| PSMs in theory can be longer than 16 bits. We do not support PSMs over 16 bits, |
| since nobody else (incl bluedroid) does either |
| |
| only basic l2cap mode is supported for now |
| |
| spec does not allow "fixed channels supported" to report real supported channels (says must be zero) |
| since spec says unknown packets RXed shall be dropped, |
| we do that, and assume any open()s to the other side are a success, since at worst other side will drop the data... |
| |
| timer support prefers system with at least Linux 2.6.39 kernel for CLOCK_BOOTTIME support |
| all others will have verying degrees of issues... |
| |
| timer assumes pointers are no more than 64-bits on your system (who is using a 128-bit proc?) |
| |
| aapi* code is a big bessy since the android interface to the BT stack is a mess and not well documented. it gets cleaned up slowly |
| aapiSock* currently only supports RFC sockets. only listening is tested for now |
| |
| |
| bt 1.1 is minimum supported version |
| |
| |
| |
| TODOs: |
| |
| |
| sniff mode for sleep mode? |
| |
| |
| (when local services change - change the device class we send - it contains it) |
| (when local services change - change EIR we send - it contains it) |
| (when local name changes - change EIR we send - it contains it) |
| |
| dont bother name discovering if EIR nas name |
| |
| |
| |
| |
| |
| |
| idea: sdp socket: open, write length, write record. till closed record is visible :) |
| |
| |
| |
| |
| current gatt server api is unclear on how cross-service queued writes work. whom do you notify about execution? |
| or perhaps we need to keep track and notify all? but them how do we maintain atomicity? |
| |
| |
| |
| [!!!] |
| The Client Characteristic Configuration descriptor |
| shall be persistent across connections for bonded devices. It is unique for each client. |
| |
