Replace the large macro HOMEAXIS with a function. This avoids the
compiler generating three copies of largely identical code. The
saving is 724 bytes of program memory.
We make use of XYZ_CONSTS_FROM_CONFIG to provide convenient
array-shaped access to MAX_LENGTH, HOME_RETRACT_MM and HOME_DIR.
Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>
SERIAL_ECHOPAIR implies, eventually, two calls to MYSERIAL.print. One
of these has FORCE_INLINE for a per-character loop, and both involve
constructing a method call rather than a simple function call.
Produce better and smaller code by providing three specialised
functions serial_echopair. This saves 672 bytes of program memory
(with EEPROM_SETTINGS and SDSUPPORT enabled).
Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>
Previously the parameters set in M206 would only be used if a G82
command was sent with specific axis home values. This limits its
usefulness.
Really, we should have a way to adjust the XYZ homing of a machine in
the eeprom. So as the first stage of this, make M206 affect every
home command. The values set using M206 are now added to the
configuration variables [XYZ]_HOME_POS.
This is achieved by replacing all uses of [XYZ]_HOME_POS in the code
by a new home_pos[] which includes the adjustment. We also have to
adjust the uses of [XYZ]_{MIN,MAX}_POS similarly - see below.
To allow axis_is_at_home to be written as a function taking an axis
index rather than a macro taking an axis letter, we provide
constant arrays in program memory containing the values of
[XYZ]_{MIN,MAX,HOME}_POS from the compiled-in configuration.
This is done with some helper macros to deal with the declaration
(XYZ_CONSTS_FROM_CONFIG) and definition of the inline function which
does the program memory access.
We also introduce the overloaded function read_pgm_any, whose
instances are produced with DEFINE_PGM_READ_ANY, which allows the
access functions to automatically produce the correct type.
The type- and pointer-massaging code in the access function boils
down, when compiled, to a simple program memory access.
A question arises: if the M206 offset is set, should this adjustment
to the home position shift or change the possible range of movement
permitted by the software endstops ?
The documentation in Configuration.h describes these limits as:
// Travel limits after homing
Since this is a file containing physical limits, and actual suggested
values for these configuration parameters appear to include a certain
amount of slop, I've taken the view that these should be regarded as
nominal physical distances from the limit switches, and that the
permissible travel should be unaffected by M206.
So for example with the (rather unrealistic)
#define X_HOME_DIR -1
#define X_MIN_POS -20
#define X_HOME_POS 0
#define X_MAX_POS 100
no matter the setting of M206 X, the machine would be permitted
to move from 20mm "beyond" the limit switch trigger point in
the negative X direction and 100mm away from the limit switch in
the positive X direction, for a total travel of 120mm.
With M206 X-10 that would be considered to correspond to X coordinates
-30 to +90. With M206 X+10 that would be considered to correspond to
X coordinates -10 to +110.
fixes#200 (in ErikZalm/Marlin).
Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>
If [XYZ]_HOME_POS and [XYZ]_MIN_POS aren't 0, these corrections are
wrong. Use the same logic as in Marlin.pde:prepare_move: ie, clamp to
[XYZ]_{MIN,MAX}_POS.
While we're here, put this cut-and-paste code in a function
clamp_to_software_endstops.
Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>
Emacs by default doesn't recognise a ".pde" file as C++ source code.
Add the annotation to the top of the file to make it work.
Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>
HOME_POS is now always where the endstop is and can be outside the limits.
The limits are now defined by MIN_POS and MAX_POS rather than HOME_POS and MAX_LENGTH.
The Z is axis now homed first if direction is away from the bed.
Saguinololu limit pins change from MIN to MAX according to the homing direction.
Since the class "MainMenu" was used within a static variable the
initialization of the object (constructor call) was done before Arduino
library startup. It always caused a crash when using AVRStudio with
JTAG debugger (caused from calling the LCD initialization / the lot of
I/O work / the stack used during this calls). By moving the LCD_INIT
out of the constructor and using an explicit call inside of Arduino
setup() implementation immediately fixed all problems and the JTAG
debugger runs fine.
