klipper-dgus/docs/Debugging.md

149 lines
4.7 KiB
Markdown

The Klippy host code has some tools to help in debugging the firmware.
Translating gcode files to firmware commands
============================================
The Klippy host code can run in a batch mode to produce the low-level
firmware commands associated with a gcode file. Inspecting these
low-level firmware commands is useful when trying to understand the
actions of the low-level hardware. It can also be useful to compare
the difference in firmware commands after a code change.
To run Klippy in this batch mode, there is a one time step necessary
to generate the firmware "data dictionary". This is done by compiling
the firmware code to obtain the **out/klipper.dict** file:
```
make menuconfig
make
```
Once the above is done it is possible to run Klipper in batch mode
(see [installation](Installation.md) for the steps necessary to build
the python virtual environment and a printer.cfg file):
```
~/klippy-env/bin/python ./klippy/klippy.py ~/printer.cfg -i test.gcode -o test.serial -v -d out/klipper.dict
```
The above will produce a file **test.serial** with the binary serial
output. This output can be translated to readable text with:
```
~/klippy-env/bin/python ./klippy/parsedump.py out/klipper.dict test.serial > test.txt
```
The resulting file **test.txt** contains a human readable list of
firmware commands.
The batch mode disables certain response / request commands in order
to function. As a result, there will be some differences between
actual firmware commands and the above output. The generated data is
useful for testing and inspection; it is not useful for sending to a
real micro-controller.
Testing with simulavr
=====================
The [simulavr](http://www.nongnu.org/simulavr/) tool enables one to
simulate an Atmel ATmega micro-controller. This section describes how
one can run test gcode files through simulavr. It is recommended to
run this on a desktop class machine (not a Raspberry Pi) as it does
require significant cpu to run efficiently.
To use simulavr, download the simulavr package and compile with python
support:
```
git clone git://git.savannah.nongnu.org/simulavr.git
cd simulavr
./bootstrap
./configure --enable-python
make
```
Note that the build system may need to have some packages (such as
swig) installed in order to build the python module. Make sure the
file **src/python/_pysimulavr.so** is present after the above
compilation.
To compile Klipper for use in simulavr, run:
```
cd /patch/to/klipper
make menuconfig
```
and compile the firmware for an AVR atmega644p, disable the AVR
watchdog timer, and set the MCU frequency to 20000000. Then one can
compile Klipper (run `make`) and then start the simulation with:
```
PYTHONPATH=/path/to/simulavr/src/python/ ./scripts/avrsim.py -m atmega644 -s 20000000 -b 250000 out/klipper.elf
```
Then, with simulavr running in another window, one can run the
following to read gcode from a file (eg, "test.gcode"), process it
with Klippy, and send it to Klipper running in simulavr (see
[installation](Installation.md) for the steps necessary to build the
python virtual environment):
```
~/klippy-env/bin/python ./klippy/klippy.py config/avrsim.cfg -i test.gcode -v
```
Using simulavr with gtkwave
---------------------------
One useful feature of simulavr is its ability to create signal wave
generation files with the exact timing of events. To do this, follow
the directions above, but run avrsim.py with a command-line like the
following:
```
PYTHONPATH=/path/to/simulavr/src/python/ ./scripts/avrsim.py -m atmega644 -s 20000000 -b 250000 out/klipper.elf -t PORTA.PORT,PORTC.PORT
```
The above would create a file **avrsim.vcd** with information on each
change to the GPIOs on PORTA and PORTB. This could then be viewed
using gtkwave with:
```
gtkwave avrsim.vcd
```
Manually sending commands to the micro-controller
-------------------------------------------------
Normally, Klippy would be used to translate gcode commands to Klipper
commands. However, it's also possible to manually send Klipper
commands (functions marked with the DECL_COMMAND() macro in the
Klipper source code). To do so, run:
```
~/klippy-env/bin/python ./klippy/console.py /tmp/pseudoserial 250000
```
Generating load graphs
======================
The Klippy log file (/tmp/klippy.log) stores statistics on bandwidth,
micro-controller load, and host buffer load. It can be useful to graph
these statistics after a print.
To generate a graph, a one time step is necessary to install the
"matplotlib" package:
```
sudo apt-get update
sudo apt-get install python-matplotlib
```
Then graphs can be produced with:
```
~/klipper/scripts/graphstats.py /tmp/klippy.log loadgraph.png
```
One can then view the resulting **loadgraph.png** file.