1-sept-1996.  In spite of my new job at the KNMI, I still fully support 
the FF library.

This is the prerelease of the new version of FF, version 2.0.  The complex 
routines are being held back, as these are as yet completely untested.  The 
main differences with the normal version are:
- better handling of non-IR masses which are zero (typically neutrinos)
- faster IR divergent boxes using the algorithms of Beenakker & Denner
- B0' (ffxdb0, note that it also returns p^2*B0' to avoid an undefined result 
  for p^2=0)
- the sixpoint function F0
- some tensor functions: B2, D2 (the others will follow)
- hooks for complex functions
- possibility to input the dotproducts (ffxd0d); experimental.
Please report any problems you might have to me, gjvo@xs4all.nl or 
t19@nikhef.nl.  Without this feedback I have a hard time killing all bugs.

The calling sequences for internal functions have been changed completely, so 
it is not possible to mix routines from theis version with previous ones.  
Also some error messages have been renumbered, so you cannot even mix the 
ffwarn.dat and fferr.dat files.

Geert Jan van Oldenborgh

Bug fixes:
- 12-07-1991: inconsistent flagging in ffxc0p.f and ffxdi.f caused undefined 
  arguments to be used in some cases.  l4also=.FALSE. helps, or the new files.
- 15-07-1993: better error messages in case of dependent momenta.
- 19-07-1993: fixed typo in ffzdbd which caused ffzzdl to be called with one 
  argument too many.  new file ffzdbd.f
- 12-10-1993: ffxc0 gave a spurious error when called with 3 spacelike momenta
- 1-dec-1993: fixed many problems with the error system, for the time being 
  only in the routines with real masses.  Added B2.  B1' arguments changed!
  Also added the possibility to input the dotproducts: ffxd0d, ffxc0d.  Please 
  report problems with this scheme if you use it.  Fixed bug in B0' for 
  xp=0.
- 1-dec-1993.  AA routines added (tensor reduction).  The B-reduction is now 
  completely stable; I am working on the other ones.  Note that instabilities 
  in the aa routines are not yet reported.  B1' still missing.
- 1-dec-1993.  As the new ff.h file is different from the old one you'll have 
  to recompile everything.
- 2-dec-1993.  Further cleaning in the error reports.
- 18-jan-1994.  Fixed a few typos in ffzb0 (file ffcb0.f), wrong # of args.
- mrt-1994. Fixed a bug in ffxb0 which caused spurious error messages when 
  lwarn was off.
- 8-aug-1994. Fixed a stupid bug in ffcxyz.f, which caused undefined values to 
  be used, thus giving (very) wrong answers if l4also was .true..  Updated 
  some files to use input ier when checking to beat back on spurious error 
  messages.
- 25-mar-1995. Fixed a bug in ffcxr.f, when a Taylor expansion was made some 
  eta terms could be undefined.
- 19-apr-1995. Fixed a bug in ffxd0, if a host of conditions was met the 
  imaginary part would be off by a term i\pi^2.  Improved error checking.
- 22-aug-1995. Fixed a row of bugs which appeared in Z -> gamma gamma gamma 
  with equal masses, mainly extra terms 2\pi^2 which caused the answer to be 4 
  orders of magnitude too large.  Some of these gave error messages; other did 
  not.  New ffxd0.f, ffdel4.f, ffdcxs.f, ffcxs3.f, ffxc0p.f, ffdxc0.f.
- 22-sep-1995.  Fixed another bug in Z->3gammma, close together algorithm did 
  not check for different i*pi^2.  Still wrong for on-shell photons, use 
  p^2=-1d-10 for the time being.
- 3-oct-1995.  Fixed rare bug that caused the Hill identity to be used when it 
  should not, giving an error message from ffzli2 that the argument is too 
  large.  Actually harmless numericallyy.  New ffcrr.f
- 6-oct-1995.  Fixed another equal-masses bug in the C0, answer was off by 
  pi^2/3 due to typo.  New ffxc0p.f.
- 16-oct-1995.  An i*eps problem which occurred in the equal masses case gave 
  the wrong imaginary part.  CHECK AGAINST l4also=.FALSE, ldc3c4=.FALSE.
  New ffcxs4.f, ffcrr.f; put warning in ffxd0h.f
- 1-dec-1995.  When p4^2=0, m2>m1 the IR C0 (using delta) would have the wrong 
  sign.
- 9-mar-1996: added the complex tensor reduction functions aacbc.f, aaccc.f, 
  ffcb2.f ffcb2p.f, aacinv.f, ffcdb0.f
- 13-mar-1996: killed a bug in ffdcxs, would appear in cases with equal masses 
  in versions newer than 22-aug-1995.  New ffdcxs.f.  Also synchronized with 
  my private copy, this entails a new ff.h for the 6-point function.  Please 
  recompile everything.
- 18-mar-1996: Fixed a stupid typo bug in ffcb0, up to now all quantities with 
  complex p^2 but real m^2 would use only the real part of p^2.  New ffcb0.f
- 22-mar-1996: Killed a bug in 13-mar's bugfix in ffdcxs.f.  Should really be 
  OK now.  Fixed a few warnings with ftnchk'ing, and a typo in the testing in 
  ffxdi.f
- 27-mar-1996: On request of CERN people, changed ffinit to ffini to avoid 
  conflict with the FFREAD tape handling package.  A dummy subroutine ffinit 
  that just calls ffini is included for older programs, if you want to link 
  against the CERN paclib you should NOT include file ffini.o in ff.a!
- 28-mar-1996. I think I have all the continuations for p^2 complex correct 
  now.  New ffcb0.f
- 4-jun-1996: Added a few safety checks in ffxc0, ffcb0, fxc0i.
- 16-jul-1996: Added a check in ffxb2q to not use an algorithm which divides 
  by xm1 when xm1.eq.0.  New ffxb1p.f
- 15-aug-1996.  Got rid of the last two instances of ffinit.  
- 23-jan-1997.  The IR-divergent routines now complain when the user attempts 
  to evaluate a mass singular D0.
- 21-jul-1997.  Moved the archive back to NIKHEF; I hope for good.
- friday-13-mar-1998: Set xalogm, xclogm to their IEEE value when the 
  optimizer kills the loop and returns zero ffini (ffinit.f)
  Added word 'path' to ffopen (ffinit.f)
  Put lwrite on warning (ffxb1.f)
- 1-oct-1998: Fixed ier bug in aaxcx (was too high for level=3), fixed
  makefile problems, updated ffmanual.tex, updated npointes slightly.