Refactoring of GCode export of color changes, extruder switches etc,
so that the "color change" like extruder switches are applied first
at the Wipe Tower / G-code export, so that adding / removing
an extruder switch at the G-code preview slider does not invalidate
slicing.
1) Don't allocate ExtruderOverrides if not necessary
2) Use boost::container::small_vector<int32, 3) for ExtruderOverrides
(usually less than 4 instances are printed))
If the Elephant foot compensation is applied to the 1st object's layer,
the uncompensated 1st object's slice is newly used for calculation
of bridges, overhans, skirt, brim, raft and supports.
Layer::slices were renamed to Layer::lslices to simplify reading
of the code, to differentiate from LayerRegion::slices.
Fixed layouts after the second opening of the dialog.
(Removing any extruder from the sequence does not cause an incorrect layout)
Validation of entered values added
(0 is not a valid value)
1) Start PrusaSlicer
2) Select printer: Original Prusa i3 MK2.5
3) Load any stl file
4) In Printer Settings turn on "Support stealth mode"
5) Press "Slice now"
-> crash
All the "machine_XXXXX" vectors in the print config contain a single value instead of two leading to a vector subscript out of range
when read into GCode::_do_export()
Layer newly remembers bounding boxes of slices,
the bounding boxes are used by G-code generator & newly the support
generator.
Slices are stored as ExPolygons, not ExPolygonCollection.
for islands with another islands in their holes.
Improvement of chaining of infill lines for 3D honeycomb, Gyroid and
Honeycomb infill: New TSP chaining algorithm is used.
to avoid layers thinner than min_layer height, and possibly due to
the fact, that empty support extrusions may get generated
for support layers with non-empty support polygons.
Fix of Empty layers detection added to GCode.cpp (6ab1cec)
If the custom toolchange gcode switches extruder to the one that is expected, PrusaSlicer does not issue its own T command
Otherwise, this command is inserted just after the custom gcode, even in case it is not empty
This should prevent issues such as #2884 and #2886 from being reported
WARNING -> After this commit the exported gcode will contain the extra lines used by the time estimator. They will be removed by a future commit when a new post-process method will be implemented.
atomically. The code was taken from the llvm project, it is complex
and hopefully it covers all the Windows file system quirks. Vojtech
has highest hopes, that this will fix the various PrusaSlicer.ini
file corruptions.
Enabled the locales switching and error handling on Linux as well,
where now the missing locales are reported and running the locale-gen
tool is recommended.
Function set_extruder can be called before moving to the first layer, m_layer_index is then -1. We definitely don't want to set temperature for second layer in that case.
Linear advance is reset by filament start gcode after a toolchange. However, not all moves to the wipe tower
end with a toolchange (brim, empty grid) and it would therefore disable linear advance until the next toolchange
This should solve https://github.com/prusa3d/PrusaSlicer/issues/2770
Reported in https://github.com/prusa3d/PrusaSlicer/issues/2752 and was a result of setting support contact z distance to negative value.
This lowered the maximum allowed print_z of the next layer, even though previous layer had object layers too, so the penalty from the contact z should not have been applied.
Fixed simply by rejecting the negative contact_z.
That could happen on empty support layers which do not necessarily matter, since their spacing is not generally synchronized with the object
The new hopefully correct logic is "if there are extrusions on a layer, check that last layer with extrusions is at most the new layer height below
This is a fixup of changes from 0de6e53 and 6ab1cec
for the GLVolumes before sending to the GPU driver. The following commits
were partially reverted:
4269c8b23c Removed GLVolume non-VBO rendering
d15698e21e GLVolume and GLIndexedVertexArray refactored to send data to gpu at the first render call
Namely, the GLVolume buffers are "shrink to size"'d before sending their
content to the OpenGL driver, and the vertex buffers are populated
as quickly as possible from the GLVolume, so that the same buffer is not
kept twice in RAM on systems, where the RAM is shared with the graphics
card.
Also the memory allocation reporting was improved for the GLVolumes.
Added detection of empty layers so the wipe tower doesn't trip on them (it is not printable anyway).
This should improve wipe tower reliability with supports, objects standing on edges, etc.
I also turned an assert into exception throw to prevent hard crashes and nonsense output.
