2.0.15-1:  December 08, 2005
	Forgot to update the release constants in mexnc.h.  No need to
	make a big deal out of that.


2.0.15:  December 07, 2005
	The 2.0.9 release tried to fix a rounding bug in the netcdf-2 code,
	but the fix itself doesn't work for negative integers.  A more
	robust method is now used.

	
2.0.14:  October 11, 2005
	Updated README to reflect new unix install instructions.
	Added ncerr message for one case in netcdf3.c.

2.0.13:  August 11, 2005
	Replaced if/then clause in test 4 of test__open m-file with try/catch
	clause.

	Added test__enddef.m to test suite.

	Additional documentation in test__create m-file.

	Fixed opcode definition for _OPEN in mexnc.h.

2.0.12:  August 8, 2005
	Allowing case of [] in 6th parameter position for VARDEF in
	mexgateway.c.

	Checking for 0 length of dimids argument list in netcdf2.c, VARDEF
	opcode.

	Backported nc__open support from netcdf-4 trunk.

2.0.11:  July 15, 2005
	Added support for nc__create, nc__enddef functions.

	Cleaned up the README and tests/README files a bit.  


2.0.10:  July 13, 2005
	The END_DEF operation is really a typo.  It should have been ENDDEF.
	So ENDDEF has been added, and END_DEF kept in case people are already
	using it.

	In netcdf3.c, was allocating one byte too many for the return buffer
	used by nc_get_att_text.

	Added new pass-thru routine mexcdf60.m.  It will serve the same
	purpose as mexcdf53.m.

	Tagged netcdf-4_branch_root.


2.0.9:  June 9, 2005
	Fixed scaling bug where truncation was used instead of rounding.

	The test suite has been altered so that the user must verify that
	their path is set correctly.

	Full type checking in the test suite implemented for all netcdf3
	functions.

	Type checking implemented for ncid, dimid, and varid arguments
	in all netcdf2 functions.

	Pretty much every function is vulnerable to a segmentation fault
	if [] is passed in as a parameter.  So I wrote two functions
	called from within mexgateway.c to check the inputs to make
	sure this does not happen.  I said "two" functions because the
	1st input argument, which identifies the operation, needs to be
	checked separately.  Once we are satisfied that the 1st input
	is ok and have extracted the operation name, we can check the
	others.  There are a couple of cases where [] has been historically
	used (but they weren't documented!!!).  So the function
	"check_other_args_for_empty_set" looks for these cases and allows them
	to pass thru unmolested.  

	Because of the methodical empty set checks now in place, the temporary
	hack introduced in 2.0.7 has been removed.

	VARGET1 no longer allowed to pass [] as the index argument.

	There is another case of [] being an allowed input, and that is for
	DEF_VAR in order to define a variable with no dimensions, i.e. a
	singleton variable.  Modified the documentation in mexnc.m to finally
	let the users know about this...

	Fixed mexopts.R14.sh so that one could more easily try to compile an
	opendap build.

	mexnc.m help section modified to include mention of the default OPEN
	mode.


2.0.8:  Apr 25, 2005
	All nctypelen function calls are now being checked for their result
	status before being used elsewhere.

	The "RepairBadDataTypes" function has been axed.  See the CVS log for
	the file "netcdf3.c", version 1.16, for the explanation.


2.0.7:  Apr 21, 2005
	Put a "temporary" piece code in netcdf2.c code that extracts the varid.
	The code checks that the empty set was not accidently passed for the
	varid.

	ATTPUT:  the code used to read that if the length were equal to
	-1, then set the length to actually be the product of mxGEtM *
	mxGetN.  I now make that the case whenever the length is negative.
	
	ATTGET:  the status was not being checked upon the return from
	ncattinq and ncattget.	 If bad cdfids or varids were passed
	into ATTGET, matlab would barf forth that it was out of memory.

        Added an m-file win32_make_mexnc for invoking the build on win32
	systems.

	Fixed minor bug in INQ_VARID section.  Some bogus code had slipped
	in, would cause a segmentation fault when a zero-length variable
	name was passed.

	Continued work on the test suite.


2.0.6:  Feb 25, 2005
	Added support for nc_set_fill.

	netcdf2.c was not adequately checking error status when handling
	VARPUT1, VARGET1, VARPUT, VARGET, VARPUTG, and VARGETG.  Specifically
	there would be a call to ncvarinq.  Failure to check for bad status
	would cause a segmentation fault.

	Removed some duplicated listings between netcdf-2 and netcdf-3 in
	mexnc.m.

	ATTCOPY routine (netcdf2.c) had an indexing error.  The 2nd varid
	should have been indexed with position 5, but it was 2 instead.

	Initial set of independent tests written for netcdf-3 code.

	Added nc_global.m m-file for the purpose of providing mnemonic for
	NC_GLOBAL constant defined in netcdf.h

	Added a static char array that used the CVS keyword Name in order to
	identify the particular release of mexnc that was checked out.  By
	using the strings command and piping the results thru "grep Name", one
	should clearly see the release id.

	Added support for nc_inq_attname.

	Added support for nc_inq_atttype.

	Got rid of default mexopts.sh file and replaced it with mexopts.R14.sh
	and mexopts.R13.sh.

	Added new convenience m-files nc_double, nc_float, nc_short, 
	nc_int, nc_byte, and nc_char.

	Added code for nc_inq_vartype, nc_inq_varndims, nc_inq_vardimid,
	nc_inq_varnatts.  Since the functionality of these routines
	is entirely provided by that of nc_inq_var, this really isn't
	terribly necessary.  Just trying to be complete.

	Added code for nc_inq_dimname and nc_inq_dimlen.  Since the 
	functionality of these routines is entirely provided by that of
	nc_inq_dim, this really isn't terribly necessary.  Just trying
	to be complete.

	Added code for nc_inq_nvars, nc_inq_ndims, nc_inq_natts.  Since the
	functionality of these routines is entirely provided by that of
	nc_inq, this really isn't terribly necessary.  Just trying to be
	complete.

	Modified mexopts.sh in order to easily switch to an opendap build.

	Fixed a warning in mexncCreateDoubleScalar routine.  The solaris 
	compiler complained about a line being unreachable.  

2.0.5:  Jan 28, 2005
	Found two bugs, one in the PUT_VARS_TEXT implementation, the other in
	the PUT_VARM_TEXT implementation.  They are related in that the code
	writer (that would be me!) used cut-and-paste to write the code
	instead of thinking it thru.  The input data set was not being
	properly identified, causing matlab to crash badly.  

	Fixed a few typos in the AUTHORS file.  Continued cleanup of the test
	suite.  Changed the MEXNC_RELEASE_NAME to include the string "MEXNC ".
	Changed the MEXCDF_H definition to MEXNC.  Pulled out any definitions
	that were unneccesary to the netcdf-3 api into a new include file
	called netcdf2.h.  
	
2.0.4:  Jan 19, 2005
	Revised the AUTHORS file some more. Continued cleanup of test suite.
	Added "get_mexnc_info" function call.  Fixed some bad instructions
	in the README file.  

	Fixed a typo in the AUTHORS file.

	The mexopts.sh file had unnecessary library flags -ltiff -lgeotiff
	in the Mac OS X section that had apparently migrated there from other
	projects.  They were removed.  Thanks to Dave Hebert for clueing me
	into this.

	Abstracted [put,get]_var_[int,float] routines in the testing suite 
	out into their own functions.
	
2.0.3:  Jan 4, 2005
	The parameter structs for GET_VAR_UCHAR and GET_ATT_SCHAR were 
	incorrect, meaning the actual C library call for those operations
	were never being invoked.  Good thing I've been making up this
	test suite, otherwise I never would have caught it.  

	Abstracted get_var_schar and put_var_schar tests into its own function
	in the testing suite.

	Abstracted get_var_short and put_var_short tests into its own function
	in the testing suite.

2.0.2:  Jan 3, 2005
	Abstracted get_var_text and put_var_text tests into its own function
	in the testing suite.

	There was some misleading text in the "mexnc ( 'open' ..." function
	description.  nc_64bit_offset_mode and nc_nofill_mode are not
	appropriate here.


2.0.1:  Dec 31, 2004
	The mexcdf module was renamed to "mexnc".  Several source files were
	renamed in order to minimize confusion between "mexcdf" and "mexnc".
	Peeled off NetCDF-3 code out of the mex entry point file into a file
	of its own, "netcdf3.c"

        The test m-file having to do with large file support was not
        writing the data correctly.


2.0.0:  Dec 20, 2004
	First builds for NetCDF 3.6.0

	Added support for GET_VARM_XXX family of function calls.

	Changed stride and imap arrays declarations to ptrdiff_t,
	which suppresses warning on solaris.

	Created a new directory called "tests".  Since the testing
	routines are going to get a lot more complicated, it makes
	sense not to pollute the main directory with this.

	Had forgotten to include ENDEF in list of NetCDF-2.4 opcodes.

	The routine "mexCreateDoubleScalar" was being used extensively,
	but it is not available on R11 platforms.  Since the poor slobs
	stuck with alphas are orphaned at R11, it was decided to stub out
	all calls to mxCreateDoubleScalar with "mexncCreateDoubleScalar".
	This new function consists of an "#ifdef #else #endif" clause,
	and if the makefile passes the compiler the flag -DMEXNC_R11,
	then comparable code for mxCreateDoubleScalar gets compiled
	instead.  This is preferable to replacing every line of code
	with mxCreateDoubleScalar with four lines of equivalent R11 code.

	New help file, "mexcdf53.m".  New nc_create and nc_open mode routines,

		nc_64bit_offset_mode.m
		nc_clobber_mode.m
		nc_fill_mode.m
		nc_lock_mode.m
		nc_noclobber_mode.m
		nc_nofill_mode.m
		nc_nowrite_mode.m
		nc_share_mode.m
		nc_write_mode.m
