Added board hardware driver support for the Olimex AVR-USB-162 development board (thanks to Steve Fawcett).

Changed the signature of the CALLBACK_USB_GetDescriptor() callback function so that the descriptor pointer is const, to remove the need for extra casting inside the callback (thanks to Jonathan Kollasch).

Optimise Endpoint_ConfigureEndpoint_Prv() function to produce smaller code when CONTROL_ONLY_DEVICE compile time token is defined.

Split out endpoint and pipe stream functions into new EndpointStream.c/.h and PipeStream.c/.h files.

Added compile time error to the AVRISP-MKII project when built for the U4 chips, as the default VTARGET detection ADC channel does not exist on these chips (thanks to Marco).

Fixed critical pipe/endpoint memory allocation issue where the bank memory address space could be silently overlapped in the USB controller if the endpoints or pipes were allocated in anything other than ascending order (thanks to Martin Degelsegger).

Removed complicated logic for the Endpoint_ConfigureEndpoint() function to use inlined or function called versions depending of if the given bank size is a compile time constant, as the compiler does a better job of optimizing with basic code.

Changed over all device demos to use a clearer algorithm for the configuring of the application's endpoints.

More spell checking of all source files -- correct missed errors, switch to EN-GB spelling dictionary.

Use static inline intermediatary functions in Endpoint.h to handle the masking for the private endpoint configuration routine, instead of performing the masking in the Endpoint_ConfigureEndpoint() macro.

Move over message ID specific header data to unions inside the standard TMC header structure, so that all header data is read in one chunk in the incomplete TMC demo.

Convert over internal pseudo-function macros to true inline functions for added type-safety and compile-checking.

Indent core library function parameters so that there is only one parameter per line, to increase readability.

Move and rename DevChaper9.c/.h and HostChapter9.c/.h to the HighLevel directory, new named DeviceStandardReq.c/.h and HostStandardReq.c/.h.

Rename LowLevel.c/.h to USBController.c/.h.

Remove dependancies on the complete <avr/io.h> header in the HighLevel USB driver directory, to ensure less work is required in later (possible) ports.

Removed the automated checking of event names in the demo, project and bootloader makefiles due to inconsistancies between the behaviour of the command line tools used to perform the check on each platform.

Removed the internal endpoint and pipe memory allocation macros, as this is already performed directly in the code. Simplify the endpoint and pipe reset procedure.

Added new Drivers/USB/LowLevel/Device.c file to house Device mode specific functions that are more complicated than simple macros. Moved USB_Device_SendRemoteWakeup() to the new Device.c source file and corrected it to unfreeze and restart the USB controller clock before issuing a Remote Wakeup request.

Removed the USB_Device_IsRemoteWakeupSent() and USB_Device_IsUSBSuspended() macros, as they are now obsolete.

Minor tweaks to the library documentation.

Spell check code and manual pages. Remove custom Doxygen CSS stylesheet, as the new 1.7 Doxygen's default stylesheet is much better.

Speed up calls to Pipe_IsPipeBound() by immediately skipping unconfigured pipes, rather than performing token check first.

The EEPROM stream read/write functions now use eeprom_update_byte() instead of eeprom_write_byte(), so that only changed bytes are written to EEPROM to preserve its lifespan.

Changed over the AVRISP-MKII and TemperatureDataLogger projects to use eeprom_update_byte() when writing non-volatile parameters to EEPROM to preserve its lifespan.

Remove incorrect check for the current device state in the Set Configuration request handler of DevChapter9, which broke Set Configuration requests to the device under most circumstances.

Move out the SPI prescaler list to a PROGMEM module-level variable in ISPTarget.c.

Fixed possible device lockup when INTERRUPT_CONTROL_ENDPOINT is enabled and the control endpoint is not properly selected when the ISR completes.

Fix nasty bug in USBInterrupt.c which would cause a lockup if the control endpoint was not selected at the point the USB endpoint interrupt fires when INTERRUPT_CONTROL_ENDPOINT is defined.

Add svn:eol-style property to source files, so that the line endings are correctly converted to the target system's native end of line style.

Fixed INTERRUPT_CONTROL_ENDPOINT compile time option preventing other interrupts from occuring while the control endpoint request is being processed, causing possible lockups if a USB interrupt occurs during a transfer.

Fixed Set Configuration requests not being stalled until the host has set the device's address.

Fixed possibility of internal signature retrieval being corrupted if an interrupt occurs during a signature byte read (thanks to Andrei Krainev).

Added new ENDPOINT_*_BusSuspended error code to the Endpoint function, so that the stream functions early-abort if the bus is suspended before or during a transfer.

Oops - relocated USBInterrupt.h header file requires adjusted include paths for the other USB driver header files.

Relocate USBInterrupt.c/.h from the HighLevel USB driver directory to the LowLevel directory as it is hardware-dependent.

USB_Init() no longer calls sei() to enable global interrupts - this must now be done in the user application once all init code has run.

Fixed Set/Clear Feature requests directed to a non-configured endpoint not returning a stall to the host.

Make TeensyHID bootloader use an 8 byte control endpoint to match the official Teensy, just in case Paul decides to authenticate on that attribute in the future.

Removed two-step endpoint/pipe bank clear and switch sequence for smaller, faster endpoint/pipe code.

Added spacing between multiple paragraphs in Doxygen \note sections.

Removed call to the clock prescaler reset function in the Teensy bootloader to save space - the user application is accessed via a watchdog reset anyway, so the prescale reset would only affect the speed of the bootloader itself.