4 Jun 2004(SCH)
 - Fixed bug when using si_mode with QTLs and lm_ratio>0
3 Jun 2004(SCH)
 - Fixed bug in gen_pen() with genetics groups
24 Feb 2004(SCH)
 - Fixed printing bug in IBD estimates with dummy founders
16 Jan 2004(SCH)
 - Fixed bug when ids are shared between families and individuals.
9 Jan 2004 (EWD)
 - Merge in QTL genotype printing code
30 Oct 2003(SCH)
 - Add support for >32 segregating alleles using the -DUSE_LONGLONG switch
   (needs more testing).  This whole section should be rewritten to remove the
	limit on allele numbers.
 - Add dump file support for IBD estimation runs
 - Modified peeling sequence determination to use node weights, and to pick 
   the best of 4 greedy strategies.
29 Oct 2003(SCH)
 - Fix compile warnings about aliasing from gcc3.3
 - Add command line switches to prep and loki to redirect output files
   -d sets the directory for output and intermediate files
	-f sets the prefix for output files (normally 'loki')
	Note that the -d option does *not* affect the control and parameter files
	Note also that the existing -d option for Loki (to set the dump file to
	restart from) has been changed to -f.
15 April 2003(SCH)
 - Finally (hopefully) fix problem with crash when no log file specified
 - Synchronize prep code from current and release versions
 - Rationalize missing data code matching as suggested by Dan Weeks to allow
   for numeric matching.  Updated manual accordingly
21 March 2003(SCH)
 - Merge in bugfix for control_parse problem with arrays in File statement
 	from current
 - Fix bug with dummy pedigree members where sex was not being initialized
 - Fix bug on Alpha in seg_pen
21 February 2003(SCH)
 - Merge in bugfix for seg_pen with genetic groups from current
19 February 2003(SCH)
 - Fix memory corruption in IBD code when singleton component is larger
   than first component
13 February 2003(SCH)
 - Fix bug in IBD reporting on Alphas
 - Fix bug in Merlin IBD output with singletons
16 December 2002(SCH)
 - Fix bug in reporting inbreeding coefficients
 - Remove reporting of loop breakers in log file (not correct in any case...)
 - Stop printing dummy individuals to logfile if pruned
30 October 2002 (SCH)
 - Lots of changes (not all today...)
 - Redo IBD output code.  Now can output in SOLAR format, with proper
	translation of IDs
 - Alter Missing command so that general classes of variables can be
   specified (Pedigree/Genotypes/Factors/Real)
 - Set sensible defaults for Missing (all categorical variables)
 - Allow both haplotype to come together in a data field to allow reading of 
   genotypes like 118/120.  Add GS variable which determines character that
   separates genotypes in this case.
 - Add skip option to file command - causes Loki to skip a numnber of lines
   at the top of a data file.
 - Alter documentation to reflect above changes.
 - Put binary tree code into library.  Remove redundant copies of code.
11 September 2002 (SCH)
 - Fix bug that caused Loki to hang if performing a segregation analysis and
    a total map length was not set.
2 August 2002 (SCH)
 - Reorganize source code into separate prep and loki directories to make
    it easier to reuse bits of the code
23 July 2002 (SCH)
 - Modify IBD output code so that be default, all zeros are not printed
 - Fix IBD output bug with >1 location
 - Add options for calculating IBD matrices on a grid or at every marker
 - Remove references to TDT and IBS analyses (old, broken)
12 July 2002 (SCH)
 - Modify IBD estimation code so that fixed values are not calculated
 - Remove printing bug in no marker data case
10 July 2002 (SCH)
 - Change default to CONSTANT for all variables.  Add MULTIPLE keyword 
    to set multiple records for variables.  Add multiple_records system
	 variable to switch back to default behaviour
9 July 2002 (SCH)
 - Add checks for failure to open log file in Loki
2 July 2002 (SCH)
 - Add code to support parallel RNGs
28 June 2002 (SCH)
 - Clean up bugs found when testing on Ellen's Alphas
    fix unitialized variable (flag) bug in loki_sample.c
    make conversion of genotype to alleles robust to rounding errors
25 June 2002 (SCH)
 - Fix memory allocation bug in Solaris
 - Remove globals from loki_peel.c
 - Finish splitting loki_peel.c
24 June 2002 (SCH)
 - Fix bug when M-sampler is used and allele frequencies are specified
 - Add new tests
** Version 2.4.2
18 June 2002 (SCH)
 - Too many changes to list.  New sampling code, many bug fixes, many
   new features.  I'll try and keep this list more up to date in future
** Version 2.3.0_beta
30 August 2000 (SCH)
 - Change Makefiles so they work with OSF1 make as well as GNU make 
 - Add README_optimization to test/ to warn of difficulties with compiler
   optimization flags that affect math operaitons.
29 August 2000 (SCH)
 - Fix overdominant flag so it now works again.
28 August 2000 (SCH)
 - Change storage of factors in control_parse.y and read_data.c to a
   balanced binary tree rather than a hash table.  Rewrite recode_factor()
	to further increase speed.
23 August 2000 (SCH)
 - Implement t-distributed error models.  Not fully tested as yet - don't use.
16 August 2000 (SCH)
 - Implement polygenic and uncorrelated random effects into loki.  Rewrite
   effect sampling so that *all* model effects (QTL, candidate genes, fixed
	effects and random effects) are updated simultaneously using sparse
	Gaussian elimination.
15 August 2000 (SCH)
 - Begin splitting source files into smaller units

** Version 2.2.1_r2
29 Feb 2000 (SCH)
 - Fix problem in read_binfiles.c where a component with no phenotype data
   but some genotype data would crash on peeling the trait locus.
 - Bug fix to control_parse.y to prevent crashing when trace_affect or
   trace_censored were used.
 - Lump components with single individuals together to simplify handling.
25 Feb 2000 (SCH) 
 - Bug fix to write_data.c and read_binfiles.c which would cause genotype
   data to be scrambled when an individual had genotype data but no phenotype
	data.  Not normally a problem for the L-Sampler as the raw genotype data
	is not used, but did affect the M-Sampler.
 - Changes to Makefiles to allow compilation in FreeBSD
23 Feb 2000 (SCH)
 - Changes to M sampler and ibd handling code to clear up several bugs and
   allow use of sloppy segregation patterns (better mixing) from the 
	L sampler.
22 Feb 2000 (SCH)
 - Modify TL_Birth_Death() again to avoid Ve problems with censored traits.
 
** Version 2.2.1_r1
9 Feb 2000 (SCH)
 - Fix TL_Birth_Death() so it works again with censored traits (broken in
   2.2.0_r3)
8 Feb 2000 (SCH)
 - Modified output routines for loki.pos so that sample_from limit is
   honoured.
 - Fixed bug in restrict_data() which prevented conditions with unary minus
   operators from working correctly.
 - Fixed memory allocation bug in loki_complex_peelop() which caused writing
   to random pointers in some circumstances when candidate genes were used.
28 Jan 2000 (SCH) 
 - Add code to SampleLoop() to output QTL position every iteration to
   loki.pos.  Only outputs when a change occurs.
22 Jan 2000 (SCH) 
 - Fixed bug added in 2.2.1 in loki_init().
** Version 2.2.1
21 Jan 2000 (SCH)
 - Changed start_time/end_time processing so elapsed, system, and user time
   is written to the dumpfiles, allowing more accurate total times
	to be printed to the logfiles in the case of the program being stopped.
	Note that the new read_dump() routine will still read in the old dump
   files which lack this time information.  Note also that the time
   information is written as doubles which may lose some accuracy, but is
	(I think) portable.
 - Changed the order of statements in loki_init() to get around optimization 
   bug in gcc 2.95.2 on Ultras.
 - Made it so alleles which are no longer found in the dataset after
   restriction statements and genotype cleaning, are deleted from the
	appropriate marker.
20 Jan 2000 (SCH)
 - Changed loki_sample() so the model is now printed out to the output file.
 - Fixed bug in read_dump() where a dumpfile would not be read in correctly
   if a censored data model was used.
** Version 2.2.0_r5
18 Jan 2000 (SCH)
 - Removed bug (introduced in 2.2.0_r4) with censored or affected data where 
   indicator variable was marked as not required and removed.
 - Change libsrc/Makefile.in and lokisrc/Makefile.in so that it is easier
   to enable libdmalloc usage (just uncomment the appropriate lines in the
	makefiles.
	
** Version 2.2.0_r4
12 Jan 2000 (SCH)
 - Cleanup restrict code so that unsed restrictions and variables are not
   kept.
 - fix bug in restrict_data.c which prevented restrictions on everything
   (i.e., with null affected list) from taking affect.
 - Change param_parse.y so files can be 'included' using include "file".
 
** Version 2.2.0_r3
   Nov/Dec 1999 (SCH)
 - Change reversible jump steps so that ve is changed simultanously with
   number of QTL to improve acceptance ratio.  **Note** that only add and 
	delete steps are now active as split/combine steps are broken!
 - Change sample_effects() in loki_sample.c so that QTL effects and covariate 
   effects are sampled jointly - again to improve mixing.  **NOTE** This has 
	the side effect of breaking the no_overdominant flag.  Correct bug in
	sample_effects() where the wrong covariate levels were used for discrete
	covariates.
 
** Version 2.2.0_r2
16 Oct 1999 (SCH)
 - Binary file formats changed so that they are (almost) platform independent.
17 Sep 1999 (SCH)
 - Many changes to all files, see summaries below
 - Long-standing bug causing random crashes after weeks of running tracked
   down to RNG occasionally producing zeros and ones.  Corrected.
 - IBD routines now in main distribution, though not documented as user
	interface not necessarily stable.  
 - Quantitative analyses can now be stopped and restarted which allows long
   runs with unstable systems, and simplifies bug tracking.
 - Output format changed yet again, but "OUTPUT TYPE" option in parameter
   file allows previous formats to be selected.
 - M and LM sampler routines present, but not fully integrated.
 - Genotype error correcting routines added.  Use 'set correct_errors 1' in
   control file to activate.  A subset of genotypes wil be zeroed leading to
   a consistent configuration.  Very handy, but use with caution.

** Version 2.1.4_r1

21 Mar 1998 (SCH)
 - setup_ped.c: Change output of component sizes - individual component
   sizes are not now listed.  Change formatting out output for pruned
	individuals.  Sort components in descending size order.
 - param_parse.y, param_lex.l, loki_sample.c, loki.h: Add LIMIT
   variances option so lower limits on variance components can be set.
16-21 Mar 1998 (SCH)
 - most files: Allow specification of sex specific maps.  Change sampling
   routines for genotypes and loci positions to reflect this.  Change
	output routines and loki_ext accordingly.
20 Mar 1998 (SCH)
 - control_parse.y, write_report.c: Add OUTPUT option to allow pedigree to
   be written to a datafile after pruning and recoding.
 - setup_ped.c: Added checks for close inbreeding/weird inbreeding patterns.
 
** Version 2.1.4

15 Mar 1998 (SCH)
 - most files...: Enable GROUP command for specification of genetic groups.
	Added GROUP ORDER command for parameter file to specify order that
	multiple allele frequencies appear.  FREQUENCY statements can now take
	multiple allele frequencies (if multiple groups exist).  Modified
	calculation of founder probabilities to use group frequencies; ditto for
	sampling frequencies.  Modifiy reversible jump steps to account for
	multiple sets of allele frequencies.
1 Mar 1998 (SCH) 
 - recode_factors.c: Change consistency checks for genetic data so now   
	genotypes rather than haplotypes are checked for consistency.
28 Feb 1998 (SCH) 
 - loki.c: Write starting seedfile to to logfile. 
 - read_data.c setup_ped.c control_parse.y write_data.c read_binfiles.c
	loki_init.c: Allow sex to be specified using sex command.  Check sex
	against pedigree data.  Modify input/output routines for loki.out to
	allow for sex information. 
 - control_lex.l control_parse.y: Fix looping bug for control files. 
	Fix some small bugs concerning format clauses.
27 Feb 1998 (SCH) 
 - read_data.c: Allow blank records in datafiles.
 - setup_ped.c: Add code to count pedigree loops.
 
** Version 2.1.3

26 Feb 1998 (SCH) gen_elim.c
 - gen_elim.c: Fix bug when pruning (level=2) resulted in singletons.