF4 file, an abort occurs rather than returning a proper error code (NC_ENOTNC). [NCF-317](https://bugtracking.unidata.ucar.edu/browse/NCF-317)
* Added a new file, `netcdf_meta.h`. This file is generated automatically at configure time and contains information related to the capabilities of the netcdf library. This file may be used by projects dependent upon `netcdf` to make decisions during configuration, based on how the `netcdf` library was built. The macro `NC_HAVE_META_H` is defined in `netcdf.h`. Paired with judicious use of `ifdef`'s, this macro will indicate to developers whether or not the meta-header file is present. See [NCF-313](https://bugtracking.unidata.ucar.edu/browse/NCF-313).
* Fixed ncdump bug for char variables with multiple unlimited dimensions and added an associated test. Now the output CDL properly disambiguates dimension groupings, so that ncgen can generate the original file from the CDL. [NCF-310](https://bugtracking.unidata.ucar.edu/browse/NCF-310)
* Converted the [Manually-maintained FAQ page](http://www.unidata.ucar.edu/software/netcdf/docs/faq.html) into markdown and added it to the `docs/` directory. This way the html version will be generated when the rest of the documentation is built, the FAQ will be under version control, and it will be in a more visible location, hopefully making it easier to maintain.
* Bumped minimum required version of `cmake` to `2.8.12`. This was necessitated by the adoption of the new `CMAKE_MACOSX_RPATH` property, for use on OSX.
* Jennifer Adams has requested a reversion in behavior so that all dap requests include a constraint. Problem is caused by change in prefetch where if all variables are requested, then no constraint is generated. Fix is to always generate a constraint in prefetch.
* Added a new option for cmake-based builds, `ENABLE_DOXYGEN_LATEX_OUTPUT`. On those systems with `make` and `pdflatex`, setting this option **ON** will result in pdf versions of the documentation being built. This feature is experimental.
* Bumped minimum CMake version to `2.8.9` from `2.8.8` as part of a larger pull request contributed by Nico Schlömer. [Pull Request #64](https://github.com/Unidata/netcdf-c/pull/64)
* NetCDF creates a `libnetcdf.settings` file after configuration now, similar to those generated by `HDF4` and `HDF5`. It is installed into the same directory as the libraries. [NCF-303](https://bugtracking.unidata.ucar.edu/browse/NCF-303).
* Renamed `man4/` directory to `docs/` to make the purpose and contents clearer. See [man4 vs. docs #60](https://github.com/Unidata/netcdf-c/issues/60).
* Removed redundant variable `BUILD_DOCS` from the CMake configuration file. See the issue at github: [#59](https://github.com/Unidata/netcdf-c/issues/59).
* Added missing documentation templates to `man4/Makefile.am`, to correct an issue when trying to build the local `Doxygen`-generated documentation. This issue was reported by Nico Schlömer and may be viewed on github. [Releases miss Doxygen files #56](https://github.com/Unidata/netcdf-c/issues/56)
* When the NC_MPIPOSIX flag is given for parallel I/O access and the HDF5 library does not have the MPI-POSIX VFD configured in, the NC_MPIPOSIX flag is transparently aliased to the NC_MPIIO flag within the netCDF-4 library.
* As part of an ongoing project, the Doxygen-generated netcdf documentation has been reorganized. The goal is to make the documentation easier to parse, and to eliminate redundant material. This project is ongoing.
* The oc .dodsrc reader was improperly handling the user name and password entries. [NCF-299](https://bugtracking.unidata.ucar.edu/browse/NCF-299)
* CTestConfig.cmake has been made into a template so that users may easily specify the location of an alternative CDash-based Dashboard using the following two options:
*`NC_TEST_DROP_SITE` - Specify an alternative Dashboard by URL or IP address.
*`NC_CTEST_DROP_LOC_PREFIX` - Specify a prefix on the remote webserver relative to the root directory. This lets CTest accommodate dashboards that do not live at the top level of the web server.
* Added a separate flag to enable DAP AUTH tests. These tests are disabled by default. The flags for autotools and CMAKE-based builds are (respectively):
* Corrected "BAIL" macros to avoid infinite loop when logging is disabled and an error occurs.
* Refactored how types are used for attributes, variables, and committed types, clarifying and categorizing fields in structs, and eliminating duplicated type information between variables and types they use.
* Made type structure sharable by committed datatypes and variables that use it.
* Enabled netCDF-4 bit-for-bit reproducibility for nccopy and other applications (thanks to Rimvydas Jasinskas and Quincey Koziol) by turning off HDF5 object creation, access, and modification time tracking. [NCF-290](https://bugtracking.unidata.ucar.edu/browse/NCF-290)
* Addressed an issue where `cmake`-based builds would not properly create a `pkg-config` file. This file is now created properly by `cmake`. [NCF-288](https://bugtracking.unidata.ucar.edu/browse/NCF-288)
* Addressed an issue reported by Jeff Whitaker regarding `nc_inq_nvars` returning an incorrect number of dimensions (this issue was introduced in 4.3.1-rc5). Integrated a test contributed by Jeff Whitaker.
* When compiling with `hdf4` support, both autotools and cmake-based builds now properly look for the `libjpeg` dependency and will link against it when found (or complain if it's not). Also added `ENABLE_HDF4_FILE_TESTS` option to CMake-based builds.
* Added hack to the DAP code to address a problem with the Columbia.edu server. That server does not serve up proper DAP2 DDS replies. The Dataset {...} name changes depending on if the request has certain kinds of constraints. [NCF-276]
* Added a new configuration flag, `NC_USE_STATIC_CRT` for CMake-based Windows builds. The default value is 'OFF'. This will allow the user to define whether to use the shared CRT libraries (\\MD) or static CRT libraries (\\MT) in Visual Studio builds.
* Integrated a fix by Quincey Koziol which addressed a variation of [NCF-250], *Fix issue of netCDF-4 parallel independent access with unlimited dimension hanging*.
* Integrated change contributed by Orion Poplawski which integrated GNUInstallDirs into the netCDF-C CMake system; this will permit systems that install into lib64 (such as Fedora) to `make install` without problem.
* Corrected an error with the CMake config files that resulted in the `netcdf.3` manpage not being built or installed.
* Migrated from the netCDF-C `subversion` repository to a publicly available GitHub repository available at https://github.com/Unidata/netCDF-C. This repository may be checked out (cloned) with the following command:
* Added a `NC_HAVE_RENAME_GRP` macro to netcdf.h, [as per a request by Charlie Zender][cz1]. This will allow software compiling against netcdf to easily query whether or not nc\_rename\_grp() is available.
* Added Greg Sjaardema's contributed optimization for the nc4\_find\_dim\_len function in libsrc4/nc4internal.c. The patch eliminates several malloc/free calls that exist in the original coding.
* Added support for dynamic loading, to compliment the dynamic loading support introduced in hdf 1.8.11. Dynamic loading support depends on libdl, and is enabled as follows: [NCF-258]
* Fix issue of netCDF-4 parallel independent access with unlimited dimension hanging. Extending the size of an unlimited dimension in HDF5 must be a collective operation, so now an error is returned if trying to extend in independent access mode. [NCF-250]
* Fixed bug with netCDF-4's inability to read HDF5 scalar numeric attributes. Also allow, in addition to zero length strings, a new NULL pointer as a string value. to improve interoperability with HDF5. This required a new CDL constant, 'NIL', that can be output from ncdump for such a string value in an HDF5 or netCDF-4 file. The ncgen utility was also modified to properly handle such NIL values for strings. [NCF-56]
* Parallel-build portability fixes, particularly for OpenMPI and gcc/gfortran-4.8.x on OSX.
* Fix contributed by Nath Gopalaswamy to large file problem reading netCDF classic or 64-bit offset files that have a UINT32_MAX flag for large last record size of a variable that has values larger than 1 byte. This problem had previously been fixed for *writing* such data, but was only tested with an ncbyte variable.
