FMSLogo Code

Fix a problem in BITBLOCKTEST.SMOKETEST where, depending on where the turtle was, it would do the BITLBLOCK in a different location, which would make the screenshot comparison fail. Now the test runs CLEARSCREEN to enforce that the turtle is always in the same location.

Update the expected test log after a recent test run.

Fix bug #486; The turtle can now walk along the screen extents in FENCE mode.

The problem was that if you tried to walk along the extents, due to the impercision of floating point arithemetic, you might end up crossing the extent by 1e-14 or something similarly small (especially after conversion from degress to radians).

My fix was to permit the same "rounding" tolerance that POS uses when showing the true turtle position. My thinking is that this tolerance is used to mask very small amounts of user-facing of imprecision that are the turtle's position, the screen extent is another user-facing boundary, so it should use the same precision.

Fix a regression that broke SETPIXEL on windows (that is, it made it as bad as it on GTK3). The problem was a typo in the "__WXMSW__" conditional compilation.

Refactoring: Add comments to wrap_* functions.
Replace > operations with equivalent < operations.

Work around a wxWidgets bug where wxDC::DrawPoint() draws two
pixels under GTK3. On this platform, a zero-pixel line is drawn
instead. This is not quite correct and causes some anti-aliasing
which bleeds into the pixel, but it's better than before.

Add a regression test to ensure that WRAP does not limit the distance that FORWARD is permitted (opened as bug #487)

Update a bug number to use SourceForge's new numbering scheme.

Rewrite SETPIXEL and PIXEL in terms of platform-independent wxWidgets function calls so that the work on GNU/Linux.

Replace usage of the win32 POINT struct with wxWidget's wxPoint. This makes the code more platform-indepenent although, by itself, doesn't make any new functions availabe on the GNU/Linux build.