on the laptop of @roesel.
fixes https://github.com/prusa3d/Slic3r/issues/672
The Intel HD Graphics hangs on the glFinish() call for some reason
with one particular graphics driver revision. Also the glFinish() call
was superfluous and it only may have had negative effect over the performance.
Both glFinish() and glFlush() were removed for performance reasons
where they were not needed, see
https://www.khronos.org/opengl/wiki/Common_Mistakes
A regression error has been introduced into Slic3r 1.38.xx series
for the float/percent config value, where the value was considered
unchanged if the percent sign has been added or removed.
the point type on Windows / Linux / OSX
to achieve the same behavior on all the 32 / 64bit systems.
(Windows always treats the long as 32bit int, while Linux treats
long as a 64bit int).
The problem was caused by the gap fill algorithm, which worked
with square extrusion width values as with rounded extrusion widths,
which sometimes lead to negative extrusion cross sections
for high height to width ratios.
The extrusion width logic has been changed to consider the input width
to be the extrusion spacing, not the extrusion width. The change certainly
removed the negative feed rates, but it also certainly increased the gap
fill width to some exent. It needs to be verified now, whether the gap fill
does not extrude too much.
https://github.com/prusa3d/Slic3r/issues/677
The previous way of checking that _putenv_s is defined does not work,
because _putenv_s is a function and not a define. This is a partial
application of commit 31115e0369747b1e1c45cbe3f2a90f6dff66666a from
alexrj/Slic3r.
I tried cherry picking the whole commit, but unicode is already handled
diffrently here, so that would have been a lot of work.
due to the differences in the M203 code
(RepRap firmware has it in mm/min, Marlin in mm/sec).
This difference is important to the G-code time estimator.
Changed the g-code flavor to Marlin for all Prusa3D bundled profiles.
