GitLab

Previous behaviour:
1. Radio #1: Pipe 0 opened for writing (Address1), pipe 1 opened for
reading (Address2)
2. The radio would end up listening on the following pipes/addresses,
because pipe0 was not assigned a separate reading address.
Pipe0: Address1
Pipe1: Address2
3. This is generally not a problem with 2 radios, but when multicasting
with three. If two of the radios transmit to the same address, then
start listening, all three radios will be listening to the same address
on pipe0, unless pipe0 has been assigned a separate reading address

New behaviour:
1. Radio #1: Pipe 0 opened for writing (Address1), pipe 1 opened for
reading (Address2)
2. When calling radio.startListening() pipe0 is closed, because it is
not assigned a reading address
3. Pipe0 is re-opened for writing only, unless a reading address is
assigned to pipe0

This wouldn't really affect things while using auto-ack, because two
radios should not be writing to the same pipe/address.

- Add closeReadingPipe for RPi
- Also adjusted timing for startListening();
- Fix failure detect variable on RPi

Found issue with dynamic payloads on RPi - writeType was not specified
properly for dynamic ACKs with a NOACK payload set - Identified thanks
to mannkind   - See
https://github.com/TMRh20/RF24Network/commit/6e25453ffabf583c0304b701a3cfcf8f98c9fc94

Added delay results in better response times. The RPi is just too quick
here.

Removed the wrong rx_flush() on RPi code after testing - caused strange
behaviour in some cases
Corrected failureDetect variable

Introduced a bug by changing payload_size value to a static value of 32
and forgot to revert it. The commit was:

https://github.com/TMRh20/RF24/commit/aac1650646b2c440c8822deb2f99514f4e2e4ddd#diff-6dbdcee25d60de0476cb76000ce10fd3

Discovered due to #12 - Issue submitted by cbolanos79

Seems to be the same as issue 9 affecting Due and RPi. Enabled the 140us
delay on startListening for all boards.

Per issue #5 by sven337
- Add optional failure handling/timeout
- Un-comment #define FAILURE_HANDLING in RF24_config.h to enable
- Add radio.errorDetected variable - indicates if an error/timeout was
detected
example:
if(radio.failureDetected){
delay(1000);
radio.begin();
radio.failureDetected = 0;
radio.openWritingPipe(addresses[1]);
radio.openReadingPipe(1,addresses[0]);
report_failure(); //blink leds, send a message, etc.
}

Additional:
- removed unused wide_band boolean
- remove rx buffer flushes for RPi also

Modified write() to use startFastWrite, then toggle CE after completion
to remove need for a delay on some boards.
Added delay to startWrite for non-Arduino boards.
Removed receive buffer flushes. This prevents an issue in the
RF24Network library, where payloads would be received, but flushed when
a transmission takes place prior to reading.
Should address issues #7 and #11

- Discovered a bug affecting Arduino Due using extended SPI methods, it
seems an extra delay is required before switching to TX mode
- Found the same bug affected the RPi, and removed all delays from
bcm2835 driver, since this bug caused the strange behaviour on RPi

- Modified spi transfernb hopefully for the last time, re #9
- Updated RPi examples to minimize CPU usage
- Updated transfer example for Arduino and RPi to exit TX mode after 4ms
per manufacturer. Improves multicast streaming.
- Bit of a cleanup on RPi examples

- Adjusted spi writes (delay vs no delay)
- Re-enabled buffer flushes on start/stop listening
- Adjusted delay for transfernb()
- Testing indicates that with gettingstarted_call_response example, ack
responses may not be written soon enough if CPU load is high on RPi. eg:
testing indicates this is a limitation of the hardware, not the driver.

- Moved printf.h to root library folder
- Modified printf.h for better compatibility with Arduino Due and Teensy
- Fixed some potential bugs preventing Teensy from working
- Improved Due support and performance
- may help correct issue #7

- moved radio.flush_tx(); to public
- adjusted call_response example to use tx_flush();
- adjusted write_payload, read_payload

- added delays to stop/start listening on RPi
- removed buffer flushes on start/stop listening for RPi
- added modified transfernbd function to BCM library - one has no
delays, also removed one delay period from both by waiting for RX data
- modified transfernb function to stop sending if RX fifo is full
- added delay to stoplistening for arduino etc. (actually seems to
improve performance)
- tests on-par with Arduino performance

- Added _BV define for Due
- adjusted defines

- Adjusted define in RF24_Config.h for teensy
- Teensy doesn't seem to support extended SPI methods, may not perform
as well as on Due
- Currently compiles, but is untested with teensy

- Changed from using reinterpret cast to memcpy for storing pipe0
address

Updated examples to use proper include path

https://github.com/TMRh20/RF24/issues/6

- Added setAddressWidth()  - allows address widths of 3 to 5 bytes
(24,32 or 40 bits)
- Addresses can now be specified via a byte array
- Thanks to Zephyrr for suggestion

Since RPi behaves more like a PC than Arduino, users should be able to
execute various programs without powering down the radio module, where
an Arduino needs to be re-programmed between code changes, so a full
power-down is not such a dramatic step.

Thanks to sej7278 (GitHub) see:
https://github.com/TMRh20/RF24Network/issues/2

Makefile fixes

Restore closeReadingPipe merge for mysensors

Restore closeReadingPipe to be compatible with mysensors

- Enabled use of hardware CS1 pin
- Removed custom_cs code for software driven CSN
- Modified BCM driver to allow use of hardware SPI pins not in use while
SPI is active.
- Updated constructor in examples

Added support for hardware SPI on RPI per jscrane fork:
- Fixed a few bugs with single SPI transfers
- Aligned with optimized RF24 library fork
- Updated bmc2835 files per
http://www.airspayce.com/mikem/bcm2835/index.html
- Merged bcm files with changes from
https://github.com/farconada/RF24Network
- Added example files
- Added setup info to readme

- fixed mistake with changes to setPALevel prior to pushing

- Using status register is faster to read TX FIFO status - Zephyrr (GitHub)
- Fixed typo on FLUSH_RX cmd
- Added support for pwr levels on SI24R1 chips with smaller code

- Updated some example files and tests
- Small changes to minimize code size

For nRF24L01+ to go from power down mode to TX or RX mode it must first
pass through stand-by mode.
There must be a delay of Tpd2stby (see Table 16.) after the nRF24L01+
leaves power down mode before
the CEis set high.

5ms is the maximum defined in the v1.0 datasheet

- Added new function: enableDynamicAck()  to enable single-write NOACK
payloads (merge closer with gcopeland fork)
- Added single NOACK writes for write(), writeFast(), startWrite() and
startFastWrite()
- powerUp() now checks to see if radio is powered on. No delay if
powered up.
- radio must still be powered up manually if powered down manually
- use setRetries function in begin()
- set CE low on powerDown
- updated readme