FMSLogo Code

Update the documentation for ELLIPSEARC2 to match fix for Bug #494.

Fix the documented order of the X and Y inputs for ELLIPSEARC2.
The last time I looked at ELLISEARC2, I had convinced myself that the order was swapped, but taking a closer look at the ELLIPSEARC2 behavior, I realize that I was wrong (or at least I was wrong the way these inputs are documented).

This is good news, because X Y makes more sense than Y X.

Finish fixing Bug #494; ELLIPSEARC2 now leaves at the point where the ellipse stopped.

This updates ELILPSEARC to account for the fix to ELLIPSEARC that makes it correctly honor the angle.

Partial fix for Bug #494 (ELLIPSEARC2 problems), as submitted by Francesco Simone Carta.

This would have been a complete fix in FMSLogo 7.7.0, but I have since changed ELLIPSEARC to calculate the angle of the sweep correctly, which made it so, in this implementation, the turtle either glides ahead or behind the drawn arc (but remains on the ellipse, which is still an improvement).

This also brings in a breaking change for when the third parameter is negative. The old ELLIPSEARC drew the arc on the opposite side of where the turtle is located. The new implementation draws the ELLIPSEARC sweeping in the opposite direction. The new behavior is more intuitive.

Fix Bug #533; ELLIPSEARC and ELLIPSEARC2 now treat the angles correctly.

The problem was that the "t" in the equation (x,y) = (a cos(t), b sin(t)) is not the angle of the ellipse, but the angle of a circle that is then squashed/elongated into an ellipse. To account for this, some math is now done to translate the given angle into the "t".

This fix may change the behavior of programs which give angles that aren't multiples of 90 degrees.

Add some regression tests for ELLIPSEARC, adapted from the ELLIPSEARC2 tests.

Update the message about why -flto cannot be used.
The latest mingw doesn't crash when given -flto, but I cannot make it work with wxWidgets.

Cleanup: Remove untranslated COPYDEF instructions from the Italian startup script.
All but one of these were commented out and so obviously did nothing.
The one that wasn't commented out redefined a NOREFRESH to itself, vacuous in itself, but also because NOREFRESH is a no-op in FMSLogo.

The primary motivation for this change is to remove warnings emitted by make-translation-tables.pl.

Cleanup: remove [copydef "px "pi] from Italian startup script.
This throws an error because px is a primitive.
It's also unnecessary since PX and PI are both already synonyms for PENREVERSE in the Italian FMSLogo.

Remove the OUTOFMEMORY tests that might fail (because of no overflow) on Windows.
This removes false-postive failures from the test pass.