[Nauty] nauty and Traces version 2.8.9 released.

[Brendan McKay]

Dear nauty people,

Version 2.8.9 is now available at the usual places:
https://users.cecs.anu.edu.au/~bdm/nauty
https://pallini.di.uniroma1.it

The biggest change this time is the use of the 'libtool' system to create
shared/dynamic libraries as well as static libraries.  To complete library
creation you probably need "make install" after "make".

Shared/dynamic libraries are notoriously system-dependent and even
though I tested it on multiple Linux-like systems as well as Mac, there
can easily be problems still. ONE REQUEST is that if someone understands
shared/dynamic libraries on cygwin, please try it and tell me if it is doing
what it is supposed to do.

Brendan and Adolfo
------------------------------------------------------------

Other changes:

* The install target puts package configuration files (*.pc) into
   ${pkgconfigdir}, not ${pkgconfigexecdir}.  The default location is now
   ${prefix}/libdata/pkgconfig for FreeBSD and ${libdir}/pkgconfig
   otherwise.  You can select another location using the option
   --with-pkgconfigdir=DIR with configure.

* The make target TLSlibs also makes nautyTS.a, nautyTS1.a and, if 128-bit
   arithmetic is available, nautyTQ.a and nautyTQ1.a. The package configuration
   files for all nautyT*.a libraries now have -DUSE_TLS in their C options.
   After that you can install with "make TLSinstall".

* There is a new utility uniqg for filtering by isomorphism class.
   Use  uniqg --help  for full instructions.
   This utility uses the SHA256 256-bit cryptographic signature to compare
   canonically labelled graphs. Though it is theoretically possible for two
   different graphs to compare equal, years of theoretical effort and
   centuries of computer time by the cryptology community has failed to
   find two bit strings with the same signature.
   Distinct signatures are held in memory, 56 bytes each, so you will run
   out of memory if you have an extremely large number (more than 19 million
   per gigabyte). The total number of graphs doesn't matter, just the
   number of isomorphism classes.
   A particularly useful facility is that you can specify a file of graphs
   to exclude from the output.
   Note that graphs are written in the order they are read. They are not
   sorted in the manner of shortg.

* vcolg has an option -O for adding loops to vertices with weight 0.

* nauty.h previously defined groupsize1 and groupsize2 as synonyms for
   grpsize1 and grpsize2 (the names of the group size fields in the stats
   structure. This dates back to the days when some compilers didn't accept
   variable names longer than 8 characters and no longer has a purpose.
   These definitions have now been removed. The correct names for the fields
   of the stats structure are grpsize1 and grpsize2.

* New options for countg and pickg:
   --O : number of orbits on edges
   --OO : number of orbits on directed edges
      (--O and --OO are the same for digraphs.)
   -tt : 2 for arc-transitive, 1 for edge-transitive, 0 for neither.

* hamheuristic -V causes graphs for which a cycle (path with -p) was found
   to be output, rather than the default which is the opposite.

* assembleg has a new option -k for restricting the number of inputs used
   to make each output. (Thus, if the inputs are connected, -k will restrict
   the number of components in the output.)

* productg has a new option -u to take the disjoint union of two graphs.

* delptg has a new option -S# for setting the random number seed (used
   only in conjunction with -r).  Without -S, the seed is set from the clock.

* The meaning of geng -bT has changed. Previously it could only make forests,
   so was useless. Now it makes chordal-bipartite graphs, which are bipartite
   graphs having no induced cycles longer than 4-cycles.

* dreadnaut can now read undirected graphs with loops. So can dretog.

* The -f option in labelg and shortg is now available for Traces and has a
   more general syntax. The argument is a string of characters considered
   extended to infinity with 'z'. Vertices are coloured by the corresponding
   character and the canonical label arranges them in (ascii) order of colour.
   The shortcut c^N, where c is a character and N is a number, means N copies
   of c. Also, starting the string with '-' colours the vertices from the
   end backwards and arranges them in decreasing ascii order.  Examples:
   -fa    fix the first vertex, placing it first
   -fa^4  fix the first 4 vertices as a set, placing them first in some order
   -fba   fix the first two vertices, placing them second and first
   -f-a   fix the last vertex, placing it last
   -f-ba^9999 fix the last vertex, placing it first (assuming n <= 10000)

* listg has a new option -m to write a graph or digraph in Mathematica format.

* The previous definition of NORET_ATTR (which says that a function never
   returns) didn't work in C++ so I replaced it by code sent by Jerry James.

* planarg now has two options for writing planar embeddings in planarcode.
   They are only different if the number of vertices is at least 256.
   -p : write multibyte integers high-order bytes first (big-endian)
   -P : write multibyte integers low-order bytes first (little-endian)

* subdivideg has a new option -i for performing a series reduction.
   For undirected graphs, repeatedly replace x--y--z by x--z if
       x,y are not adjacent and y has no other neighbours.
   For digraphs, repeatedly replace x->y->z by x->z if x,z are
       distinct, x->z is not present, and y has no other neighbours.