[Mspgcc-users] Mayday ! Insn errors ???

[Matthew . Cook]

Howdy...

It seems that my code has started giving msp430-gcc a fur ball, which is
choking upon.  This morning porting some functions from one of my previous
avr-gcc projects it started generating insn errors which I assume are bad !

Here is the scenario...   My project is basically split into multiple
source files.  In my adcv.c module you will find the following function.

----------------------------------------------------------------

void ADCV_InitChannel( u08 Channel, u16 InitialValue, \
u08 SampleFrequency, u08 TimeConstant, u08 Flags)
{
    /* code removed to protect the innocent */
    return;
}

In the adcv.h header file;

extern void ADCV_InitChannel( u08 Channel, u16 InitialValue, \
u08 SampleFrequency, u08 TimeConstant, u08 Flags);

note:

typedef unsigned char     u08;
typedef unsigned short    u16;

----------------------------------------------------------------

In my other modules (eg temp.c) the function above is called like so;

void  TEMP_InitSW(void)
{
    ADCV_InitChannel(ADCV_TEMP_Channel, TEMP_DEFAULT_ADC, \
                     TEMP_SAMPLE_FREQ, TEMP_FILTER_TC, \
                     ADCV_ENABLED);
    return;
}

where

#define ADCV_TEMP_Channel     0
#define TEMP_DEFAULT_ADC      1936
#define TEMP_SAMPLE_FREQ      60
#define TEMP_FILTER_TC        4
#define ADCV_ENABLED          (1<<0)

Note the sample freq, filter timeconst and default filter values change
depending on module calling the Init Function.

----------------------------------------------------------------

Now from the output window I get the following when I try to compile the
above;

msp430-gcc src\adcv.c -g -Os -Wall -fshort-enums -mmcu=msp430x447 \
    -Iinc -Wa,-ahlms=tilst\adcv.lst -c -o tiobj\adcv.o
msp430-gcc src\temp.c -g -Os -Wall -fshort-enums -mmcu=msp430x447 \
    -Iinc -Wa,-ahlms=tilst\temp.lst -c -o tiobj\temp.o
src/temp.c: In function `TEMP_InitSW':
src/temp.c:190: unrecognizable insn:
(insn 63 62 64 (set (mem/f:QI (pre_modify:HI (reg/f:HI 1 r1)
                (plus:HI (reg/f:HI 1 r1)
                    (const_int -2 [0xfffffffe]))) [0 S1 A8])
        (const_int 1 [0x1])) -1 (nil)
    (nil))
src/temp.c:190: Internal compiler error in extract_insn, at recog.c:2148
Please submit a full bug report,
with preprocessed source if appropriate.
See <URL:http://www.gnu.org/software/gcc/bugs.html> for instructions.

----------------------------------------------------------------

My testing has shown that this error revolves around the number of
variables I am passing to the ADCV_InitChannel() function.   If I remove
one variable ie "u08 Flags"  then the code compiles without the insn error.
eg see below;

void ADCV_InitChannel( u08 Channel, u16 InitialValue, \
u08 SampleFrequency, u08 TimeConstant)
{
    /* code removed to protect the innocent */
    return;
}

void  TEMP_InitSW(void)
{
    ADCV_InitChannel(ADCV_TEMP_Channel, TEMP_DEFAULT_ADC, \
                     TEMP_SAMPLE_FREQ, TEMP_FILTER_TC);
    return;
}

----------------------------------------------------------------

It does not matter which variable I remove from the function call, I only
have to reduce the number of variables by one byte and it works.  I do not
believe it to be a variable type problem, just the number I am trying to
pass.

I have used this code previously and I am happy that all my chickens and
eggs line up with respect to the relevant includes and dependencies between
my source modules.

I know that the above function is not efficient, I could pass a pointer and
prevent this situation, but I was under the impression there were no
limitations on how many variables can be passed to a function. ie once the
number of available registers were consumed the compiler would start using
the stack.

For the time being I will revert to using a pointer to pass the relevant
data, but I am hoping this is not a bug ?

Cheers

Matthew