thatch/cammill

2D CAM tool for Linux, BSD, Windows, Mac

cammill's Issues

Treeview calls in batchmode

When using batchmode, the stderr is full of messages like

(process:10359): Gtk-CRITICAL **: 12:13:56.637: IA__gtk_tree_store_set_valist: assertion 'GTK_IS_TREE_STORE (tree_store)' failed

(process:10359): Gtk-CRITICAL **: 12:13:56.637: IA__gtk_tree_model_foreach: assertion 'GTK_IS_TREE_MODEL (model)' failed

I've gotten stack traces in gdb, this seems fairly straightforward unless the sort has a side effect that is being relied upon. Hint for gdb: https://stackoverflow.com/questions/5785902/how-do-i-get-gdb-to-break-on-a-glib-assertion-failure

Helix for just holes

Right now helix is an all-or-nothing option. When doing a combined holes+outline, would be nice to remove a step and simply select helix-holes or helix-interior.

Infinite loop when sign of Depth and Z-Step differ

Ability to select by layer

Similar to object-<n>-use, I'd like something layer-<name>-use to choose based on layer. This is probably an input-only param, and doesn't need to be recorded in the output file.

Nice to have would be a way to have a default of off, then include just the one layer I want, rather than disabling them individually. I will provide a sample dxf that includes layers.

The Dynamic Array Project

Basically, convert all the staticly allocated arrays into using malloc/realloc to grow them as necessary. This would let you load an arbitrary size dxf file, reduce the overall memory footprint (6GB!) and speed up a few of the worst netsted-loop algorithms.

I started on this, but it doesn't yet produce the correct output.

Fix static analysis or valgrind errors

For example, here's some valgrind output from a sample run: valgrind_errors.txt

Using valgrind or some other tool, try to find many of these and adjust loop conditions/have default values to avoid undefined behavior.

Number editing in treeview

Some minor issues here:

1. validation -- requires deleting before typing anew. only one decimal point and limits the number of digits after decimal point. (is the double-click behavior odd here?)
2. spinner + mousewheel results in gui lag. See cammill#52
3. display shows more decimal digits than you can input

Trapezoidal tabs

There are already square (cfg:Holding-Tabs-Triangle-Type: 0) and triangular (cfg:Holding-Tabs-Triangle-Type: 1); this should probably use the value 2, and will need an additional parameter for ramp angle. The code near P_T_TYPE in calc.c is where most of the changes will be.

Tab grid interacts oddly with Zero: 1

It appears to operate as if Zero: 0 is used; that is, the lower-left grid line is always even with the lower-left object edge.

Option for mouse controls to match openscad

I think the left mouse orbit, right mouse pan works well on a trackpad. If this works well enough it could be made the default.

Better lead-in and lead-out options

For example, you could follow the same path at some stepdown angle -- similar to helix, but not spread over the whole cut.

Unnecessary moves

This is from a pocketing operation on pipe_hanger. object 1 moves to near origin before cutting (and also doesn't include the full comment preamble):

(Object: #1)
(Layer: 0)
(Pocket-Operation)
(--------------------------------------------------)
G00
G00 X-11.000000 Y-13.175000
G00
G00 X4.080000 Y224.605000
G00 Z1.000000
G01 Z-1.000000 F500
G01 X6.620000 F2500
G01 Y229.685000
G01 X4.080000

Segfault while editing treeview

I don't have a good repro, just clicking around sometimes causes it. This build is too optimized to easily tell what the line of code is.

#0  0x00007ffff56a4eec in g_type_check_instance_cast () at /usr/lib/libgobject-2.0.so.0
#1  0x00005555555605b3 in gui_cell_renderer_param_editing_done ()
#2  0x0000555555573a17 in  ()
#3  0x00007ffff642a7cc in  () at /usr/lib/libgtk-x11-2.0.so.0
#4  0x00007ffff5680475 in g_closure_invoke () at /usr/lib/libgobject-2.0.so.0
#5  0x00007ffff569377d in  () at /usr/lib/libgobject-2.0.so.0
#6  0x00007ffff569c045 in g_signal_emit_valist () at /usr/lib/libgobject-2.0.so.0
#7  0x00007ffff569d030 in g_signal_emit () at /usr/lib/libgobject-2.0.so.0
#8  0x00007ffff6545235 in  () at /usr/lib/libgtk-x11-2.0.so.0
#9  0x00007ffff65515e2 in gtk_widget_send_focus_change () at /usr/lib/libgtk-x11-2.0.so.0
#10 0x00007ffff6551871 in  () at /usr/lib/libgtk-x11-2.0.so.0
#11 0x00007ffff655bef0 in  () at /usr/lib/libgtk-x11-2.0.so.0
#12 0x00007ffff655c140 in  () at /usr/lib/libgtk-x11-2.0.so.0
#13 0x00007ffff642a7cc in  () at /usr/lib/libgtk-x11-2.0.so.0
#14 0x00007ffff5680475 in g_closure_invoke () at /usr/lib/libgobject-2.0.so.0
#15 0x00007ffff5692f97 in  () at /usr/lib/libgobject-2.0.so.0
#16 0x00007ffff569c045 in g_signal_emit_valist () at /usr/lib/libgobject-2.0.so.0
#17 0x00007ffff569d030 in g_signal_emit () at /usr/lib/libgobject-2.0.so.0
#18 0x00007ffff6545235 in  () at /usr/lib/libgtk-x11-2.0.so.0
#19 0x00007ffff6428f1b in gtk_main_do_event () at /usr/lib/libgtk-x11-2.0.so.0
#20 0x00007ffff60a1d5e in  () at /usr/lib/libgdk-x11-2.0.so.0
#21 0x00007ffff55a3ca7 in g_main_context_dispatch () at /usr/lib/libglib-2.0.so.0
#22 0x00007ffff55a3f59 in  () at /usr/lib/libglib-2.0.so.0
#23 0x00007ffff55a4272 in g_main_loop_run () at /usr/lib/libglib-2.0.so.0
#24 0x00007ffff6427df3 in gtk_main () at /usr/lib/libgtk-x11-2.0.so.0
#25 0x0000555555571e10 in main ()

Postprocessor can only be selected by "number"

The postprocessors actually have filenames, e.g. /usr/lib/cammill/posts/EMC.scpost and the numbers it uses aren't stable between multiple installs (on my laptop, EMC=2, desktop EMC=0).

My usecase is in (cfg:Machine-Post: 0) which I'm using from a makefile, but it's also possible to specify as -mpt on command line.

Nice-to-have would be a way to list selectable postprocessors and/or reference them using a relative path to include in a git repo.

File load after first is sometimes wrong

This tends to not happen when reloading the same file, but I don't have a good repro. The symptom is that lines will sometimes be in the wrong place, or be skipped entirely.

Better pocketing strategy

There are two major deficiencies right now -- only doing X-major moves, and doing a full retract when passes don't line up exactly. We could do analysis of the angles and pick a better direction, even if it's just Y instead of X.

Limited holding tabs

Would be nice to specify an initial offset to holding tabs, e.g. right now we have X-Grid for their spacing starting from zero. If I have two of the same part horizontally each 100mm wide, and want tabs right in the middle, I would ideally like X-Offset 50 then X-Grid of 100.

Nice to have would be specifying exactly where I want the tabs by position.

Tiny segment offsetting problem

For example, where these circles don't quite intersect properly:

Lobes(5, 16, 6, 11, $fn=128);

Results in moves into the workpiece when it shouldn't.

Touch off on table option

Right now it's hard-coded that 0 is the work top, and it does set down until the configured (negative) depth. This is near the comment "offset for each depth-step" in calc.c (although the calls are to object_draw_offset_depth, it handles both opengl and postcam moves).

This same piece of code is responsible for #14

I think we should add another parameter here, maybe even called work top, and start there instead of zero. This is because we can't just repurpose depth (I typically cut until somewhat below zero, so we'd need a separate param)

A way to reference the last object

Right now there's

(cfg:object-0-use: 0)

to ignore the first object. I'd like to find the last, or maybe next-to-last, but don't have a good syntax for it.

Documentation fixes

There is both a commandline '-bm 1' parameter as well as cfg:Misc-Batchmode. This is missing from the wiki and --help.