Check that the buffers are our fake data before counting them (so that refused stun packages dont trigger it)

Porting libnice to the new stun usage API. I know it's a big commit, but it can't be divided into smaller ones, it's all one big chunk

A much better way of generating the username and getting the password. This allows for msn compat username generation too.

Check if the alternate_server is given in the bind_process, so it can now be set to NULL if you don't want it.

Added a stun_usage_ice_conncheck_process in order to correctly process a connectivity check (same as bind process but adds role conflict checking)

validating the error response will give us a success since now the agent checks if it's an unauthorized error, it won't check the integrity

Do not force the password on error responses, let the agent take care of that, so 'unauthorized' responses won't get a message-integrity

Avoid segfaulting if no message integrity attribute is found. Also make sure that if it happens, it's only accepted on error responses with codes bad-request and unauthorized.

correctly handle the IGNORE_CREDENTIALS usage flag. And add the optionial fingerprint whenever we're in 3489bis mode. The USE_FINGERPRINT usage flag is only used to discard received messages without that flag

Move bind and stun-ice back into a usages/ directory and update the includes and makefiles everywhere

Revert "add the ice usage and the corresponding test"

This reverts commit b6153207.

Add a test for the agent's unmatched response error, and fix test-parse after the change to the agent that automatically uses the right key for the message integrity

Make a difference between an unauthorized validation because username/password do not exist and with the unauthorized when the key doesn't match. Also commented the validation return values