-------------------- AmigaOS 3.1.4.(1) project -----------------------

Changes for V45.6 (28.12.2016)

- Added a magic libbase replacement that automatically switches the
library base if the math hardware changes, for example through setpatch.
No need for setpatch to play with the library base anymore.

- Re-did the integration into the V45 build from the V50.1 version in
a better way. No longer depends on the hx68 assembler, now uses
clear-text FPU primitives.

- Now builds with SAS/C and its assembler.

Changes for V45.7 (2.1.2017)

- Apparently, the magic library replacement could have resulted in
situations where the dead library base is removed while some programs
are still using it, and hence causing a crash. This happens only if
a program re-opens the math.library, but does not update its internal
pointers, and then closes the library through the old, non-updated
pointer, finally pulling the rug under its feed. Arexx seem to be
notorious about this, but other programs may exist that assume that
the library base stays constant. Thus, to avoid any further trouble,
the obsolete library stays resident in memory forever, causing a slight
loss of memory.

Changes for 45.8 (4.3.2018)

- Since the latest version of setpatch removes copies of mathieeesingbas
prior loading the CPU libraries, we will now (hopefully) always get
the right version anyhow, so the version check is no longer required.

Changes for 45.9 (6.5.2018)

- Due to a misconfigured assembler, some harmless code was only runable
on a 68020 or beyond. Fixed that.

---------------------- AmigaOS 3.2 project ---------------------------

Changes for 47.1 (20.6.2019):

- Changed the rounding mode for double to single precision conversion to
  IEEE aware round to nearest.


Changes 47.2 (24.8.2020)

- The DOUBLE and FLOAT types should not be redefined after they have
  already been set up by <exec/types.h>. The internal structure
  definitions of these types in "math.h" are now called INTERNAL_DOUBLE
  and INTERNAL_FLOAT, respectively.
