Changes version 1.34
================================
  o fix deprecated function declaration without prototype
  o fix LaTeX style issue
  o fix netlib.org URL

Changes version 1.33
================================
  o back-compatibility with R <= 3.6.2
    (patch provided by Davis Vaughan) 

Changes version 1.32
================================
  o USE_FC_LEN_T for Fortran BLAS/LAPACK routines
    to ensure compatibility with R > 4.2

Changes version 1.31
================================
  o replace deprecated S constant DOUBLE_XMAX by DBL_MAX

Changes version 1.30
================================
  o add doi to references

Changes version 1.29
================================
   o replace legacy S macros PROBLEM and ERROR in C code with Rf_error()
   o replace http:// with https:// if possible

Changes version 1.28
================================
  o minor: avoid an implicit type conversion at the C level

Changes version 1.27.1
================================
  o C level change to fix symbol clash on MacOS
    solution kindly provided by Brian Ripley

Changes version 1.27
================================
  o C level changes to ensure compatibility with gcc 10
    solution kindly provided by Brian Ripley

Changes version 1.26
================================
  o fix outdated class checks to ensure R 4.0.0 compatibility
    (Karline, Thomas)
  o Fortran modernization: initialization of variables (Karline)

Changes version 1.25
================================
  o add original authors of LINPACK, ODEPACK and SPARSKIT to Authors@R
    (Thomas, Karline, Woody)

Changes version 1.24
================================
  o fix compiler warnings to improve Fortran compatibility 
    (Thomas, Karline, Woody)
    thanks to Brian Ripley and Kurt Hornik
  o iteration: set attribute before calling diagnostics (Karline)
  o add open researcher id, ORCID (Thomas)

Changes version 1.23
================================
  o add FME to 'Suggests'

Changes version 1.22
================================
  o small updates of examples (Thomas)
  o improve Fortran compiler compatibility (Thomas)

Changes version 1.21
================================
 o change way how PROTECT/UNPROTECT is handled (Thomas)
   (suggested by Tomas Kalibera / R-Core)
 o fixed inconsistency in the aphid model example
   (suggested by Sarah Kintner)
   
Changes version 1.20
================================
 o register native routines (Thomas)
 o check if event data frame has ordered time (and if not, order)
 o change 'event list' to event matrix or data frame in docs
 o intentional version jump to indicate chances at the C level

Changes version 1.14
================================
o matplot.deSolve is not anymore exported as matplot to avoid the 
  respective startup message
o please use matplot.deSolve or the alias matplot.0D instead  (Thomas)
o small fix that allows parameters in list format for DLLfunc and DLLres
o a little bit Fortran modification (e.g. avoid real*8 and complex*16 types)

Changes version 1.13
================================
 o observed data and plot.deSolve / matplot for multiple outputs  (Karline)
 o combining compiled code function with R code event function (Karline)
 o check sorting of event times (Karline)
 o fix bug related to negative event time (patch supplied by J. Stott)
 o relax setting of tcrit to make integration with events slightly faster
   (patch from J. Stott)
 o adapt maxstep calculation for rk methods,
   print a warning if maxsteps is exceeded, fix diagnostics (Thomas)
 o more argument checking for rk solvers (Thomas)
 o add reference to book of Soetaert, Cash and Mazzia (2012)

Changes version 1.12
================================
 o new functions matplot.deSolve and matplot.1D
 o fix valgrind issue (detected by new compilers)
 o small improvments of plotting functions
 o import standard packages as required by upcoming R versions

Changes version 1.11
================================
 o compiledCode vignette now with dede example
 o warning and error bug resolved
 o Time SEXP incompatibility with R 3.1.1 resolved
 o CFunc compatibility (compiled code)

Changes version 1.10.9
================================
 o documentation updates, hyperlinks to examples and vignettes
 o moved example directories

Changes version 1.10.8
================================
 o remove redundant .R files from inst/doc
 o fixed bug in event code (patch contributed by Jonathan Stott)

Changes version 1.10.7
================================
 o Fortran examples of compiled dede models (Woody)
 o vignettes moved to /vignettes
 o roles of authors (Authors@R)
 o function timestep is now internal
 o small documentation updates

Changes version 1.10.6 (Thomas) 
================================
 o change declaration of variable dimensions from (1) to (*) in legacy
   Fortran code to pass automatic bounds check
 o remove the Jacobian examples from ?ode because banddown=0 can
   lead to problems on some systems; examples will come back in a
   next release
 o fixed bug in the "iteration" solver
 o small documentation updates

Changes version 1.10.5 (Karline, Thomas)
================================
 o extended subset.deSolve with argument arr, when TRUE returns an
   array for >2-D output
 o fixed the R compiler notes
 o plot.ode.2D now has an mtext argument, via the ..., to label multiple 
    figures in margin... CHECK - see ode.2D
 o subset can also be a vector with indices in addition to logical
 o image with legend = TRUE changed size of plot in different layouts - now solved
   (by adding par(mar = par("mar")) )
 o new method to output warnings and error messages 
 o add data type check for external outputs in rk_util.c
 o add interface for compiled dede models
 o emphasize consistent order of states in y and return value of func
 o changes of Fortran error messages (to be continued)

Changes version 1.10-4 (Thomas, Karline)
================================
 o allow reverted time vector for fixed step solvers
   - todo: find solution for dense output methods, and Livermore solvers
 o all solvers now have default atol = 1e-6; before this daspk and vode had 1e-8.
 o multiple warnings from daspk if num steps = 500 toggled off.
 o added input argument "nind" to daspk, to make it compatible with radau.
   this also changes the way the variables are weighed, 
   hence this differs from the original daspk 2.0 code.
 o improved warning printing in daspk and vode 
 o extended sparse Jacobian input in lsodes. (2-D and 3-D sparsity
   with mapping var and arbitrary sparsity in ian/jan format).

Changes version 1.10-3 (Karline)
================================
 o rwork and iwork in lsodes from Fortran -> C (to remove compiler warnings)
 o roots + events:  now certain roots can stop simulation + fixed bug in radau root
 o improved events\roots help file
 o diagnostics(out) gave error in case method=iteration (no rstate) now fixed
 o the package authors agreed to assign the maintainer role to T.P.,
   but the order of authorship and credits remain unchanged.

Changes version 1.10-2 (Karline)
================================
 o remove NAs from forcing functions - when used in DLL (file forcings.R)
 o new argument "restructure" in ode.1D, for use with implicit solvers not in deSolve
 o removed requirement to have eventfunc in compiled code when func is in compiled code
 o subsetting on summary.deSolve
 
Changes version 1.10-1 (Thomas)
===============================
 o remove several redundant variables from C code
 o add NEWS file

Changes version 1.10 (Karline, Thomas)
======================================
 o compiled code using mass in daspk
 o cleanEventTimes

Changes version 1.9+ (Karline)
==============================
 o roots, events, lags in radau
 o roots in lsodes
 o lags in daspk
 o ode (method = "iteration")

Changes version 1.9 (Karline, Thomas)
=====================================
 o summary.deSolve
 o subset.deSolve
 o plotting deSolve objects improved:
   - plot more than one output in same figures (scenarios), 
   - add observations
 o vignette improved
 o fixed bug in 'timesteps'

Changes version 1.8.1 (Thomas, Woody, Karline)
==============================================
 o fixed compiler warnings using valgrind
 o fixed compiler warning C-code

Changes version 1.8 (Thomas)
============================
 o Dormand-Prince 8(7) coefficients use now common 
   instead of decimal fractions


Changes version 1.8 (Karline)
=============================
 o Runge-Kuttas:
   - extra output: number of failed steps (see also 2)
   - number of function evaluations + 1 for initial condition
   - dense output for cash-karp
   - dopri8(7) added
   - radau added!! implicit runge kutta, solves also DAE up to index 3!
 o other:
   - image function for ode.2-D added.
   - changed warning printing in FORTRAN code
   - common interface for radau and daspk:
     both can solve systems written as M*dy = f(x,y).
     daspk can also solve systems written as 0 = g(x,y,dy) (=default for daspk)
   


