Among the TODOs:
* Refactor obtaining LPC serial number and chip ID into separate API and header/source files. Remove from main().
* Create a usb_set_serial_number_descriptor() or similar function to be called before USB stack is started.
* Ensure USB serial number descriptor is valid even if code forgets to initialize the serial number before the USB stack is started. May be as simple as providing default initializer for usb_descriptor_string_serial_number[].
* Create a #define/constant for the usb_descriptor_string_serial_number length.
* Identify what's causing intermittent crashes in hackrf_transfer when no serial number is specified. I'm probably misusing getopt.
* Permit serial number without leading zeros so you don't have to type as much.
* Add support for serial number argument in other hackrf_* tools.
* Provide libhackrf support for enumerating multiple HackRFs, so that hackrf_info can list all devices. May require an additional libhackrf function, outside of hackrf_open().
...and anything else that makes this less of a hack.
Previously CPLD programming involved a large 64kbyte buffer into which
the the entire bitstream would be completely downloaded, and at which
point the programming process would commence. This is needlessly
wasteful of memory. Moreover, it could lead to USB timeouts as the
firmware needs to wait the entire duration of the programming process
before returning an ACK to the host after the final SETUP data phase
packet.
Instead, we now receive 512 byte chunks of the bitstream and stream them
to the CPLD one at a time. We wait for each packet to be streamed out to
the CPLD before ACKing the packet to prevent active data being
overwritten.