Change History

1.8	06-Sep-2000	New distribution organization, with recursive make
			scripts; utilities more properly organized, all the
			tools now build off libmpi.a in the main directory.

1.8.1	24-Sep-2000	Tweaked mp_add() and mp_sub() so that they will only
			allocate temporaries when they really need to.  Also
			played around in s_mp_mul() and s_mp_sqr() to get rid
			of several indirections that slow down the code even
			with a good optimizer.  This seems to have improved
			the performance of multiplies by as much as 20% in 
			ad-hoc testing.

1.8.1p1	17-Jan-2001	Fixed a bug in mp_div_d() where it would try to fix
			the sign of zero on the output parameter, rather than
			on the computed quotient (this caused failure in the
			case where the output quotient was NULL, allowed by
			the API to ignore the quotient).  Thanks are due to
			scroussette@yahoo.com for spotting this.

1.8.1p2	23-Jan-2001	Various tweaks to improve performance.  This results
			in significant speedups in some operations.  Fixed
			a small, rare memory leak in mp_exptmod().  Tweaked
			the configurations so you can select how many primes
			you get when you build mpprime.c.  Timing results
			updated to reflect the performance improvement.

1.8.1p3 31-Jan-2001	Fixed a sign-related bug in mp_sub().  This would
			cause the sign of the output to be incorrect in the
			case where mp_sub(a, b, c) was computed, a < 0, 
			b > 0, and neither a nor b is the same mp_int as
			c.  Thanks to Tushar Udeshi <tudeshi@zyvex.com> for
			finding this bug!

1.8.1p4 17-Feb-2001	Fixed a sign-related bug in mp_mul().  This would
			cause the sign of the output to be incorrect in the
			case where the output variable was equal to the first
			input, e.g. mp_mul(a, b, a), and a < 0.  The trouble
			was in not computing the output sign before calling
			the internal unsigned multiply function, which uses
			a temporary for output and replaces a.

1.8.1p5 24-Feb-2001	Fixed a precision-related bug in s_mp_mul_d(), the
			internal function implementing mp_mul_d().  In some
			cases, when precision extends to an additional digit
			beyond the existing precision, the code neglected to
			register the increase in the mp_int structure, so 
			that the digit was lost during clamping.

1.8.2   27-Feb-2001     Added mp_tomag() and mp_mag_size() functions, to 
                        deal with output of unsigned magnitude.  The mp_toraw()
                        function is rewritten to use it.  This should make it 
                        easier to implement things like PKCS#1 padding, where
                        only the magnitude is desired, and no sign indication.

1.8.2p1 01-Aug-2001     Changed 'char' to 'unsigned char' where appropriate, to
                        fix the buglet with output on systems that have signed
                        characters by default.  You could also circumvent this
                        by telling your compiler to use unsigned characters, but
                        this takes care of that issue.  Of course, this causes
                        lots of warnings when compiling mpi-test, but they are
                        spurious warnings, and can be ignored.  Also fixes a bug
                        with mp_tomag() and mp_mag_size() which would not handle
                        zero (0) correctly.

1.8.3	25-Aug-2001	Changed the names of the mp_toraw() and mp_tomag()
			functions to mp_to_signed_bin() and mp_to_unsigned_bin().
			This means more typing, but the names are clearer in 
			meaning (I had some folks writing in confused).  Also,
			mp_raw_size() is now mp_signed_bin_size(), and, as you'd
			expect, mp_mag_size() is mp_unsigned_bin_size().

			The old names are defined as macros, if you enable the
			MP_COMPAT_MACROS flag in mpi-config.h.  It is off by 
			default, though.  This release also adds two new 
			functions, mp_init_array() and mp_clear_array().

1.8.4   14-Sep-2001	Added mp_mul_2d() and mp_count_bits().  Changed the way
			mp_radix_size() computes the size of the buffer, so that
			it is now exactly the size required.

1.8.4p1	28-Dec-2001	Fixed a subtle but important bug in mp_mul() that would
			give incorrect results when the target variable was the
			same as the multiplier.  Thanks to Tom St. Denis for 
			spotting this one.

1.8.4p2 03-Jan-2002	Fixed a subtle bug in s_mp_div() whereby the "guessed"
			digit of the quotient could be GREATER than the radix,
			not just equal to it, and in that case, you got the 
			wrong answer.  Thanks to Tom St. Denis for
			identifying this bug, and to Bryan Olson for
			spotting where I had gone wrong.

1.8.5   24-Dec-2002     Fixed sign-related bugs in several routines whereby a
                        zero value would sometimes be incorrectly left with a
                        negative sign flag.  Fixed mp_gcd() which was giving 
                        the incorrect sign on output to gcd(0, x) with x < 0.

                        Thanks to Tom St. Denis for bringing these bugs to my
                        attention.

1.8.6	20-Oct-2003	Fixed a bug in s_mp_outlen() which would generate the
			wrong number of digits for some cases.  Thanks to 
			Miguel Oyarzun for catching this one.

			Fixed a bug in mp_lsh(), in mplogic.c, where small 
			shifts would not be done properly.  Thanks to 
			Kenichiro Katayama for finding this one.