Model here: http://www.thingiverse.com/thing:13408
Model prints mostly fine (way better than SF :) ), but below each pillar for 1 or 2 layers there is space left in in the infill. It actually tries to print the outlines of the columns as if they are inset into the model versus starting at the surface layer. Similar thing happens with Hollow Pyramid Short on thingiverse.
This is 11/4 build!
I'm seeing this as well with the Hollow Calibration Pyramid (http://www.thingiverse.com/thing:11846). It almost seems like the logic for determining infill is taking precedence over the logic for making bottom layers solid. So it ends up carving out an area of the otherwise solid layers and making it non-solid.
Can you please post a screenshot of the wrong layer(s)?
You can preview the print from Pronterface: load the file and click on the preview window; there you can view all layers.
Here you go: http://www.newdiversions.com/skimages//pyramid_layer_issue-20111110-095424.jpg . You can see the "hole" left for the pillar before it starts, this lasts a couple of layers.
Sorry that's of the pyramid from Pointedstick, here's the original model with the round pillars: http://www.newdiversions.com/skimages//retract_model_layers-20111110-095700.jpg
The pyramid thing there is exactly what I see as well.
Please post your config.ini and also any command-line arguments used.
bottom_layer_speed_ratio = 0.3
duplicate_distance = 6
duplicate_x = 1
duplicate_y = 1
end_gcode = M84 ; disable motors
filament_diameter = 1.76
filament_packing_density = 1.056
fill_angle = 0
fill_density = 0.1
infill_every_layers = 1
layer_height = 0.2
nozzle_diameter = 0.35
perimeter_feed_rate = 30
perimeter_offsets = 3
print_center = 60,60
print_feed_rate = 45
retract_before_travel = 2
retract_length = 1
retract_lift = 0
retract_restart_extra = 0
retract_speed = 18
rotate = 0
scale = 1
skirt_distance = 6
skirts = 3
solid_layers = 3
start_gcode = G92 X0 Y0 Z0 E0 ; reset distance
temperature = 185
travel_feed_rate = 100
use_relative_e_distances = 0
z_offset = 0
This might be fixed (I didn't test it). Can you test it with current git or new packages once they're out? Thank you! :)
Here  is an illustration of the pattern that is breaking the "retract test.STL" object. The in-fill is somehow breaking into the solid-layers. The illustration was made with today's Slic3r (20111114) using 100% default settings.
Hello Peter, thank you for the excellent screenshot. That helped me much.
I regret to say that it's not a bug but an intended behaviour: those areas in layers 2 and 3 are not given a solid infill because they're internal areas (no top, no bottom). The solid infill is only drawn in the areas that are actually exposed to open air: those areas below the pillars are not exposed to open air. Let me know if my explanation wasn't clear enough.
That said, we could have encountered a situation (pillars/small areas) where this behaviour is dangerous for the quality of the print. While it makes sense for most of cases, you're making me think about handling this particular situation as an exception.
My question is: does this impact print quality?
Do you have pictures of bad results?
If so, the solution is to detect these situations (small internal surfaces inside solid surfaces) and merge them in order to have a single solid surface. (Note to self: small could be defined as those surfaces which cannot be offsetted by the number of perimeters + n.)
Print solid infill without cutting an internal surface under small re…
…gions such as pillars. #39