RF12_COMPAT mode broken since Jun 1 checkin

[cmoore42]

The checkin on Jun 1 (adapted changes from fixInterrupts branch) broke the RF12_COMPAT mode. Turning on RF12_COMPAT in RF12.h results in a compile failure in RF69.c:

...\jeelib\RF69.cpp: In function 'uint16_t RF69::recvDone_compat(uint8_t*)':
...\jeelib\RF69.cpp:171:27: error: lvalue required as left operand of assignment
rxfill = rf12_len = 0;
^
But rf12_len is a #define in RF12.h, you can't assign 0 to it.

I'm not entirely clear why RF69.cpp is being compiled at all, but the only change I made was to set RF12_COMPAT to 1 and that caused the compile to fail.

[JohnOH]

#if RF12_COMPAT
#define rf12_len (rf12_buf[1] - 2)
extern volatile uint8_t rf12_buf[];

I think the attempted code line is:
rxfill = (rf12_buf[1] - 2) = 0;

Which looks odd to me too.

Regarding compilation of RF69.cpp, I believe Arduino compiles all modules in a library into its temporary files directory and only the linker decides if they need to be pulled into the final image.

[JohnOH]

I take it that you are using #define RF12_COMPAT 0 to trigger the almost undocumented fixed packet length transport mechanism. Since use of the RFM12B radio is controlled by:

#define RF69_COMPAT      0   // define this to use the RF69 driver i.s.o. RF12 

before the jeelib include in the main sketch.

[JohnOH]

If at some point we look at the fixed packet length transport mechanism again I believe there is a much simpler run time option available as in: ffafa00

[jcw]

The problem comes from this change - line 164 / 171.

I'm inclined to change it back so it compiles, but it may not be the right thing to do for fixed-size packets. Probably better to break it and fix the most common case...

[JohnOH]

I did think up a run time fix for packets not needing the Jee header. There is a load of conditional stuff for the current compile time approach.