$Header: /cvsroot/nco/nco/doc/TODO,v 1.1005 2008/05/11 21:43:48 zender Exp $ 

# Purpose: TODO items for NCO

# URL:
http://nco.sf.net/TODO

# Usage: Upload to homepage:
/usr/bin/scp ~/nco/doc/TODO nco.sf.net:/home/groups/n/nc/nco/htdocs
/usr/bin/scp ~/nco/doc/TODO dust.ess.uci.edu:/var/www/html/nco

******************************************************************************
NCO Wish list:
Numbers were assigned in chronological order in three categories:
Building (bldXXX), ncap-related (ncapXXX), and generic NCO (ncoXXX):
bldXXX: Items specifically related to building and installation.
ncapXXX: These items do not affect rest of NCO
ncoXXX: 15 high priority items are shown first, then another ~100 items
Question marks indicate items or implementations are debatable
Please help out wherever your experience allows.
******************************************************************************

************************************************************************
Begin ncoXXX TODOs
************************************************************************
Highest priority NCO-wide items that You can help with:
nco59. Add "averaged over dimension x from y to z" attribute to ncwa-processed variables
nco97. Add x_op=average... attributes to averagers for ncar_csm
nco124. Use `install' program and mkdir -p in bld/Makefile method
nco128. Ensure target 'dir' is made before *.d dependency files 
nco152. Change make or nco_dst.pl to add docs to /usr/doc in RPM
nco164. Spiff up NCO homepage web page (Logo?)
nco165. Create Makefile dependency to do 'make dir' if lib, obj, bin directories do not exist
nco212. Ensure structures are always passed by reference (default is pass by value)
nco312. Make TODO list more legible (change to, e.g., Emacs list/outline mode, LaTeX, Wiki?) 
nco313. Redo TODO subject categorization by discriminating based on skills required to tackle. Initial cateogorization might be e.g., autotools, bison, C, C++, Debian, flex, HTML, Perl, RPM, www, TeXInfo
nco356. ncwa benchmark yields incorrect answers for arrays exceeding 10^7 elements with xlc and gcc, why?
is this related to nco354?
zender@esmf04m:~/nco/data$ date;ncwa -O ${DATA}/mie/big.nc ${DATA}/mie/foo.nc;date;ncks -H ${DATA}/mie/foo.nc
Mon Jul 19 16:58:22 PDT 2004
ncwa: INFO No dimensions specified with -a, therefore reducing (averaging, taking minimum, etc.) over all dimensions
Mon Jul 19 16:58:47 PDT 2004
wvl_1e0 = 1
wvl_1e1 = 1
wvl_1e2 = 1
wvl_1e3 = 1
wvl_1e4 = 1
wvl_1e5 = 1
wvl_1e6 = 1
wvl_1e7 = 1
wvl_1e8 = 0.167772 (NB: this is approximately 1/6)
nco405. Modify multi-slabbing so that it places the output slabs in the order in which they were specified rather than the order they appear in the input file (Martin Schultz)

NCO-wide items of lesser priority:
nco50. Allow user-specified scalar weights to ncwa
nco51. Make getopt() a shared subroutine?
nco53. Add wall clock timer options? (to provide more info than, e.g., timex())
nco55. nco_var_cnf_dmn() should be able to expand (x,64,128) array into (nx,64,128) array (maybe)
nco60. Make something that can mask field without having to average over dimension
nco66. Make history attribute get appended, rather than overwritten, when in append mode
nco69. ncks option for file/field summary: var name list,var min,max,avg,mss_val,# mss_val,
nco75. Add warning to ncwa triggered by averaging over lat without weighting by gw with NCAR_CSM_FORMAT
nco79. Allow ncrename .$var_nm syntax to work on all operators (so variables not present cause warning, not error)
nco82. Make libnco namespace-safe by prefixing all library subroutines with nco_
nco86. Extend stride/wraparound capability to all operators possible
nco92. ncrcat -O -v base_time,time_offset arm.cdf arm.cdf foo.nc;ncks -H -C -h foo.nc | m gives bad time_offset on LINUX not SUNMP (problem with LINUX adding record coordinate to open file?)
nco99. Allow ncatted to dump file in ncatted format
nco103. Implement trapezoidal rule option in ncra, ncea
nco105. Add easy-to-read tabular display option for file contents for ncks, i.e, allow ncks -s to print multiple variables in column format
nco106. Add switch to keep degenerate dimensions with ncwa, remove them from ncra
nco111. Investigate whether using ncwa to average files where variables of type 
        NC_CHAR have a record dimension causes core dumps or other problems.
        i.e., LSM variable timecom caused subtle ncra problems for a while.
        Problem caused when weight or mask has dimension not in dimension extraction list
        ncwa -O -C -D 5 -v fl_nm -w gw ~/nco/data/in.nc ~/foo.nc
        fails because fl_nm has dimension char_dmn_lng but gw has dimension lat
        call to nco_var_fll() for gw requires that lat be in dmn_lst
nco113. Rethink normalization switch options in ncwa
nco114. Fix nco_var_cnf_dmn() so that returned weight always has same size tally array as template variable
nco115. SIGFPE when sum of denominator in ncwa (i.e., sum of weight) is zero
        Users should not be doing weighted averages if denominator sums to zero, but...
        Desired fix here is graceful exit instead of core dump
nco127. Make mss: signal retrieval from NCAR mass store
nco143. Add -W or -q switch to turn off squelch most warning messages in ncra, ncrcat (especially warnings about non-monotonicity)
nco144. Add option(s) to select all N-D variables where N=0,1,2...
nco145. Talk to Brian and implement latest netCDF and CF conventions, e.g., time_op
nco146. Allow proxies for coordinate variables so that, e.g., hyperslabs
        may be specified for "date" even though "time" is the coordinate. 
        Proxies must be one dimensional and monotonic, of course.
nco147. Use builtin mkstemp() so file is built in fast directory rather than, say, NFS-mounted directory like /fs/cgd  
nco148. Put in more useful error diagnostics for Schweitzer's data holes
        These only appear in ncra/ncrcat because ncks treats same hyperslab as a wrap!
        Must define a consistent convention here
        cd ~/nco/data
        ncks -O -C -d lon,10.0,80.0 -v time_lon in.nc foo.nc;ncks -H foo.nc
        ncrcat -O -C -d lon,10.0,80.0 -v time_lon in.nc foo.nc;ncks -H foo.nc
nco149. Add ncks hyperslab tests to nco_tst.pl to catch inconsistencies like #148
nco151. Dynamically modify web pages so files with changing version numbers can be directly linked
nco154. Add mirror capability to web pages
nco158. Add policy for hyperslabbing single level values to User's Guide
        Multi-file operators with record coordinate
        treat single point hyperslabs differently than single file operators 
        or than any operators with single point cuts on non-record coordinates.
        See explanatory notes in nco_lmt_evl().
nco159. Add capability to handle superfluous intermediate files to ncra,
        ncrcat when lmt_typ = dim_idx.
        This would allow these multifile operators to handle hyperslabs where,
        for example, every other file was not needed. This capability is already
        implemented for lmt_typ = crd_val. 
        This is an issue only when stride is non-unity.
        Does multislabbing algorithm fix this?
nco162. Add GiNaC (Computer Algebra System) support to ncap www.ginac.de (maybe?)
nco166. Document 'make test' so people know it exists and use it
nco173. Add switch to ncra which allows user to remove degenerate time dimension from result
nco174. Setup ncra,ncea,ncwa to do standard deviations sdn with one call
nco182. Add -L, --license switch to print license terms (as suggested by GNU coding style)
nco190. Make typ_pck et al. use nc_type enum or something sane for default
nco192. ncdiff will sometimes subtract files of incommensurate sizes without complaint
        ncks -O -v one_dmn_rec_var -d time,0,1 in.nc foo1.nc
        ncks -O -v one_dmn_rec_var -d time,0,2 in.nc foo2.nc
        ncdiff -O foo1.nc foo2.nc foo3.nc
        This may make sense, e.g., for 1-D record variables of different lengths, but not for multidimensional variables
        Former case should print warning, latter case should exit with error
nco193. Convert dimension sizes from type "long int" to type "size_t"?
nco195. Itanium builds!
nco199. Internationalize (i18n) NCO with gettext()
nco203. Perform benchmarking tests to quantify arithmetic performance improvements, I/O bottlenecks, etc.
nco209. Add rules to automagically make MANIFEST and update Sourceforge download area
nco215. nco_var_cnf_dmn() does not abort when variables do not conform because dimension list of one is subset of other but ORDER of dimensions differs, e.g., a(lat,lev,lon) !~ b(lon,lev). Attempting this will return incorrect answers!
nco222. Warn when ncea/ncra/ncrcat/ncecat operate on scaled variables ("Results will be meaningless if scale_factor and add_offset are not identical in each file")
nco225. Pass aed by reference in nco_aed_prc()
nco226. Is xrf in nco_var_cnf_dmn() is really necessary? If not, remove it and make wgt arg const var_sct * const
nco228. g++ linking requiring #ifndef GNUC++ around fabsf,fmodf defs in nco_var_scv.c
nco230. Keep track of type of _FillValue field in variable structure for reasons described in nco_cnf_var_typ()/nco_cnv_mss_val_typ()
nco232. Warn when arithmetic operators encounter packed variables?
nco233. Makefile.old rule for make tst appears broken for long FTP rule
nco237. Break down nco.texi into bite-size chunks, e.g., one file per section
nco238. Replace y=(typ *)nco_malloc(nbr*sizeof(typ)) with y=(typ *)nco_malloc(nbr*sizeof(*y))?
nco239. Create one global compatibility file (nco_xpf.[ch]?) activated by HAVE_ or NEED_ actions so do not have multiple platform dependent #includes as is case now with HAVE_STRCASECMP, HAVE_STRDUP requiring nco_sng_utl.h
nco241. Add run-time switch which requests that output be _FillValue where ANY input field is _FillValue. NCO discussion forum 20021212.
nco245. Eliminate all instances of passing naked constants
nco246. Deprecate USE_FORTRAN_ARITHMETIC from Makefile
nco250. ncks -s should refuse to print strings formatted with %s unless they contain NUL-terminator. This will avoid segfaults like
        nconcks -C -H -s "%s" -v fl_nm ~/nco/data/in.nc
nco251. Does ncks -B correctly write packed variables in binary format?
nco255. Desktop icon for NCO gui interface
nco256. Turn udunits into .deb
nco258. Develop syntax/procedure for specifying lists of hyperslabs for single pass, low memory consumption MSA jobs like Martin Schultz's in Discussion forum on 20030506
nco259. Variadic function extension to nco_malloc() to support printing debugging messages?
nco260. Warn (but work) when coordinate variable has more than two dimensions
nco261. Switch from toggling defaults to --no-option format for all Booleans. e.g., --abc should ensure alphabetization is True, and --no-abc should ensure it is false, rather than toggling.
nco265. ncbo: Promote before operations using nco_typ_cnv_rth() in ncbo
nco266. ncbo: Finish ncbo.1 manpage
nco270. Automate uploading of tagged .deb builds to ftp://ftp.debian.org
nco275. Use nco_mss_val_cnf() in ncflint
nco276. Use nco_mss_val_cnf() in ncap
nco277. Use nco_mss_val_cnf() in ncwa
nco287. Develop dianostics for OMP OpenMP implementation in ncra
nco292. nco_fl_utl.c: Use autoconf HAVE_NETWORK instead of WIN32
nco293. nco_fl_utl.c: Use autoconf HAVE_MKDIR_M instead of SUN4
nco294. Globals variable in NCO operators are prg, prg_nm, dbg_lvl 
        If possible, modify so that shared-library (rather than main.c) holds
        and returns thread-specific values for each, set with, e.g., prg_set().
        Make prg, prg_nm, and dbg_lvl local to main(), not global to main.c
nco295. Understand/maintain re-entrant behavior of shared library.
nco298. Localize ncap's three globals: ncap_ncl_dpt_crr, ncap_ln_nbr_crr, ncap_fl_spt_glb?
nco299. Prefix prg,prg_nm,dbg_lvl globals with nco_ for namespace cleanliness
nco300. Activate ccmalloc libdl into configure.in for enable-debug-custom
nco301. Unable to use DDD on autotools NCO executable with --enable-debug-custom
nco302. Add speedy GCC optimization flags to --enable-optimize-custom
nco303. Make test suite more comprehensive
nco309. Free memory allocated in nco_fl_utl.c: fl_lst_in=(char **)nco_malloc(sizeof(char *)); /* fxm: free() this memory sometime */
nco310. Add ENABLE_DEBUG_CUSTOM, ENABLE_OPTIMIZE_CUSTOM, ... to bld/Makefile
nco311. Automatically convert int,short to float in nco_var_pwr()
nco315. GNUC non-ANSI rank reduction methods in nco_var_avg.c fail on large files. Problem with stack memory allocation?
nco316. Change offending nco_malloc() to nco_malloc_flg() calls for #315
nco317. Understand stack memory problem. ulimits? compiler switches?
nco318. Solve stack memory problem and benchmark vs. ANSI code branch
nco320. How to do TeX-like hyphens in TeXInfo?
nco321. Do not promote NC_FLOAT to NC_DOUBLE in ncra/ncea(?)
nco322. Implement integer empowerment in var_scv_pwr() with, e.g., GSL
nco323. Collect all past ANNOUNCE files and place in reverse chronological order in ANNOUNCE file
nco324. Add release date to "Version Summary" section on homepage
nco326. Add manual override of writing *tmp* file for intermediate results?
nco328. Ability to reverse specified dimensions (use negative stride?)
nco333. Clean up preprocessor symbols with cd ~/nco/src/nco;ifnames *.c
nco335. ncflint: implement interpolation/mss_val algorithms two and three as options?
nco336. Allow input file lists to be specified in a text file and to supercede command line arguments. This will be useful on platforms with limited buffers for command line arguments and globbing expansions (e.g., AIX)
nco337. When input file is a symbolic link, print name of linked-to-file as debugging information
nco340. o/a/e stdin is overwritten when stdin input method used
        cd ~/nco/data;echo "in.nc" | ncecat -D 3 -o foo.nc
nco342. Get canonicalize_file_name() working in nco_fl_info_get()
nco344. Create sophisticated etags regular expressions for change version number in all documentation tags_doc.sh
nco352. Probably related to 351: OpenMP directive must appear in main() or ldd will not associate /usr/lib/libxlsmp.a(smprt_64.o) with executable so rtld errors occur on AIX
************** Important ***************************
nco354. ncwa dumps core on AIX ESMF on large files regardless of OpenMP: 
	fxm: Does this show need to use "long long" instead of "long" for counts/buffers?
        ncwa -D 5 -O -a wvl -v ss_co_alb ${DATA}/mie/mie_1000000.nc ${DATA}/mie/foo_avg_1000000.nc (works)
        ncwa -D 5 -O -a wvl -v ss_co_alb ${DATA}/mie/mie_10000000.nc ${DATA}/mie/foo_avg_10000000.nc (fails)
        fails with 
        Segmentation fault (core dumped)
nco359. Allow spaces between variables names in -v var_lst arguments (assuming list whitespace is quoted, of course)
nco362. mmr: reduce memory during run by free()'ing var_fix[idx],var_fix_out[idx],var_prc[idx],var_prc_out[idx] after writes
nco363. ncpdq: document netCDF limitations in nco.texi
nco375. Do something more consistent with: nco_exit(EXIT_FAILURE)
nco385. ncpdq: dynamically allocate arrays in nco_var_dmn_rdr_mtd(), nco_var_dmn_rdr_val()?
nco386. Remove purely diagnostic arrays from nco_var_dmn_rdr_mtd(), nco_var_dmn_rdr_val()?
nco389. ncpdq: Single dimension in re-order list with no negative sign is always identity transformation, warn? exit?
nco390. ncpdq: make -a optional? (all switches are optional in all other operators)
nco392. OpenMP: thread main loop of ncatted
nco395. ncpdq: document rdr shr lists
nco396. free() memory at end of each operator
nco397. make ncpack symbolic link to ncpdq? Re-order/pack are completely independent yet need same structure of program
nco404. Does Ersoy's XML stylesheet nco.css work?
nco406. Add advantages/disadvantages of NCO (compared to e.g., ANTS) description?
nco421. ncpdq: Make packing offer to move missing value to inside range
nco425. ncpdq: Make unpacking offer to unpack missing value to preserve its usefulness
nco426. ncpdq: Make packing offer to pack missing value to the packed type to preserve its usefulness
nco427. ncpdq, ncra: Fix pck_dbg problems in nco_var_get() which automatically down-convert missing value to typ_dsk. How to reconcile with nco457?
nco441. Packing NC_DOUBLE->NC_CHAR, NC_INT->NC_CHAR, or
   NC_SHORT->NC_CHAR and then unpacking, places 0.0
        instead of 1.0 in first element, why? implicit conversion of NUL byte when value = 0.
        ncpdq -D 5 -O -C -P all_new -M flt_chr -v three_dmn_var_dbl ~/nco/data/in.nc ~/foo.nc
        ncks -m -H -v three_dmn_var_dbl ~/foo.nc
        ncpdq -D 3 -O -C -P upk ~/foo.nc ~/foo_upk.nc
        ncks -m -H -v three_dmn_var_dbl ~/foo_upk.nc
nco453. Re-think handling of dangling pointers in nco_pck_val()?
nco454. ncpdq: Combine nco_pck_map,nco_pck_typ into new nco_pck_sct to avoid proliferation of argument passing
nco457. ncpdq: _FillValue is not converted to variable type when variable is not packed three_dmn_var_dbl. How to reconcile with nco427?
nco461. Always plenty of great ideas for new NCO functionality in http://tinyurl.com/6pzqd
nco462. ncpdq: pack/unpack missing_data just like variable data? The reason I did not do this originally is because I did not want the _FillValue to contribute to the dynamic range of values which need packing. Since my _FillValues are always huge (1.0e36), packing _FillValues would greatly diminish the precision of the valid packed data. Best solution appears to be to check if _FillValue distorts valid_range, and, if so, pack _FillValue to default value of packed type.
nco463. add optional scale_factor, add_offset arguments to ncap pack()?
nco464. ncpdq: convert _FillValue to packed type
        ncpdq -O -D 4 -C -P all_new -v mss_val ~/nco/data/in.nc ~/foo.nc
        ncpdq -O -C -U -v mss_val ~/foo.nc ~/foo_upk.nc
        ncks -C -m -H ~/foo_upk.nc
nco467. Keep track of CDO "Climate Data Operators" http://www.mpimet.mpg.de/~cdo/
nco469. Use default _FillValue e.g., NC_FILL_FLOAT, or values from <limits.h>, <float.h>?
nco474. Incorporate alternative "chunking" for ncrcat when data are one-dimensional, to avoid high disk/network access incurred by retrieving one float at a time
nco477. Make default to copy but not to print associated coordinate variables?
nco479. netCDF/L from Loci: http://loci.cs.utk.edu/modules.php?name=News&file=article&sid=40
nco480. Add boxcar averaging option (average in groups of n records) to ncra
nco482. Remove return's from void functions dated 20050109 once gcc-3.4.2 behavior fixed or gcc upgraded on i386 platforms
nco483. Compiling with ABORT_ON_ERROR and calling without an option dumps core
        Fix so inputting 0 filenames merely prints usage, does not abort?
nco485. Add md5sum hash functionality to ncks
nco487. Add ncks option to suppress over-writing attributes in output file by attributes of source file#1
nco494. Implement gzip library wrapper library build option for znetCDF?
nco514. MPI builds call superfluous linker steps:
        make MPI=Y > foo 2>&1
nco516. Implement SDO-wide data pipes, e.g., ncra in*.nc | ncwa | ncrename t,tpt foo.nc
        This would bring us one step closer to grid services, too
nco527. Purify ncatted memory: 
        valgrind --leak-check=yes --show-reachable=yes --suppressions=${HOME}/nco/doc/valgrind.txt --tool=memcheck ncatted -D 5 -O -a float_att,,c,f,74 -p ${HOME}/nco/data -l /tmp in.nc ~/foo.nc > ~/foo 2>&1
        valgrind --leak-check=yes --show-reachable=yes --suppressions=${HOME}/nco/doc/valgrind.txt --tool=memcheck ncatted -D 5 -O -a char_att,att_var,a,c,"and appended Sentence three,with,lots,of,commas" -a short_att,att_var,c,s,37,38,39 -a float_att,att_var,d,,, -a long_att,att_var,o,l,37 -a new_att,att_var,o,d,73,74,75 -p ${HOME}/nco/data -l /tmp in.nc ~/foo.nc > ~/foo 2>&1
        valgrind --leak-check=yes --show-reachable=yes --suppressions=${HOME}/nco/doc/valgrind.txt --tool=memcheck ncatted -D 5 -O -a float_att,,a,f,74,75,76 -p ${HOME}/nco/data -l /tmp in.nc ~/foo.nc > ~/foo 2>&1
# Commas at end of following aed_sct appear never to be free()'d, are responsible for three unfree'd bytes
        valgrind --leak-check=yes --show-reachable=yes --suppressions=${HOME}/nco/doc/valgrind.txt --tool=memcheck ncatted -D 5 -O -a float_att,att_var,d,,, -p ${HOME}/nco/data -l /tmp in.nc ~/foo.nc > ~/foo 2>&1
nco533. Implement C99 mathematical special functions: lgamma_r, j0, j1, jn, y0, y1, yn
nco540. bm: regression tests report some apparent successes as failures on SUNMP, e.g., 
	Result: [1e+36] != Expected: [1e+36]
	always recorded as failure. 
	Seems like any "Expected's" containing "+" string may fail.
	Perl problem?
nco542. Put batch scripts, executables in ${DATA}/nco_bm/CASEID? Where to put created files
nco546. Use makeatletter command in CCSM example in nco.texi?
nco547. Format CCSM example in nco.texi
nco548. Hyper-link all switches to explanatory sections?
nco552. ncbo: symmetric w/r/t fl_1 <--> fl_2, Step 1: allow smaller variables to be in file_1 (pre-compute broadcast size and place in var_out before var_def(out_id,var_out))
nco560. Change FTP access script to facilitate mget *?
	NCO cannot process files obtained this way unless mget feeds
	information on files obtained back to main()
	Script modification would be simple:
	ftp -i climate.llnl.gov << END
	bin
	cd sresa1b/atm/mo/tas/ncar_ccsm3_0/run2
	lcd /data/zender/sresa1b/atm/mo/tas/ncar_ccsm3_0/run2
	mget *
	quit
	END
nco564. Implement --64bit switch in ncatted?
nco565. bm: Add ncatted bm to quantify nc__enddef() tuning improvements
nco566. Implement --64bit switch in ncrename?
nco567. Add option to turn off default-creation of .tmp file?
nco568. bm: test NC_SHARE in ncatted/ncrename?
nco569. bm: Add SRB test
        Sifang Lu <sifang@sdsc.edu>
        /home/zender.uci-esmf/zender/nco/data/in.nc
        Sinit -v
        Scd zender
        Sls
        Sput ~/nco/data/in.nc nco/data
        Sget nco/data/in.nc
nco570. Merge ncatted/ncrename file opens into nco_fl_out_open()
nco573. Consolidate -4, --64bit into one option a la ncks --fl_fmt
nco576. Re-order arguments so var_sbt computes op2:=op1-op2 rather than op2:=op2-op1
nco577. ncbo: get rid of dangling pointers in var_1, var_2 after ncap_var_cnf_dmn()
nco578. ncbo: Use ncbo.c as template for [mp]ncflint re-write
nco581. ncbo: symmetric w/r/t fl_1 <--> fl_2, Step 2: allow nbr_xtr_out=max(nbr_xtr_1,nbr_xtr_2) when nbr_xtr_2>nbr_xtr_1 if discrepant variables are type fix (e.g., grid properties) not type prc (NB: already works for nbr_xtr_1>nbr_xtr_2)
nco587. Command line MPI on AIX PPE with poe
	Promising documentation at:
        http://publib.boulder.ibm.com/infocenter/clresctr/index.jsp?topic=/com.ibm.cluster.pe.doc/pe_42/am10600203.html
	echo ${HOST} > ~/nco/bld/host.list
        cd ~/nco/bld;poe -hostfile=~/nco/bld/host.list -node=1 -tasks_per_node=3 $MY_BIN_DIR/mpncwa -O -D 3 -p ${HOME}/nco/data in.nc ~/foo.nc
        zender@esmf04m:~/nco/bld$ cd ~/nco/bld;poe -hostfile=~/nco/bld/host.list -node=1 -tasks_per_node=3 $MY_BIN_DIR/mpncwa -O -D 3 -p ${HOME}/nco/data in.nc ~/foo.nc
        ERROR: 0031-212  pmd: node esmf04m: user zender denied from access from host esmf04m
        ERROR: 0031-024  esmf04m: no response; rc = -1
        AIX POE sample hello world at http://publib.boulder.ibm.com/infocenter/clresctr/index.jsp?topic=/com.ibm.cluster.pe.doc/pe_411/am105001/am10500113.html
        ocd ~/nco/bld;poe -infolevel 6 -node 1 -tasks_per_node 3 -rmpool 1 $MY_BIN_DIR/mpncwa -O -D 3 -p ${HOME}/nco/data in.nc ~/foo.nc
nco593. mpncrcat "sometimes" hangs while concatenating
	command line: -np 2 seems to always work on sand, ashes, elnino
	command line: -np 3 rarely hangs on ashes, usually hangs on sand, never hangs on elnino
                      When it does hang on ashes, it hangs after this message:
                      DEBUG: prc_rnk 1 waiting for msg from Mgr for final write
	command line: -np 4 always hangs on ashes and sand, never hangs on elnino
        ncrcat -O -D 5 -p ~/nco/data in.nc in.nc ~/foo.nc (works)
	mpirun -np 2 mpncrcat -O -D 5 -p ~/nco/data in.nc in.nc ~/foo.nc (fails, depending on np and machine)
        /bin/rm $MY_OBJ_DIR/mpncra.o;make MPI_FAKE=Y mpncra
	mpncrcat -O -D 5 -p ~/nco/data in.nc in.nc ~/foo.nc (works)
        valgrind --leak-check=yes --show-reachable=yes --suppressions=${HOME}/nco/doc/valgrind.txt --tool=memcheck mpncrcat -O -D 5 -v one_dmn_rec_var -p ~/nco/data in.nc in.nc ~/foo.nc > ~/foo 2>&1 (perfect)
        /bin/rm $MY_OBJ_DIR/mpncra.o;make MPI=Y mpncra
        Shorten for GDB:
        ncks -O -C -d time,0,1 -v one_dmn_rec_var ~/nco/data/in.nc ~/in_short.nc
        mpirun -np 3 $MY_BIN_DIR/mpncrcat -O -C -D 1 -v one_dmn_rec_var ~/in_short.nc ~/foo3.nc (hangs)

        DEBUG: input file opened in prc_rnk 0 inside the loop
        **** prc 0 hangs somewhere between these lines ***
        DEBUG: prc_rnk 0 at the end of ncra/rcat
        DEBUG: prc_rnk 0 is out of file idx loop

        ~/nco/bm/nco_bm.pl --mpi_prc=3 --regress ncrcat
        mpirun -np 2 /usr/bin/valgrind --leak-check=yes --show-reachable=yes --suppressions=${HOME}/nco/doc/valgrind.txt --tool=memcheck $MY_BIN_DIR/mpncrcat -O -D 5 -v one_dmn_rec_var -p ~/nco/data in.nc in.nc ~/foo.nc > ~/foo 2>&1
nco597. mpncea hangs on ncea regression test 4 (on sand with LAM)
	Hangs occur _only_ on sand with LAM in nco_bm.pl, not from command line!
	valgrind shows no NCO-related memory problems on sand with -np 2 or 3
        ~/nco/bm/nco_bm.pl --mpi_prc=3 --regress ncea
        ncea -O -D 5 -C -v one -p ~/nco/data in.nc ~/foo.nc (works)
        mpirun -np 2 /home/zender/bin/LINUXAMD64/mpncea -O -D 5 -C -v one -p ~/nco/data in.nc ~/foo.nc (leaves processes hanging in nco_free(var_out[0]))
        mpirun -np 2 /usr/local/bin/valgrind --leak-check=yes --show-reachable=yes --suppressions=${HOME}/nco/doc/valgrind.txt --tool=memcheck $MY_BIN_DIR/mpncea -O -D 5 -v one -p ~/nco/data in.nc in.nc ~/foo.nc > ~/foo 2>&1
	Daniel finds mpncea tries to free dangling pointer on sand with:
        mpirun -np 4 valgrind -v /home/zender/bin/LINUXAMD64/mpncra -Y ncea -h -O -D 2 -C -v pck -p ../data in.nc ~/foo.nc > ~/foo 2>&1
	Charlie cannot reproduce that.
	Charlie finds that master exits but workers reproducibly hang on sand with:
        mpirun -np 2 valgrind -v /home/zender/bin/LINUXAMD64/mpncra -Y ncea -h -O -D 2 -C -v pck -p ../data in.nc ~/foo.nc > ~/foo 2>&1
nco599. Allow targets to determine Makefile variables so, e.g., 'make mpi' automagically defines MPI=Y
nco607. mpnc*: Define structures (rather than int arrays) for MPI messages
nco609. Tune (or remove?) sleep interval in MPI operators
nco612. ncks ncdump-like features for netcdf4 files structures
nco618. XML data discovery on NCAR CDP via Luca Cinquini protocol?
nco622. NULL'ify opt_crr unless opt_lng detected?
nco624. Document MPI in users guide/manual
nco630. Change if thr_nbr > 0 to if thr_nbr > 1 for nco_thr_att_cat()?
nco631. Change usleep() to nanosleep()
nco635. ncwa: MACOSX "regressions" caused by broken FREEBSD/MACOSX 'cut' binary
        Solution may be to install GNU textutils?
        cd ~/nco/bm;~/nco/bm/nco_bm.pl --dbg=0 --regress --udpreport ncwa
        ncwa -h -O -D 0 -y min -v three_dmn_var_dbl -a lon -p ~/nco/data in.nc ${DATA}/nco_bm/foo.nc
        ncks -C -H -s '%f,' -v three_dmn_var_dbl ${DATA}/nco_bm/foo.nc > ${DATA}/nco_bm/foo
        cut -d, -f 7 ${DATA}/nco_bm/foo # (breaks on MACOSX,FREEBSD? fine on AIX, LINUX)
nco636. Try pnetcdf: http://www-unix.mcs.anl.gov/parallel-netcdf/
nco639. Fix packing code to be parallel over in_id without requiring 
	extra var->nc_id initialization everywhere
nco640. Naming: mpnc* (as currently implemented) or ncmpi* (as in pnetcdf)?
nco641. Use netCDF4 MPI I/O hooks in nco_open/nco_create (mpncecat.c is testbed)
nco642. Create nco_pnetcdf.h, nco_pnetcdf.c wrapper replacements?
nco643. bld/Makefile fails with netCDF4 and MPI simultaneously enabled
	make NETCDF4=Y MPI=Y ends with error in first compile (nco_att_utl.o)
	/usr/include/mpi.h:95: error: two or more data types in declaration specifiers
	/usr/include/mpi.h:278: error: two or more data types in declaration specifiers
nco644. Where is nc_open_par()? not in libnetcdf.a?
nco645. Does wrapping MPI-IO routines for netCDF4 and pnetCDF require ENABLE_MPI?
	Perhaps same MPI must be used to build netCDF4 as is used for NCO?
nco646. pnetCDF builds on elnino, not ashes. Why?
        ashes pnetCDF builds fail with both LAM and MPICH2 here in system header:
	/usr/bin/mpicc -c -O -g -I.   -O -g mpinetcdf.c
	In file included from nc.h:14,
                 from mpinetcdf.c:8:
	/usr/lib/gcc/i486-linux-gnu/4.0.2/include/stddef.h:152: error: two or more data types in declaration specifiers
	/usr/lib/gcc/i486-linux-gnu/4.0.2/include/stddef.h:214: error: duplicate unsignednco
nco648. Does mpncra create dangling pointers when it packs/unpacks variables?
nco649. Add ncks --prn_cln switch to print coords/vars naked (no names, indices)
nco651. Use hdr_pad standard to add documentation for long switches without short counterparts, e.g., --cmp, --mpi 
nco652. Polish/finish ncap random number generator
        ncap -O -v -s 'x=rnd_nbr(one)' ~/nco/data/in.nc ~/foo.nc;ncks ~/foo.nc
nco653. Segregate MPI from non-MPI builds completely in bld/Makefile on AIX
nco655. Incorporate BLAS if possible?
nco658. netcdf4 regression in ncrename test #1: Regression re-verified 20060924, 20070831, 20080220 PBN4
        make NETCDF4=Y FL_FMT=netcdf4 tst 
        ncks -O --fl_fmt=netcdf4 -v lat_T42 ~/nco/data/in.nc ~/foo.nc;od -An -c -N4 ~/foo.nc
        ncrename -O -d lat_T42,lat -v lat_T42,lat ~/foo.nc
nco659. C++: nco_mmr.c: errno presence triggers g++ error: undefined reference to `__errno_location()'
nco660. C++: NCO does not link to libudunits.a
nco664. ncatted & ncrename will not work with DAP because there is not any
	way currently to retrieve the DAP files to local files for them to
	operate on.  see nco_fl_utl.c:~813 in nco_fl_mk_lcl().  
	Currently just fails semi-gracefully.
nco666. Grid regridding/interpolation operator ncrebin?
nco668. Prefix nco.png fractal icon with text "NC" to look like "NCO"
nco669. Command-line switch for float->double math-promotion to avoid underflow?
nco675. ncap2: Makefile generates redundant antlr runs
nco680. ncap memory double-free:
	valgrind --leak-check=yes --show-reachable=yes --suppressions=${HOME}/nco/doc/valgrind.txt --tool=memcheck ncap -O -s 'foo=unpack(pck)' ~/nco/data/in.nc ~/foo.nc > ~/foo 2>&1
        ncap -O -s 'foo=unpack(pck)' ~/nco/data/in.nc ~/foo.nc
==15384== Invalid read of size 4
==15384==    at 0x80772E2: nco_var_free (nco_var_utl.c:692)
==15384==    by 0x8077429: nco_var_lst_free (nco_var_utl.c:720)
==15384==    by 0x804BEF1: main (ncap.c:772)
==15384==  Address 0x0 is not stack'd, malloc'd or (recently) free'd
nco681. Rename nco++:
	Invoke.cc		sdo_nvk.cc
	Makefile		"
	Makefile.orig		"
	Ncap2.hh		sdo.hh // [] ncap_type
	NcapVar.hh		sdo_var.hh // [cls] NcapVar class definition and declaration
	NcapVarVector.cc	sdo_var_vct.cc // [] NcapVarVector class declaration
	NcapVarVector.hh	sdo_var_vct.hh // [] NcapVarVector class definition
	NcapVector.hh		sdo_vct.hh // [] Ncapvector template class
	VarOp.hh		sdo_var_opr.hh // [] VarOp template class
	libnco++.hh		"
	ncap2.cc		"
	ncap2.hh		"
	ncap2_utl.cc		"
	ncoGrammer.g		sdo_grammar.g
	ncoLexer.cpp		sdo_lexer.cc
	ncoLexer.hpp		sdo_lexer.hh
	ncoParser.cpp		sdo_parser.cc
	ncoParser.hpp		sdo_parser.hh
	ncoParserTokenTypes.hpp	sdo_tkn_typ.cc
	ncoParserTokenTypes.txt	sdo_tkn_typ.txt
	ncoTree.cpp		sdo_tree.cc
	ncoTree.hpp		sdo_tree.hh
nco683. Convert to subversion http://sf.net/docs/E09/
nco685. Run static analysis tool on NCO codebase
nco689. Multislabs (multiple limits on same dimension) fail in arithmetic operators. Currently, arithmetic operators silently ignore all limit specifications after the first. Solution: Diagnose and fail with informative warning until nco690 implemented.
        ncwa -O -a lat,lon -w area -d lat,30.,90. -d lon,345.,360. -d lon,0.,195. isccpd_0112_clm_landmask.nc foo4.nc
nco690. Implement multislabs in arithmetic operators (no reason not to!)
nco693. ncks regression for offset past end of file test
	ncks -h -O -C -d time,0,1 -v time ~/nco/data/in.nc ~/foo.nc
	ncks -C -H -s '%g' -C -d time,2, ~/foo.nc
        ncks: ERROR User-specified dimension index range 2 <= time <= 1 does not fall within valid dimension index range 0 <= time <= 1
        ncks behaves perfectly here. Unfortunately, the perl test script does not recognize the expected answer string as being valid
nco694. Change nco_bool to C99 standard bool as per Wikipedia?
nco702. How to mix new ncwa behavior with -I wgt_msk_crd_var?
nco705. Add lmn_nbr_wgt and DID_BROADCAST outputs to nco_var_cnf_dmn()?
nco706. Add DID_REDUCE outputs to nco_var_avg_reduce()?
nco709. Change wgt_brd_flg (may have wrong meaning)
nco717. Averaging empty record variable creates an element in output!:
	ncwa -O ~/nco/data/in_rec_zero.nc ~/foo.nc
	ncwa -O -a lon ~/nco/data/in_rec_zero.nc ~/foo.nc
	This seems unavoidable since variable must be a scalar in output,
	and scalars automatically have one element of space reserved, whereas
	zero-record arrays have not elements of space reserved.
	Scalar is ill-defined of course.
	Hence averaging the file increases the data storage.
	What to do?  Nothing.
nco722. mpncwa: DDRA diagnostics should work in MPI mode, not just SMP mode
nco725. Add attribute propagation regression test
        ncap2 -v -O -s "cnv_CF_crd2=cnv_CF_crd" -s "lon2=lon(0)" -p ~/nco/data in.nc ~/foo.nc
nco726. attribute propagation: document
nco737. Add regression for regex'ing consecutive Q's?
	ncks -O -v 'Q[0-1][0-9]|2[0-3]' ~/nco/data/in.nc ~/foo.nc;ncks -H ~/foo.nc
	ncks -O -v 'Q[0-9][0-9]$' ~/nco/data/in.nc ~/foo.nc;ncks -H ~/foo.nc
nco738. nco_cnv_cf_cll_mth_add() CF cell methods
nco744. Add option to eliminate degenerate dimensions in ncra
nco745. Add regression test for ncwa --rdd
nco747. Change lat to -45,45 and add lat_grd = -90,0,90 in in.cdl
nco754. Make sure private/firstprivate aspects of ddra_info are correct
nco755. Test timers on ESMF under openmp
nco756. Make nco_ddra() thread-safe
nco757. ncbo bug: works with --mmr_drt, fails with --mmr_cln
	ncwa -O -a time -p ~/nco/data in.nc ~/foo_avg.nc
	ncbo --mmr_cln -O -D 1 ~/nco/data/in.nc ~/foo_avg.nc ~/foo.nc
	valgrind --leak-check=yes --show-reachable=yes --suppressions=${HOME}/nco/doc/valgrind.txt --tool=memcheck ncbo --mmr_cln -O ~/nco/data/in.nc ~/foo_avg.nc ~/foo.nc > ~/foo 2>&1
	valgrind --leak-check=yes --show-reachable=yes --suppressions=${HOME}/nco/doc/valgrind.txt --tool=memcheck ncbo --mmr_cln -O -v one_dmn_rec_var ~/nco/data/in.nc ~/foo_avg.nc ~/foo.nc > ~/foo 2>&1
nco757. Make sure above ncbo bug is not in ncflint
nco759. Document --mmr_cln flag
nco761. Merge timer srt/end calls to MPI codes
nco763. Access IPCC AR4 database via OPeNDAP: http://www-pcmdi.llnl.gov/software-portal/esg_data_portal/dapserver
	Sample datasets look like:
        http://climate.llnl.gov/cgi-bin/dap-cgi.py/ipcc4/sresa1b/ncar_ccsm3_0/pcmdi.ipcc4.ncar_ccsm3_0.sresa1b.run1.atm.mo.xml.html
        ncks -M -p http://zender:password@climate.llnl.gov/cgi-bin/dap-cgi.py/ipcc4/sresa1b/ncar_ccsm3_0 pcmdi.ipcc4.ncar_ccsm3_0.sresa1b.run1.atm.mo.xml
	ncks -M -p http://dust.ess.uci.edu/cgi-bin/dods/nph-dods/dodsdata in.nc
nco768. Change some common flags from toggles to hard-coded true/false
        to avoid weird behavior when toggle switches accidentally passed twice,
        e.g., ncks -O -O in.nc foo.nc
nco772. ncpdq fails MPI regression tests on pbs, possibly others
nco776. Minor, fixable memory leaks in ncecat, fix with valgrind
nco777. ncra regression on ipcc/pathscale (only) w/w/o mmr_drt with > 1 thread.
	May indicate problem with timer behavior in OpenMP code.
	ncra -O -t 1 --mmr_drt -D 1 -v one_dmn_rec_var -p ~/nco/data in.nc in.nc ~/foo.nc # works fine
	ncra -O -t 2 --mmr_drt -D 1 -v one_dmn_rec_var -p ~/nco/data in.nc in.nc ~/foo.nc # fails
	cra: INFO omp_set_num_threads() used to set execution environment to spawn teams of 2 threads
	ncra: CONVENTION File "Conventions" attribute is "CF-1.0"
	ncra: TIMER Metadata setup and file layout before main loop took    0.00 s
	ncra: INFO Input file 0 is /home/zender/nco/data/in.nc
	ncra: INFO Input file 1 is /home/zender/nco/data/in.nc
	ncra: INFO Moving /home/zender/foo.nc.pid7069.ncra.tmp to /home/zender/foo.nc...done
	**** Segmentation fault!  Fault address: 0x20
	ncra: TIMER Wallclock-elapsed time for command is    0.00 s
nco778. ncrename should allow renaming dimensions to existing dimensions so long as dimension size does not change. Problem is that netCDF nc_rename_dim() disallows this.
	ncks -O -v lev,lat_grd -p ~/nco/data in.nc ~/foo.nc
	ncrename -O -d lev,lat_grd ~/foo.nc ~/foo.nc
nco779. Add non-record stride checks to regression tests e.g.,
	ncks -O -v three_dmn_rec_var -d lon,,,2 -p ~/nco/data in.nc ~/foo.nc
	ncrcat -O -v three_dmn_rec_var -d lon,,,2 -p ~/nco/data in.nc in.nc ~/foo.nc
	ncks -C -v three_dmn_rec_var ~/foo.nc
nco781. Add ncks option to suppress creating temporary output file in -A mode, simply copy input file first? Seems unlikely this will save time.
nco784. Build NCO on NMI test lab http://nmi.cs.wisc.edu/
nco789. Add regression test for behavior when _FillValue defined and missing_value not	      
	ncwa -O -a lon -v fll_val,mss_val ~/nco/data/in.nc ~/foo.nc
        ncks -H -C ~/foo.nc
nco794. Automagically use missing_value when _FillValue not present
nco796. pgCC does not support exception handling with codes compiled with OpenMP (-mp) switch so ncap2/antlr files must be built serially
nco799. Extend -n 3,4,1 NINTAP switch to handle regular expressions
nco800. Extend -n 3,4,1 NINTAP switch to -n 3,4,1,offset to work with CCSM patterns
nco803. Link to UCSD proposal by Kraig Winters /data/zender/prp/prp_winters.pdf
nco807. Merge ncra.c v.1.192-1.193 changes (per-file record limit checks) into mpncra.c
nco808. Dave Allured request on 20070213 for ncea to resolve hyperslab indices for all dimensions on each file so this works:
        ncea -d time,"2010-1-1 0:0:0","2020-12-31 0:0:0" run*.nc output.nc
        ncea -O -d time,"2010-1-1 0:0:0","2020-12-31 0:0:0" ~/giss*.nc ~/output.nc
	ncea -O -D 5 -d time,1.,5. -p ~/nco/data in.nc in.nc ~/foo.nc
nco809. Implement --quiet option to suppress WARNINGs and INFOs
nco814. Add regression test for forgiving exclusion list handling
        ncks -O -D 2 -x -v not_present ~/nco/data/in.nc ~/foo.nc
nco816. Add regression tests for rmssdn when N=0, N=1, N=2
	ncra -O -y rmssdn -v '^rec_var_flt_mss_val_flt_all.?' ~/nco/data/in.nc ~/foo.nc
	ncks -H -m ~/foo.nc
nco817. Document how to work with 1D/2D coordinate grids
	ncks -H -v '.?_[0-9]D.?' ~/nco/data/in.nc
nco821. ncrename netCDF4 regression
	cd ~/nco/bm;./nco_bm.pl --fl_fmt=netcdf4 --regress --udpreport ncks
	ncks -h -O -C --fl_fmt=netcdf4 -v lat ~/nco/data/in.nc ~/foo.nc
	ncks -O -C --fl_fmt=classic -v lat ~/nco/data/in.nc ~/foo.nc
	ncrename -O -D 1 -d lat,new_lat -v lat,new_lat ~/foo.nc # works
	ncks -O -C --fl_fmt=netcdf4 -v lat ~/nco/data/in.nc ~/foo.nc
	ncrename -O -D 1 -d lat,new_lat -v lat,new_lat ~/foo.nc # fails
nco822. ncap2 netCDF4 regressions with nco_bm.pl:
	cd ~/nco/bm;./nco_bm.pl --fl_fmt=netcdf4 --regress --udpreport ncap2
	ncap2 -O --fl_fmt=netcdf4 -D 5 -v -S ~/nco/data/ncap2.in ~/nco/data/in.nc ~/foo.nc # fails, causes ncap2 regression with nco_bm.pl
	Script dies with:
	ncap2: DEBUG ncap_def_ntl_scn(): i3 - defined in output.
	ERROR: program exiting through nco_err_exit() which will now call abort()
	nco_err_exit(): ERROR nco_enddef
	HDF error
	Aborted (core dumped)
nco826. ncwa optimization: do not allocate tally when var_sz=var_avg since this this always indicates a no-op, e.g., removing a degenerate dimension.
nco828. ncks: returns file ID = 65536 on netCDF4 files: why?
	ncks -O --fl_fmt=netcdf4 ~/nco/data/in.nc ~/nco/data/in_4.nc
        ncks -m ~/nco/data/in_4.nc
	od -An -c -N4 ~/nco/data/in_4.nc
nco834. Add atomic type string NC_STRING to nco_c++ get/put var
nco835. ncks uint64_var(ULLONG_MAX) = int64_var(LLONG_MAX) (problem with libnetcdf4?) PBN4
	ccc
	ncks -v '.?_var$' ~/c++/foo.nc
nco839. What, if anything, should NC_FILL_STRING be?
nco845. Add -Wno-switch to quiet GCC enum warnings on netCDF4 case labels?
nco846. netCDF4: switch to long long array offsets everywhere? 
        e.g., can long tally hold > 2 GB array?
nco847. switch sz, tally to type (unsigned long), perhaps size_t, and perhaps (unsigned long long)
nco848. Add regression to demonstrate breakage on variables with 2--4 GB elements
nco853. Enable arithmetic for NC_BYTE and NC_UBYTE everywhere?
nco854. Use NC_BYTE instead of NC_SHORT for logical in ncap2?
nco855. nco_netcdf.c netCDF4 forward compatibility stubs generate beaucoups 'unused parameter' warnings (only when compiled with netCDF3)
nco865. Re-cast output variable to original type in ncbo.c (like in ncflint.c)?
        ncbo -O -v '.?pck.?' -p ~/nco/data in.nc in.nc ~/foo.nc
        ncbo -O -v rec_var_dbl_mss_val_dbl_pck -p ~/nco/data in.nc in.nc ~/foo.nc
        ncks -O -C -d time,0,3 -v rec_var_dbl_mss_val_dbl_pck -p ~/nco/data in.nc
	ncks ~/foo.nc
nco866. Why don't ncra and ncwa make NC_BYTE etc. as fix var in var_lst_dvd()?
nco871. Add compression regression test
	ncks -O --dfl_lvl=0 ~/nco/data/in.nc ~/foo.nc
	ls -l ~/foo.nc
	ncks -O --dfl_lvl=9 ~/nco/data/in.nc ~/foo.nc
	ls -l ~/foo.nc
	ncks -O -4 --dfl_lvl=0 ~/nco/data/in.nc ~/foo.nc
	ls -l ~/foo.nc
	ncks -O -4 --dfl_lvl=9 ~/nco/data/in.nc ~/foo.nc
	ls -l ~/foo.nc
	ncks -O -4 --dfl_lvl=9 -v one ~/nco/data/in.nc ~/foo.nc
nco873. ncks remik bug: ncks BUG: coords are zero after 2 var ext 
	intel compiler only, possibly only ia64
	"superficially injecting "nc_sync(out_id);"
	into ncks.c:559 at the end of the "Copy variable data" for-loop 
        the output is correct, although this isn't a solution we'd want.
nco874. Implement CDL abbreviations for netCDF4 atomic types (when known)
nco875. Squelch compiler warnings like: 
	../src/nco/nco_pck.c:819:39: warning: integer constant is so large that it is unsigned
	that are triggered by OK statements like
	case NC_UINT64: mss_val_dfl_dbl=NC_FILL_UINT64; break;
nco880. ncpdq: netCDF4 problem _only_ with fll_val/mss_val variables: PBN4
	ncpdq -O                     -D 4 ~/nco/data/in.nc ~/foo.nc # works fine
	ncpdq -O       -C -v fll_val -D 4 ~/nco/data/in.nc ~/foo.nc # works fine
	ncpdq -O -4 -x    -v fll_val -D 4 ~/nco/data/in.nc ~/foo.nc # borken
	ncpdq -O -4    -C -v fll_val -D 4 ~/nco/data/in.nc ~/foo.nc # borken
	nco_err_exit(): ERROR Short NCO-generated message (usually name of function that triggered error): nco_copy_att()
	nco_err_exit(): ERROR Error code is -36. Translation into English with nc_strerror(-36) is "NetCDF: Invalid argument"
nco885. ncpdq: netCDF4 problem with CAM datasets variable AERASM_v. AERASM_v does have a _FillValue attribute---is this a clue?
	ncpdq -D 6 -4 -v AERASM_v -O sncpd05_ts_ANN.nc -p $DATA/anl_sncpd05 ~/foo_5.nc
	nco_err_exit(): ERROR Short NCO-generated message (usually name of function that triggered error): nco_copy_att()
	nco_err_exit(): ERROR Error code is -36. Translation into English with nc_strerror(-36) is "NetCDF: Invalid argument"
nco886. Decide next project: SWAMP? paper on...ncap2? SWAMP? IPCC benchmarks?
nco891. dbn: fix lintian RPATH messages like
	Binary /usr/bin/ncwa compiled with an RPATH of /usr/lib/nco
	libtool error: libtool --finish /usr/lib
nco895. udunits2 implement new API http://www.unidata.ucar.edu/software/udunits/udunits-2/udunits2lib.html
	/bin/rm ~/obj/LINUX/nco_lmt.o;cd ~/nco/bld;make OPTS=D UDUNITS2=Y NETCDF4=Y allinone
	ncks -O -D 0 -C -H -v wvl -d wvl,'0.4 micron','0.7 micron' -s '%3.1e' ~/nco/data/in.nc
nco900. Get UDUnits2 working with time coordinates:
        ncks -O -D 0 -C -d time_udunits,'1999-12-08 12:00:0.0','1999-12-09 00:00:0.0' -p ~/nco/data in.nc ~/foo.nc
nco901. mss_val in code and comments now (usually) refers to _FillValue (except in nco_mss_val.c). Change mss_val to fll_val?
nco905. TKN2SNG technique for "_FillValue" adds extra quotes in nco.h but not c.c! The work-around is an extra function of global scope (nco_mss_val_sng_get()) that is worth getting rid of.
nco906. Add regression test for DAP via THREDDS server:
	ncks -m http://eclipse.ncdc.noaa.gov:8080/thredds/dodsC/isccp-d1/2001/ISCCP.D1.1.GLOBAL.2001.01.01.0000.GPC
nco907. doc: document new wget capabilities
nco909. Does autoconf for netCDF4-enabled NCO require additional -lm flag?
nco911. Periods in attribute names choke OPeNDAP (had to remove att_nm.dot):
	ncks -O -D 7 -v var_nm.dot -p ~/nco/data dap_dot_bug.nc ~/foo.nc # works fine
	ncks -O -D 7 -v var_nm.dot -p http://dust.ess.uci.edu/cgi-bin/dods/nph-dods/dodsdata dap_dot_bug.nc ~/foo.nc # borken
	nco_err_exit(): ERROR Short NCO-generated message (usually name of function that triggered error): nco_copy_att()
	nco_err_exit(): ERROR Error code is -43. Translation into English with nc_strerror(-43) is "Attribute not found"
nco912. ncks: print '_' or '_FillValue' instead of numeric values a la ncdump?
nco913. nco_cmp_get() fails on _AIX with different compiler
nco915. GCC OpenMP support in configure.in
nco917. GCC OpenMP with autoconf
nco919. 'make check' (i.e., nco_bm.pl) breaks when ncap2 disabled (i.e., --disable-nco++) reported by Tim Hoar
nco920. Add ncwa (ncks?) capability to remove all degenerate dimensions --delete-degenerate-dimensions (ncl-talk 12/5/07) (assign to hmb?)
nco922. ncbo with OpenMP on netCDF4 files fails (not with netCDF3 in.nc):
	ncbo -O -t 1 -D 4 -p ${HOME}/nco/data in.nc in.nc ~/foo.nc # marche
	ncbo -O -t 2 -D 4 -p ${HOME}/nco/data in.nc in.nc ~/foo.nc # march
	ncbo -O -t 1 -D 4 -p ${HOME}/nco/data in_4.nc in_4.nc ~/foo.nc # marche
	ncbo -O -t 2 -D 4 -p ${HOME}/nco/data in_4.nc in_4.nc ~/foo.nc # fails
	ncbo -O -t 1 -D 4 -p ${HOME}/aca swnb.nc swnb.nc ~/foo.nc # marche
	ncbo -O -t 2 -D 4 -p ${HOME}/aca swnb.nc swnb.nc ~/foo.nc # fails
nco923. OPTS=D and OPTS=R should not do -O4 optimization on nco++
nco924. Document ncks --msa_usr_rdr
nco929. drop ncap, or, rather, link it to ncap2, next release
nco930. ERROR: ld.so: object 'libfakeroot-sysv.so' from LD_PRELOAD cannot be preloaded: ignored.
nco931. nco_c++ nc_put_var1_string wrapper currently kludged and untested
nco932. ncwa and packed data: help forum 20080214: When using ncwa to calculate a weighted average from a packed file (NC_SHORT, scale_factor=0.1, _FillValue=32766), the data in the resulting file is in double precision. However, the _FillValue attribute is not changed to the double precision value (3276.6)....it is left at 32766. Shouldn't ncwa writeout the new _FillValue attribute?
nco934. Add ncecat -M option to other operators?
nco935. Make single-file operators (SFOs) default out.nc to in.nc
nco936. Add -X to all MSA operators
nco937. 
************************************************************************
End ncoXXX TODOs
************************************************************************

************************************************************************
Begin valgrind section
************************************************************************
Valgrind: Archive of valgrind tests which are memory-error and leak-free
vg1. ncap
   valgrind --leak-check=yes --show-reachable=yes --suppressions=${HOME}/nco/doc/valgrind.txt --tool=memcheck ncap -O -S ${HOME}/nco/data/ncap.in -p ${HOME}/nco/data -l /tmp in.nc ~/foo.nc > ~/foo 2>&1
     ncap2
   valgrind --leak-check=yes --show-reachable=yes --suppressions=${HOME}/nco/doc/valgrind.txt --tool=memcheck ncap2 -O -D 1 -S ~/nco/data/ncap2.in ~/nco/data/in.nc ~/foo.nc > ~/foo 2>&1
vg2. ncbo
   valgrind --leak-check=yes --show-reachable=yes --suppressions=${HOME}/nco/doc/valgrind.txt --tool=memcheck ncbo -t 2 -O -C -d lon,1 -p ${HOME}/nco/data -l /tmp in.nc in.nc ~/foo.nc > ~/foo 2>&1
vg3. ncea
   valgrind --leak-check=yes --show-reachable=yes --suppressions=${HOME}/nco/doc/valgrind.txt --tool=memcheck ncea -t 2 -O -C -D 3 -d lon,1 -n 3,4,1 -v one -p ${HOME}/nco/data -l ${HOME}/nco/data h0001.nc ~/foo.nc > ~/foo 2>&1
vg4. ncecat
   valgrind --leak-check=yes --show-reachable=yes --suppressions=${HOME}/nco/doc/valgrind.txt --tool=memcheck ncecat -t 2 -O -C -D 3 -d lon,1 -n 3,4,1 -p ${HOME}/nco/data -l ${HOME}/nco/data h0001.nc ~/foo.nc > ~/foo 2>&1
vg5. ncflint
   valgrind --leak-check=yes --show-reachable=yes --suppressions=${HOME}/nco/doc/valgrind.txt --tool=memcheck ncflint -t 2 -O -C -d lon,1 -p ${HOME}/nco/data -l /tmp in.nc in.nc ~/foo.nc > ~/foo 2>&1
vg6. ncks
   valgrind --leak-check=yes --show-reachable=yes --suppressions=${HOME}/nco/doc/valgrind.txt --tool=memcheck ncks -O -C -d lat,1,1 -d lev,2,2 -d lon,0,3 -d lon,1,3 -p ${HOME}/nco/data -l /tmp in.nc ~/foo.nc > ~/foo 2>&1
vg7. ncpdq (re-order)
   valgrind --leak-check=yes --show-reachable=yes --suppressions=${HOME}/nco/doc/valgrind.txt --tool=memcheck ncpdq -t 2 -O -D 3 -a lat,lev,lon -v three_dmn_var ~/nco/data/in.nc ~/foo.nc > ~/foo 2>&1
vg8. ncpdq (pack)
   valgrind --leak-check=yes --show-reachable=yes --suppressions=${HOME}/nco/doc/valgrind.txt --tool=memcheck ncpdq -t 2 -O -D 3 -P all_new -v three_dmn_var ~/nco/data/in.nc ~/foo.nc > ~/foo 2>&1
vg9. ncra
   valgrind --leak-check=yes --show-reachable=yes --suppressions=${HOME}/nco/doc/valgrind.txt --tool=memcheck ncra -t 2 -O -C -D 3 -d lon,1 -n 3,4,1 -v one_dmn_rec_var_sng -p ${HOME}/nco/data -l ${HOME}/nco/data h0001.nc ~/foo.nc > ~/foo 2>&1
vg10. ncrcat
    valgrind --leak-check=yes --show-reachable=yes --suppressions=${HOME}/nco/doc/valgrind.txt --tool=memcheck ncrcat -t 2 -O -C -D 3 -d lon,1 -n 3,4,1 -v one_dmn_rec_var_sng -p ${HOME}/nco/data -l ${HOME}/nco/data h0001.nc ~/foo.nc > ~/foo 2>&1
vg11. ncrename
    valgrind --leak-check=yes --show-reachable=yes --suppressions=${HOME}/nco/doc/valgrind.txt --tool=memcheck ncrename -O -D 3 -d lon,new_lon -v scalar_var,new_scalar_var -a long_name,new_long_name -p ${HOME}/nco/data -l /tmp in.nc ~/foo.nc > ~/foo 2>&1
vg12. ncwa
    valgrind --leak-check=yes --show-reachable=yes --suppressions=${HOME}/nco/doc/valgrind.txt --tool=memcheck ncwa -D 3 -t 2 -O -C -a lat,lon,time -w gw -m gw -v PS -p ${HOME}/nco/data -l /tmp in.nc ~/foo.nc > ~/foo 2>&1
    callgrind ncwa -D 3 -t 2 -O -C -w gw -m gw -p ${HOME}/nco/data -l /tmp in.nc ~/foo.nc > ~/foo 2>&1
    valgrind --suppressions=${HOME}/nco/doc/valgrind.txt --tool=callgrind ncwa -D 3 -t 2 -O -C -a lat,lon,time -w gw -m gw -v PS -p ${HOME}/nco/data -l /tmp in.nc ~/foo.nc > ~/foo 2>&1
    kcachegrind ncwa &
************************************************************************
End valgrind section
************************************************************************

************************************************************************
Begin MPI section
************************************************************************
MPI: Archive of quickie MPI tests
NB: mpirun requires explicit path to executable
Status 20050920: 
1. LAM-MPI, MPICH, MPICH2 all work fine
   Currently recommend running with MPICH2, though LAM is easier since it's in Ubuntu
   See ~/doc/mpi.txt for installation details.
2. Charlie's default compilation/testing environments:
   ashes: Ubuntu, x86, LAM-MPI, gcc-4.0
   elnino: Libranet, x86, MPICH2, icc-9.0
   sand: Ubuntu, dual x86_64, LAM-MPI, gcc-4.0
   soot: Fedora Core 3, dual x86, LAM-MPI, icc-8.0

0. MPI with valgrind example
   mpirun -np 3 /usr/bin/valgrind --leak-check=yes --show-reachable=yes --suppressions=${HOME}/nco/doc/valgrind.txt --tool=memcheck mpirun -np 3 ${MY_BIN_DIR}/mpncbo -O -D 3 -p ${HOME}/nco/data in.nc in.nc ~/foo.nc > ~/foo 2>&1
1. mpncbo
   mpirun -np 3 ${MY_BIN_DIR}/mpncbo -O -D 3 -p ${HOME}/nco/data in.nc in.nc ~/foo.nc > ~/foo 2>&1
2. mpncea
   mpirun -np 3 ${MY_BIN_DIR}/mpncea -O -D 3 -p ${HOME}/nco/data in.nc in.nc ~/foo.nc > ~/foo 2>&1
3. mpncecat
   mpirun -np 3 ${MY_BIN_DIR}/mpncecat -O -D 3 -p ${HOME}/nco/data in.nc in.nc ~/foo.nc > ~/foo 2>&1
4. mpncpdq
   mpirun -np 3 ${MY_BIN_DIR}/mpncpdq -O -D 3 -a lat,lev,lon -p ${HOME}/nco/data in.nc ~/foo.nc > ~/foo 2>&1
   mpirun -np 3 ${MY_BIN_DIR}/mpncpdq -O -D 3 -P all_new -p ${HOME}/nco/data in.nc ~/foo.nc > ~/foo 2>&1
5. mpncra
   mpirun -np 3 ${MY_BIN_DIR}/mpncra -O -D 3 -p ${HOME}/nco/data in.nc in.nc ~/foo.nc > ~/foo 2>&1
6. mpncrcat (hangs just before write)
   mpirun -np 3 ${MY_BIN_DIR}/mpncrcat -O -D 3 -p ${HOME}/nco/data in.nc in.nc ~/foo.nc > ~/foo 2>&1
7. mpncwa
   mpirun -np 3 ${MY_BIN_DIR}/mpncwa -O -D 3 -p ${HOME}/nco/data in.nc ~/foo.nc > ~/foo 2>&1
************************************************************************
End MPI section
************************************************************************

************************************************************************
Begin OPeNDAP section
************************************************************************
OPeNDAP: Archive of quickie DAP tests:
ncks -p http://dust.ess.uci.edu/cgi-bin/dods/nph-dods/dodsdata -l . in.nc
ncks -p http://soot.ess.uci.edu/cgi-bin/dods/nph-dods/dodsdata -l . in.nc
ncks -p https://srb.npaci.edu/mysrb331all.shtml/home/zender.uci-esmf/zender/nco/data -l . in.nc
http://sasa.sdsc.edu:8003/cgi-bin/nph-dods/data/nccoos-duck-codar-latest.nc.html
http://orion.sdsc.edu:8003/cgi-bin/nph-dods/data/nccoos-duck-codar-latest.nc.html
************************************************************************
End OPeNDAP section
************************************************************************

************************************************************************
Begin hpctoolkit section
************************************************************************
/bin/rm -r ~/nco/bm/hpctoolkit # Remove old profiles
mkdir -p ~/nco/bm/hpctoolkit
lsmod | grep oprofile # Make sure oprofile module is not loaded
sudo opcontrol --deinit # Unload oprofile iff it is loaded
modprobe perfctr # Load perfctr module
# Do not monitor more than one function per hardware counter, e.g., PAPI_FP_OPS and PAPI_FP_OPS

cat > ~/nco/bm/prf_ncbo.sh << EOF
cd ~/nco/bm/hpctoolkit
hpcrun -r -o ~/nco/bm/hpctoolkit -e PAPI_TOT_CYC:32767 -e PAPI_FP_OPS:32767 -e PAPI_TOT_INS:32767 -e PAPI_L2_DCM:32767 -- ${MY_BIN_DIR}/ncbo -O --op_typ='-' -p ${DATA}/nco_bm stl_5km.nc stl_5km.nc ${DATA}/nco_bm/prf_ncbo.nc
EOF
chmod a+x ~/nco/bm/prf_ncbo.sh
~/nco/bm/prf_ncbo.sh

hpcquick -I ~/nco/src/nco -P ~/nco/bm/hpctoolkit/ncbo.PAPI_TOT_CYC-etc.clay.ess.uci.edu.14169.0
hpcviewer ~/nco/bm/hpctoolkit/hpcquick.db/hpcquick.hpcviewer &
************************************************************************
End hpctoolkit section
************************************************************************

************************************************************************
Begin oprofile section
************************************************************************
cd ~/nco/bm
xosview &
# To enable idle cycle counts, add 'idle=poll' to kernel spec in /boot/grub/menu.lst:
# kernel          /boot/vmlinuz.old root=/dev/hda1 ro console=tty0 idle=poll quiet splash
sudo opcontrol --vmlinux=/usr/src/linux-2.6.11/vmlinux # 
sudo opcontrol --separate=cpu # Use this on SMP machines
sudo opcontrol --start # Start profiling
sudo opcontrol --shutdown

cat > ~/nco/bm/prf_ncbo.sh << EOF
cd ${DATA}/nco_bm
sudo opcontrol --reset
time ${MY_BIN_DIR}/ncbo -O --op_typ='-' -p ${DATA}/nco_bm gcm_T85.nc gcm_T85_00.nc ${DATA}/nco_bm/prf_ncbo.nc
opreport --demangle=smart --exclude-dependent --symbols > ~/nco/bm/prf_ncbo.txt
EOF
chmod a+x ~/nco/bm/prf_ncbo.sh
~/nco/bm/prf_ncbo.sh

cat > ~/nco/bm/prf_ncwa.sh << EOF
cd ${DATA}/nco_bm
sudo opcontrol --reset
time ${MY_BIN_DIR}/ncwa -O -a lat,lon,lev,time -w lat -p ${DATA}/nco_bm gcm_T85.nc ${DATA}/nco_bm/prf_ncwa.nc
opreport --demangle=smart --exclude-dependent --symbols > ~/nco/bm/prf_ncwa_avgall.txt
EOF
chmod a+x ~/nco/bm/prf_ncwa.sh
~/nco/bm/prf_ncwa.sh

cat > ~/nco/bm/prf_ncwa.sh << EOF
cd ${DATA}/nco_bm
sudo opcontrol --reset
time ${MY_BIN_DIR}/ncwa -O -a lat,lon,lev,time -w lat -z "lat!=1.0" -p ${DATA}/nco_bm gcm_T85.nc ${DATA}/nco_bm/prf_ncwa.nc
opreport --demangle=smart --exclude-dependent --symbols > ~/nco/bm/prf_ncwa_avgall_msk.txt
EOF
chmod a+x ~/nco/bm/prf_ncwa.sh
~/nco/bm/prf_ncwa.sh

cat > ~/nco/bm/prf_ncwa.sh << EOF
cd ${DATA}/nco_bm
sudo opcontrol --reset
time ${MY_BIN_DIR}/ncwa -O -F -d lat,,64 -d lon,,128 -d lev,,16 -d time,,4 -a lat,lon,lev,time -w lat -p ${DATA}/nco_bm gcm_T85.nc ${DATA}/nco_bm/prf_ncwa.nc
opreport --demangle=smart --exclude-dependent --symbols > ~/nco/bm/prf_ncwa_hlf.txt
EOF
chmod a+x ~/nco/bm/prf_ncwa.sh
~/nco/bm/prf_ncwa.sh
************************************************************************
End oprofile section
************************************************************************

************************************************************************
Begin ncapXXX TODOs
************************************************************************
ncap-specific items:
ncap07. Combine binary operations into binary_op_var_var and binary_op_var_att types to reduce number of functions
ncap09. Add outer loop over files so script may apply to each file
ncap12. Ensure lexer does not perform un-necessary work during initial_scan (e.g., LHS casting...)
ncap13. Test that LHS casting works regardless of value of LHS template. If true, remove value initialization block
ncap15. Link to GSL functions, especially gsl_sf_* (special functions)
ncap16. Get ncap build working with pure_parser turned off so lex/yacc builds work
ncap17. Start in console mode if no script and no output file specified, e.g., nco> 
ncap19. Make all binary operators handle ordering of operands consistently, e.g., divide(1,2)=1/2 not 2/1
ncap20. Print out contents of each non-comment statement when dbg_lvl==1
ncap21. Add two-argument functions like fmodf to ncap_sym_init() (using va_args?)
ncap25. Turn arbitrarily fixed size array tokens NCAP_ATT_LST_NBR_MAX, NCAP_SPT_NBR_MAX into dynamically allocatable arrays
ncap26. Make sym_tbl_nbr dynamic by implementing sym_add(), sym_rm() functions or linked list
ncap28. Make provision for pack() and unpack() to copy all attributes except scale_factor and add_offset from disk into new variable on disk?
ncap29. Rewrite nco_var_pck() to use var_scv() arithmetic operations rather than more complex var_var() operations
ncap30. Allow -S to specify multiple input scripts to be sequentially parsed
ncap31. Put globals fl_spt_crr,ncl_dpt_crr,ln_nbr_crr in prs_arg or remove fl_spt from it
ncap32. Need filename expander and $NCO_DATA_PATH handling so #include's may be generic
ncap34. Make parse errors always print current line to terminal rather than silently failing
ncap36. cygwin builds of ncap.l and netcdf fail because of dependence on syslimits.h. non-kludge fix may require updated gcc and mingw packages
ncap37. Implement float function definitions in nco_rth_flt.c as inline to improve speed for MACOSX builds (inline is OK in C99 but will it break C89?) using AC_C_INLINE
ncap38. Finish dimension-reducing operations (min/max/avg)
ncap41. Make ncap function prototypes consistent with rest of NCO (ncap.h mainly)
ncap42. Add switch allowing derived variable to take on all attributes of specified variable
ncap43. Alphabetize ncap function list prior to printing it with -f
ncap46. Define YYMALLOC/FREE/REALLOC etc so lexer/parser use nco_malloc() etc.
ncap48. Search ncap_yacc.y for non-symmetric arithmetic operations, implement symmetric operations
ncap49. Cause "**" in arithmetic context to cause error, print HINT to Fortran syntax users
ncap52. Combine pow() with ^ parsing in parser ncap_yacc.y 
ncap53. Bug in ncap pack() on Alpha/Tru64: ncap -O -D 10 -s 'hsig=pack(hsig)' foo.nc foo2.nc details on sourceforge
ncap54. Add option (true by default?) to retain variable attributes when variable re-defined or defined in terms of single other variable or perhaps always retain attributes of left-most variable on RHS of definition?
        ncap -v -O -s "one=one*1" -s "two=one*2" -s "pck=pack(upk)" ~/nco/data/in.nc ~/foo.nc
        ncks -m -H -v one,two,pck ~/foo.nc
ncap55. Add pre-defined named_constants, e.g., nco_pi, nco_e, to make math easier
ncap56. Introduce way of making function synonyms, e.g., ln=log, without growing symbol table?
ncap57. Raising transcendental functions of NC_FLOATs to powers fails on AIX
AIX thinks it has the trancendental float functions, so NCO uses system-supplied versions.
Symptoms are consistent with breakage of system-supplied logf, sinf, etc.
# Fails on AIX
ncap -D 1 -O -v -s "foo=log(e_flt)^1" ~/nco/data/in.nc ~/foo.nc
ncap -D 1 -O -v -s "foo=sin(e_flt)^1" ~/nco/data/in.nc ~/foo.nc
ncap -D 1 -O -v -s "foo=log(e_flt+0.01f)^1" ~/nco/data/in.nc ~/foo.nc
# Works on AIX
ncap -D 1 -O -v -s "foo=log(e_dbl)^1" ~/nco/data/in.nc ~/foo.nc
ncap -D 1 -O -v -s "foo=sin(e_dbl)^1" ~/nco/data/in.nc ~/foo.nc
ncap59. ncap always gives points in first quadrant for
ncap -O -v -s 'water_dir=(180/4*atan(1))*(atan(water_v/water_u) + 4 * atan(1) * ((water_u-abs(water_u))/(-2*water_u)))'
ncap60. Add capability to create character string variables that contain any string, given only a pre-existing dimension size capable of holding the string
ncap66. make ncap errors fatal, either in yyerror(), or call nco_error_exit()
ncap70. gamma() function broken on AIX
ncap71. Remove erf and gamma functions from AIX build
ncap72. Stop linking to -lC on AIX. Get all float math functions by casting double versions
ncap79. make abs() synonym for fabs() and pow() and ** synonyms for ^
ncap80. ncap2+ openmp + netcdf4 - script only works with threads>1 if there are no atts in input file
************************************************************************
End ncapXXX TODOs
************************************************************************

************************************************************************
Begin bldXXX TODOs
************************************************************************
Build items:
bld14. Replace #ifdef tokens in NCO code with autoconf-style tokens, e.g., #ifdef HDF5 -> #ifdef HAVE_HDF5, etc. Other tokens to consider are WIN32, _OPENMP, USE_FORTRAN_ARITHMETIC (deprecated). Of course must modify bld/Makefile to pass these style tokens as well so that config.h is not _required_, and that token use is consistent.
bld23. Implement ncra i18n support I18N with autotools: Add .po files, update doc/i18n.txt
bld30. Search for libstdc++ with AIX/DODS/GCC breaks NCO build
bld32. Why have --enable-shared and --enable-static both default true? override one?
bld33. Use PACKAGE_BUGREPORT, PACKAGE_NAME, PACKAGE_STRING, PACKAGE_TARNAME, PACKAGE_VERSION?
bld35. Add pgcc, pgCC switches to configure.in 
bld36. Improve precision/wording of nco_ctl.c tokens on shared/static libraries
bld37. Ensure C99 used by compilers with AX_CFLAGS_SUN_OPTION AX_CFLAGS_HPUX_OPTION AX_CFLAGS_AIX_OPTION AX_CFLAGS_IRIX_OPTION
bld38. Use OPeNDAP opendap-config script to ease build process, e.g., `opendap-config --cxx --cflags --libs` -o sample sample.c
bld39. Fix autoconf build warnings (from including ncap_yacc in ncwa?):
       Makefile:472: warning: overriding commands for target `ncap_yacc.h'
       Makefile:440: warning: ignoring old commands for target `ncap_yacc.h'
bld40. Automate splicing trailer ~/nco/doc/man_end.txt to man pages
bld41. Add configure check for libssl functionality, if not found (Fedora) to explicitly adding libssl
bld42. Adopt orphaned Debian NCO http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=273322
bld43. Implement Martin Schmidt's suggestion for AIX autoconf builds posted to developer's forum 20050201: {$CC} is not tested for xlc* but for (arbitrary_path/xlc*)
bld44. Add support for autopackage format? http://www.wildgardenseed.com/apkg/
bld45. Replace ENABLE_LARGEFILE with test for 64-bit offset support (i.e., netCDF 3.6) in configure.in and nco_ctl.c
bld47. Add regression test for 64bit file format capability
bld61. Verify netCDF4 benchmarks work for Ed Hartnett
bld70. hjm nco_c++ on Mac OS X: Get "make check" working (add -lstdc++, remove -lnco)
bld74. hjm Do not pass -std=c99 unless GCC=gcc rather than g++
bld76. Add SMP support switches to configure.in for pathcc, pgicc, icc
bld77. Make sure C99 switches like -std=c99 do not get passed to CXX
bld78. Fix CXXFLAGS for icc enable-custom-optimize
bld79. 
************************************************************************
End bldXXX TODOs
************************************************************************

************************************************************************
Begin bnchXXX TODOs - benchmarking & analysis/presentation of same
************************************************************************
Benchmark items:
bnch01. hjm make benchmark script return remote benchmark data as netcdf 
	data, or re-write server side to convert returned data into netcdf
	variables and append to a netcdf file.
bnch02. hjm add code to analysis/plotting script (at least until bnch01 is
	ready) to filter data points that vary more than 1 SD from mean to remove
	bad data.
bnch03. hjm should plotting script use linear regression or a log()
	transform to analyze the relationship?  Check with other MPI analyses.

************************************************************************
End bnchXXX TODOs
************************************************************************

************************************************************************
Begin dlwXXX plan
************************************************************************
1. one or two operator complete server side DAP script server side only
2. CVS. write client wrapper script which parses regular command
   line(s) and feeds complex script to server (shell)
         
    wrapper_script ncra -n 100,1,2 model_01.nc model_avg.nc

3. Complete SS CGI script to work with all operators 
4. Complete client side to work with all operators
5. Make sure wildcards/globbing works

Milestone: regression tests can work in nco_bm.pl
           Check if design decisions/interfaces seem good
           ncra -n 100,1,2 -p http://server/foo/model_01.nc model_avg.nc
Goal: end of February

6. Multiple sequential commands with SS intermediate output files and one
   output file to client
7. SS intermediate filename magic allow deterministic chaining of
   output files so sequential command may depend on previous commands
8. SS and client side mods to send back multiple output files
8.1 db-backed job control
8.1a. client-side multi-file wrapper first iteration
8.1b. server-side file-waiting (block-until-send) first iteration
8.1c. client-side real async: job status, etc.
8.1d. server-side wakeup/signalling for scalability


Milestone: Mark's climate model analysis script "works"
           Check if design decisions/interfaces/initial parser seems good 
           
9. SS understand dependencies (i.e., dependence of commands within script)
10. Smarter scheduling of operations (i.e., non-sequential script execution)
11. Basic security: file path validator (configurable?)
12. Basic security: argument validator (watch for backticks, pipes and other shell constructs)
13. Docs about security concerns
14. Charlie xml filefetch test.

Random cleanup items:
1. Rework/consolidate tests in src/ssdap for less redundancy, less clutter, 
   more simplicity, and understandability.  
2. Create simple SSDAP functionality test so a user can verify the ability to 
   use SSDAP scripts. Want to call this script from the bm script during 
   regression testing.
3. Modularize parser (clear place to add new operators)
4. Modularize executor (clear interface to build new frontend)

************************************************************************
End dlwXXX plan
************************************************************************


************************************************************************
End NCO Wish list
************************************************************************