This removes the need for doing an extra pointer dereference on every
access to the buffers, shrinking the code by around 100 bytes.
The members for these buffers must be public for now, since the
interrupt handlers also need to access them. These can later be made
private again.
Furthermore, the struct ring_buffer was removed. This allows the all
head and tail pointers to be put into the HardwareSerial struct before
the actual buffers, so the pointers all end up in the first 32 bytes of
the struct that can be accessed using a single instruction (ldd).
References: #947
Since the buffers aren't bigger than 64 bytes, these values can be
smaller. This saves a few bytes of ram, but also saves around 50 bytes
of program space, since the values can now be loaded using a single
instruction.
To prevent problems when people manually increase the buffer size, a
compile-time check is added.
Closes: #1078
Otherwise, you get an error when compiling for processors with no serial port because the header file is always compiled.
See, for an example of the problem: https://github.com/damellis/attiny/issues/8
Before, the UMSELn1 bit was being to set to 1, putting the UART into a reserved mode. Now, we only set the high (0x80) bit to 1 for the ATmega8, which is needed to access UCSRnC (whose i/o address is shared with UBRRH).
Also, no longer bitwise-or the new config with the existing register value, because we're actually configuring all the settings in the register. (We're not using UCPOL, but it's supposed to be 0 in asynchronous mode.)
Mostly useful for Leonardo - simple way to test whether the port is actually opened by an application and ready to receive data. For Serial objects attached to real UARTs always returns true.
write(), print(), println() now return size_t (and don't use negative values to signal errors).
Print adds writeError() for checking for write errors, clearWriteError() to reset the flag to false, and a protected setWriteError() for signalling errors.
http://code.google.com/p/arduino/issues/detail?id=598
Because UBBR is only 12 bits, we were overflowing it at 300 baud because of the use of the U2X bit. Now we turn off U2X if it would yield a UBBR value that would overflow.
Note that this breaks 300 baud communication with the computer on the Uno and Mega 2560 because the 8U2 USB-serial firmware has this same bug (and previously they cancelled each other out). Since, however, it seems more likely that people will need to use 300 baud to communicate with other (legacy) hardware than with the computer, I'm making this change. Issue for 8U2 firmware bug: http://code.google.com/p/arduino/issues/detail?id=542http://code.google.com/p/arduino/issues/detail?id=522
Now Serial.write() places characters in the transmit buffer, and the data register empty interrupt reads and transmit them. Based loosely on the implementation here: ftp://wookey.org.uk/arduino.
http://code.google.com/p/arduino/issues/detail?id=262
- moving back to multple cores per platform
- using target instead of platform
- moving per-board and per-programmer preferences out of Preferences.java and into a new Target class
- adding a new "target" preference
- support for platform:value values in board preferences for bootloader path and core
- XXX: need to support platform:value syntax for board upload.using preferences.
Cristian Maglie
Matthijs Kooijman
David A. Mellis
Alarus
Zach Eveland