Changelog for LHAPDF
--------------------
5.8.8b1 - 28-2-12
	* fix bug in lhaglue.f causing inability to access the abkm etc
	sets...

5.8.7 - 17.02.2012 - Addendum

    * grid files for (all work with standard 5.8.7

        NNPDF21_nnlo_mc_15_100
        NNPDF21_nnlo_mc_16_100
        NNPDF21_nnlo_mc_17_100
        NNPDF21_nnlo_mb_425_100
        NNPDF21_nnlo_mb_45_100
        NNPDF21_nnlo_mb_50_100
        NNPDF21_nnlo_mb_525_100
  
     and replaced gird file NNPDF21_FFN3_100  (error in old one)

5.8.7 - 15.02.2012

    * 5.8.7b3 - 17.01.2012
    
    * adds code files for AMB11 sets 

    * 5.8.7b2  - 18.11.2011

	* adds code for NNPDF21_nnlo_nf5_100 and NNPDF22_nlo_100 sets (and
	grid files)

	* src/wrapmrstqed*.f: Fix calls to the jeppe2 function (patches from Genser)

	* src/commonlha*.inc: Change path and name variable lengths to
	1024 characters to avoid overflows on e.g. Grid deployments.

	* configure.ac: Adding a check to make sure that no-one tries to
	install into --prefix=$PWD.
    
    * Fixes the problem that the u and d, and also the ubar and dbar, 
      partons were interchanged in error for the HKN nuclear PDF sets.

    * 5.8.7b1 identical to 5.8.6
        
5.8.6 - 02.08.2011

	* All of additions in beta 5.8.6 versions. including

	* HERAPDF 1.5 (NLO and NNLO, eigenvalue sets, alphas sets and variation sets)

	* NNPDF21 lo, lostar and nnlo sets

	* Corrected versions of abkm09 wrappers

	* Various fixes.

	* 5.8.6b5  - 11.07.2011

	* bin/lhapdf-config.in: fix --libdir treatment by defining the variable.

	* 5.8.6b4 - 28.6.2011

	* bin/lhapdf-config.in: fix --libdir treatment by defining the $exec_prefix
	variable. Bug reported by Hans-Werner Boschmann.

        * add code to handle NNPDF_21 nnlo and lo/lostar sets

        * corrected versions of abkm09 wrappers

	* pyext/setup.py.in: use CXXFLAGS and LDFLAGS variables safely in
	the Python extension build definition.

	* add -flat_namespace to the Mac OS X linker command, as requested
	by James Monk.

	* 5.8.6b3 - 05.5.2011

	* add code to handle herapdf1.5

	* bug fix in HERAPDF lite version causing strange values around
		x=10^06

	* 5.8.6b2 - 14.4.2011

	* corrected error in HKN code which hadd cause the u and d partons to
	  be interchanged, and also the ubar and dbar partons.

	* Updating Python wrapper generation to be compatible with GCC 4.6.


5.8.5 -02/02/2011

	* pdfsets moved to new location (SVN -> downloads) which requires
	  a new lhapdf-getdata script

	* addition of lots of NNPDF21 sets

	* fixing bug in mstw code which causes NaN outputs in various extreme
	  kinematic regions (q<M_c)  with Mc= 1.05 and 1.10 GeV

	* fixing AFG photon sets which gave parton distributions (not gluon)
	  2x too large.

	* fixing bug in several CT09 and cteq66alphas file causing failure in
	  lite mode

	* fixing bug in lite version of the HKN sets (using the wrong alphas
       	  code was causing failure)

	* stopping re-initialising sets already current (non-lite only)

	* Use less verbose compilation output. Use 'make V=1' to get a
	  verbose compile for debugging.

	* Adding better handling of failed data file loading: the error
	  message should now be helpful!

	* Fixing memory leaks in LHAPDF's use of binreloc dynamic path
	  lookup functions and inside binreloc itself.

	* Fixing a Python bug in handling of failed PDFsets index file download.

******************************************************************************
5.8.4 -21/7/2010

        * addition of CT10,CT10as,CT10w and CT10was sets.

        * Addition of MSTW2008 heavy-quark variation sets

	* Fix to Alphas calcualtion in HERAGRID sets

	* Addition of lhaglue numbers for all the sets in the NNPDRF2.0 alphas
	  variation sets.  Only the average set (0) was allocated a number
	  previously.

	* Adding EPS09 nuclear correction sets
	  These corrections are used along with a free proton set, selected
	  in the usual way,  with the call setlhaparm('EPS09') - also EPS09LO
	  and EPS09NLO.  In addition there are 15 +/- sets for uncertainties
	  used by setlhaparm('EPS09NLO,1') etc...

        * Small changes to abkm wrapper code (wrapabkm09.f..) to bring it in
	  line with the code on the abkm web page.

	* Adding abkm09_4_nlo - 4 flavour set. (40780-40805)
	* Adding abkm09_4_nnlo - 4 flavour set. (40810-40835)

	* Correcting abkm09 numbering in PDFsets.index

5.8.3 -18/5/2010

        * splitting cteq66a sets into cteq66a0,a1,a2,a3 to fix the problem
	of different q**2 bin edges. cteq66a gave incorrect pdf values for
	this reason.

	* adding HERAPDF10 alphas variation sets

	* adding NNPDF20 alphas variation sets

 	* fix for correct lamdbaqcd values for GJR and JR sets

	* fix for correct alphas values for cteq4 and 5

	* octave/wrapoctave.cc: Fix to avoid falling through switch
	statements (i.e. add break statements!).

	* bin: Converting PDF set downloading function in lhapdf-getdata to use
	urllib.urlretrieve, which is more efficient than the urllib2
	approach for large PDF grid files. Thanks to Dmitri Konstantinov
	for the suggestion.


5.8.2 - 17/3/2010

    * New PDF sets:   NNPDF20_100    (90800-90900)
                      NNPDF20_1000   (97000-98000)

                      abkm09_3_nlo   (40650-40675)
                      abkm09_5_nlo   (40750-40775)
                      abkm09_3_nnlo  (40850-40875)
                      abkm09_5_nnlo  (40950-40975)

                      CT09MCS        (10770)
                      CT09MC1        (10771)
                      CT09MC2        (10772)

                      cteq66alphas   (10595-10599)

    * Addition of USER grid file option (wrapusergrid)

    * Fix bugs 1) using multisets with lite options
               2) alphas values in gjr non zero sets
               3) MST2006 idef option in wrapevolve

5.8.1 - 11/12/2009

    * Added Octave interface, by Philip Ilten.

    * Added new sets HERAPDF10_EIG.LHgrid
                     HERAPDF10_VAR.LHgrid

5.8.0 - 07/10/2009

    * Adding the JR09VFnnloE pdfsets.

    * Adding --enable-sets=LIST option

    * Adding --incdir, --libdir, --cppflags and --ldflags arguments to lhapdf-config.

    * Adding definition of $datarootdir to lhapdf-config. Thanks to Thorsten Ohl for the bugfix.

5.7.1 - 08/07/2009

    * Adding new PDF sets:
        Nuclear PDFs  HKNlo and HKNnlo (LHgrid)  (19 sets central values  only -  various nuclei)
        JR09FFnnloE.LHgrid (Dynamic pdfs from JR collab)
        MSTW2008nlo_asmzrange (+ nnlo) (+68cl and 90cl error sets) alphas variations from MSTW
        NNPDF11_100 and NNPDF12_100.LHgrid ...  100 and 1000(1.2only) replica sets for v1.1 and 1.2

    * Added has_photon() function in Fortran and LHAPDF::hasPhoton in
	C++/Python, for detecting if the current set has an extra photon
	member.

	* Adding checks for the Python.h header, allowing the Python
	extension build request to fail more gracefully in the configure
	phase.

	* Adding --list, --force and other improvements to lhapdf-getdata.
	The PDFsets.index file is no longer used, preferring instead to get
	an always up-to-date list of PDF files from the SVN directory listing.

	* Committing the pdf-error-plot.py script for visualising PDF
	error set bands.

	* Committing the pdf-plot.py and pdf-ratio-plot.py example
	scripts, for plotting LHAPDF PDF sets with the Python matplotlib
	library.

	* Adding the lhapdf-getdata script, and bug-fixing Python logging
	initialisation for SL4 systems in lhapdf-query.

	* Removing bundled PDF sets. Now only one tarball is made, and
	only the PDFsets.index file is installed.

	* Fixing Mac compilation bug with NNPDF sets: replacing
	intrinsically looped READ statement with an explicit loop and
	normal READ (thanks to James Monk).

	* Adding DESTDIR to Python module install prefix (thanks to
	Nikolay Dyankov).


5.7.0 - 16/02/2009

        * Added MSTW and MSTW-lite code and grid files

	* Fixed problem with lhaglue numbering of NNPDF sets

	* Added a case-insensitive pattern match argument to the
	lhapdf-query "listnames" and "listsets" arguments,
	e.g. "lhapdf-query listsets CTEQ6".

	* Converted "go to" to "goto", since some Fortran compilers don't
	like the version with a space.

	* Fixed missing inline of operator<< for PDFInfo objects.

	* Replaced binreloc with an upgraded and symbol-independent copy.


5.6.0 - 23/10/2008

	* Added LHAPDF::setParameter(parm) function to the C++ interface
	--- it is just a thin wrapper on the Fortran setlhaparm. (AB)

	* Added LHAPDF/LHAPDFConfig.h header, automatically populated with
	LHAPDF_VERSION, LHAPDF_LOWMEM and LHAPDF_NMXSET. (AB)

	* Fixed error calculation and x-range in CCTest2, and improved
	efficiency of low mem mode example initialisation strategy. (AB)

	* Added Example7.f - fortran version of CCTest2.cc to demonstrate
	the calculation of hessian type errors using MRST2006nnlo.LHgrid (MW)

        * Fixed the set number for cteq6lg in the lhaglue strictm routine.
	(MW)

        * Added HERAPDF01 sets (MW)

	* Added NNPDF sets (MW)

	* Added GJR sets (MW)

        * Added several new 'lite' versions to reduce v-mem size in
	--enable-low-memory mode. : incl... mrstqed-lite, a02m-lite,
	h1-lite etc... (MW)

	* Removed wrapa02.f since this was not used and had been superseded
	by wrapa02m.f (MW)

	* Added "listsets", "listids", "listnames" and "version" commands
	to lhapdf-query. (AB)


5.5.1 - 26/09/2008

	* Improvements to archive building: svn export should no longer be
	needed before release since SVN files are automatically removed. (AB)

	* Fixed lhaprint to set silent if a stream other than #6 is used,
	and to call the common initialisation routine first. (MW)

	* Fixed MRST QED set to use arrays of compatible size with the
	"jeppe" interpolation routines. Spotted by Peter Richardson. (AB)

	* Removed broken caching of alpha_s computation in MRST
	sets. Thanks to Simon Plaetzer and Giuseppe Bozzi for alerting us
	to the problem. (MW)

	* Fixed units of lambda_qcd5 in cteq6ab, cteq65 and cteq66 PDF
	grid files. This was causing crashes in MC@NLO. Thanks to James
	Ferrando for spotting and solving the problem. (MW)

	* Added building of lhapdf-*.tar.bz2 tarball (better compression
	than gzip). (AB)

	* Added "by LHAPDF ID code" initialisation to C++ interface. (AB)

	* Fixing deprecated uses of GOTO to jump to end of a DO loop. (AB)

	* Added regression scripts to dump out randomly sampled PDF
	values (with logarithmic measures on x and Q) and to then read
	them back in to check for deviations between DF values between old
	and new LHAPDF versions. (AB)

	* Provided lhapdf-query script for command-line querying of
	LHAPDF. (AB)

	* Added C++/Python API functions to read PDF set metadata from
	PDFsets.index. (AB)

	* Regression of F90 steering modules to use common blocks and an
	initialising subroutine, which is not completely robust but keeps
	SLC4 LCG applications happy (they have no F90 compiler). Regressed
	F90 file extensions from .f90 to .f, since g77 refuses to even
	attempt to compile .f90 extensions, even if it can handle the
	contents. (AB)

	* Conversion of all Fortran source files to F90-type
	free-format. With a .f suffix and -ffree-form flag, g77 will parse
	these files successfully. (AB)

5.5.0 - 08/08/08

	* Doxygen now attempts to generate a Doxyfile compatible with the
	version being used, in an attempt to avoid some reported segfaults
	on old systems. (AB)

	* Fixed several uninitialised variables in Fortran routines. (AB)

	* Provided "lite" versions of the CTEQ and MRST sets, without the
	43 extra arrays allocated for Hessian sets. Use the
	--enable-low-memory configure switch to disable the extra 43 sets:
	this is mainly of interest for batch use on systems where the
	batch accountancy is militant about vram occupancy: on standalone
	systems the unused memory pages are never really allocated, so
	there is no danger in having them allocated. (MW, AB)

	* Added configure flag to control max number of concurrent PDF
	sets. Affects virtual memory overhead, so if LHAPDF is being used
	on a system which uses vram-based accountancy, you may need to
	ensure that this is set as low as possible. By default, it's set
	to 1. (AB)

	* Compatibility with pgf90, pgf95 and ifort compilers. (AB)

	* Version number in banner now set from autoconf spec. (AB)

	* Various F90 conversions, clean-ups and fixes. Improved
	portability as a result of e.g. double standardisation. Tab
	characters removed from F77 files (converted to 8 spaces) since
	the fixed format makes them dangerous. (AB)

	* Fixed uninitialised steering common blocks (block data doesn't
	work from libraries) by moving to using F90 "modules"
	instead and fixing some scoping bugs in the getdirpath function.
	LHAPDF now requires a F90-compatible compiler. (David Grellscheid, AB)

	* Renamed C++ test programs, added an installcheck target and
	added Fortran tests #5 and #6 to "make check". (AB)

	* Fixed C string errors in path-finding functions. These functions
          have been converted to C++ instead, since the C string copying /
          pointer issues were extremely fragile. The test now also checks for
          existence of LHAPDF-specific directory contents before returning,
          for extra safety. (AB)


5.4.1 - 10/07/08

	* Fixed bug x/q limit bug in cteq6 when using different members
          the same set. (MW)

	* Updated obsolete Fortran code which produced annoying compilation
	  warning messages. (MW)

	* Default member 0 now set automatically before call to InitPDF. (MW)

	* Fixed configure bugs with disabling old ccwrap interface and
	interaction of ccwrap with pyext. (AB)

	* Added more pedantic compiler warning messages and fixed
	highlighted errors. (AB)

	* Headers are now installed to the proper "include/LHAPDF"
	directory, rather than "include/lhapdf" (AB)

	* Added functions to get the install prefix, PDF sets directory
	and PDF sets index file path, and a corresponding test program for
	the C++ wrapper. (AB)

	* Added binreloc control to configure. (AB)

	* Rationalised C++ initialising functions (AB)

	* C++ Doxygen improvements (AB)

	* Provided C array versions of the multi-set C++ functions. (AB)

	* Renamed C++ "M" functions to use C++ style function signatures. (AB)

	* Fixed a few issues with C++-Fortran interfacing and headers. (AB)

	* Fixed potential C string length bug in getdatapath_ function. (AB)

	* Changes for GCC 4.3 compliance. (AB & MW)

	* Added Mac OS X build detection. (AB)

	* Removed constituent library linking with $FCFLAGS arguments,
	since this results in a double declaration of the gfortran dummy
	main function on some systems. (AB)

	* --disable-doxygen configure flag added. (AB)

5.4.0 - 23/05/08

	* New PDF sets included: (MW)

		cteq66.LHgrid  - cteq 6.6 fit with error sets (1+44)
		cteq66c.LHgrid - cteq 6.6 intrinsic charm sets (4)
		cteq66a.LHgrid - cteq 6.6 alternate alphas sets (4)

		MRST2007lomod.LHgrid - PDFs for LO generators (1)
		MRSTMCal.LHgrid - LO MC fit - alphas KT^2 (1)

	* New Nuclear PDF formula EPS08 available as an alternate to
	  the older EKS98.

        * Various bug fixes from patches to 5.3.1

	* Improved handling of error reporting through  setlhaparm('SILENT')
	  and setlhaparm('LOWKEY')

	* LHAPDF C++ headers all located in include directory, properly
	installed into $prefix/include/LHAPDF, and distributed in the
	tarball. (AB)

	* Merged main C++ wrapper into libLHAPDF, retaining libLHAPDFWrap
	for backwards compatibility. (AB)

	* Added libtool library versioning to both the core and C++
	wrapper libraries. (AB)

	* LHAPDF banner now obeys the SILENT/LOWKEY verbosity control
	though the LHAPDF/C++ interfaces, as for LHAGLUE. (AB)

	* Added enum for semantic choice between LHgrid and LHpdf
	interpolation/evolution to the C++ wrapper.

	* Added SWIG-based Python extension and test program. (AB)

	* Using Fortran examples and C++ wrapper test programs as automake
	unit tests. (AB)

	* Deprecated class-based C++ wrapper and provided replacement
	version based on global functions in a new "LHAPDF" namespace. The
	new interface is declared in LHAPDF.h and also provides a greater
	subset of the Fortran configuration functions and more efficient
	versions of the xfx functions which bypass the copying in and out
	of STL vectors. (AB)

	* Added Doxygen build and install for C++ wrapper. (AB)

	* Added enable/disable flags for LHAGLUE and C++ wrapper. (AB)

	* Fixed C++ wrapper to pass hidden string length parameters to
	Fortran subroutines, use explicit pointers rather than
	references (which are implementation-sensitive) in Fortran
	interface, and to use STL strings in its C++ API. (AB)

	* Added make check call to lhapdf-example1, and a make hook to
	build symlinks so that the PDTsets will be found even if the
	src/libLHAPDF library is linked, rather than the installed
	version. (AB)

	* Disabled internal use of system("lhapdf-config --pdfsets ...")
	call to work out position of the PDF sets. This is now done via
	the binreloc (http://www.autopackage.org/docs/binreloc/) system,
	with a fallback value set by a preprocessor define, whose value is
	the same as that set via lhapdf-config. This makes LHAPDF
	relocateable, as defined on the binreloc Web page, so it can now
	be considered for user-space packaging with e.g. autopackage. The
	binreloc function call must be done in C, so a new getdatapath.c
	source file has been added. The LHAPATH common block variable and
	environment variable can still used to override this
	mechanism. (AB)

	* Using AC_FC_WRAPPERS to deal with Fortran compilers which
	require a dummy MAIN__ function and to portably use different
	Fortran symbol manglings (AB);

	* Added the --disable-install-pdfsets switch to configure (default
	is to install PDF sets: this provides a mechanism to disable
	it. (AB)

	* Added --enable-debug configure flag, and corrected handling of
	user-set FFLAGS variable (previously it was always overridden by
	the default). (AB)

5.3.1 - 16/11/07

        * STRUCTM/STRUCTP now allow constants in calling argument string.

        * New MRST 2001/2204 code for better extrapolation, which also fixes the
          bug which caused failure of these on extrapolation.

        * Fixes VZERO name conflict with (for example) CERNLIB

        * Corrects error in MRST2006 code causing PDFs to be frozen at Q=1.43 when
          called with Q between 1.43 and 4.3.

5.3 - 28/06/2007

	* Addition of the following new PDFsets: mrst2004qed mrst2006nnlo
	cteq65 cteq65c cteq5f3 cteq5f4

	* GetNf() now returns '-1' for the .LHgrid files - it only returns
	the correct number of flavours if evolution is done in the PHAPDF
	program -ie .LHpdf files

	* C++ wrapper library (LHAPDFWrap.a and .so) is now built
	automatically in the make/make install procedure and placed in the
	standard LHAPDF library directory.

	* New Fortran routines and C++ methods to extract the max and min
	X and Q**2 values in the raw LHAPDF code.

	* 'extrapolate' the PDFs beyond the max/min x/Q2 now works with
	the raw LHAPDF code.

	* Addition of wrapUSER.f routine to allow a hook for user written
	PDF sets to be added.

	* Finding PDF sets with lhapdf-config now writes into
	$HOME/.lhapdf/, rather than /tmp. This avoids write permission
	conflicts between users.

	* Added PDFsets.index file

	* Distributions are now built in a more standard way with "make
	dist" and the dist-hook mechanism

5.2p3 - 6/11/2006

	* Fixed error causing anti-b in zeus2002 to be zero.

	* For consistency the argument lists in a02init,a02minit and
	grxlim corrected.

5.2p2 - 19/7/2006

	* Fixed error in print statement in Example5.f

	* Removed reference to the H12000lo2 sets in lhaglue.f

	* Added common LUDAT1 to lhaglue.f to allow working with older
	Pythia versions

	* Fixed problems with 's in decriptions causing errors with
	gfortran compiler

	* Fixed problem causing QCDNUM (in MRST and ZEUS) to work slowly
	with V5.

5.2.1 - 21/6/2006

	* Addition of MRST2004 Fieed flavour FF3 and FF4 sets

	* H12000 set back to original grid files (problems with extended
	fit)

5.2.0 - 13/6/2006

	* Addition of Nuclear PDFs (as in PDFLIB - STRUCTA(),
	evolvePDFa(), evolvePDFaM())

	* Fix to bug in CTEQ at x=1.0 (now identically zero)

	* $LHAPATH environmental variable now works with LHAPDF raw as
	well as LHAGLUE

5.1.0 - 26/4/2006

	* Libraries now installed by libtool, including shared
	libraries (.so).

	* STRUCTM fixed to work for photon PDFs (sets IP2=0,P2=0.0)

	* Changed function names BETA and GAMMA to BETA_LHA and GAMMA_LHA
	to avoid conflicts.

	* Fixed bug in SMRSPI.LHgrid file

	* Fixed small bugs to work with Intel compiler.

5.0.0 - 1/3/2006

	* Addition of multiset initialisation (LHAPDF and LHAglue)

	* Setting the path to PDF files by function call

	* Setting the path to PDF files using environmental variable
	$LHAPATH

	* Setting control parameters via function calls.

	* Renaming include files as .inc in src directory

4.2.0 - 14/11/2005

	* Addition of new cteq6AB (variable alphas(mz)) PDF sets.

	* Mods by cteq to other cteq code to improve speed.

	* New a02m_nnlo.LHgrid file (previous one was erroneously the same
	as LO).

	* Fix to SMRSPI code which was wrongly setting usea to zero.

4.1.1 - 19/9/2005

	* Patch to fix bug affecting cteq6(LHgrid) at x>0.9763

4.1.0:
	* Converted build system to use GNU autotools

	* Introduced path-independent references to PDF sets via the
	lhapdf-config script and the InitPDFsetByName subroutine. Note
	that the script must be in the user execution path and must have
	been installed by autotools to work properly (AB).

4.0.0:
	* The addition of PDF sets for photons and pions

	* Additions to lhaglue including STRUCTP for photons and PDFSTA
	for statistics

	* More PDF sets included.

	* A new simpler directory structure

3.0.0:
	* The addition of the routines PDFSET and STRUCTM plus a unique
	PDF numbering scheme (all from the lhaglue package) to enable the
	LHAPDF to be used in the same way as PDFLIB.

	* More PDF sets included.

2.0.0:
	* Single compilation of code for ALL PDF sets. (In v1 the user had
	to recompile the code when using different evolution code
	packages)

	* Ability to use the original interpolation grid files (.LHgrid)
	produced by the authors of the PDFs, as well as parameter
	files (.LHpdf) to define the PDFs with seamless switching between
	the two types.

	* More PDF sets included.