[asterisk-dev] wrong usage of AST_FLAG_END_DTMF_ONLY in ast_senddigit()?

[Michael Neuhauser]

Hello!

I think that there is something wrong how ast_senddigit() uses
AST_FLAG_END_DTMF_ONLY (head of 1.4, r63286): ast_senddigit_begin() is
only called if AST_FLAG_END_DTMF_ONLY is not set.

This flag is only set in ast_channel_bridge():
	if (!c0->tech->send_digit_begin)
		ast_set_flag(c1, AST_FLAG_END_DTMF_ONLY);
	if (!c1->tech->send_digit_begin)
		ast_set_flag(c0, AST_FLAG_END_DTMF_ONLY);
>From this code and how AST_FLAG_END_DTMF_ONLY is used elsewhere, I
conclude that it is set iff the *OTHER* channel's tech has no
send_digit_begin function (not the channel that holds the flag).

So why is a property of another channel (the bridged one) checked when
the digit is send to the "own" channel? Shouldn't ast_senddigit() look
something like this

        int ast_senddigit(struct ast_channel *chan, char digit)
        {
        	/* original: if (!ast_test_flag(chan, AST_FLAG_END_DTMF_ONLY)) { */
        	if (chan->tech->send_digit_begin) {
        		ast_senddigit_begin(chan, digit);
        		ast_safe_sleep(chan, 100); /* XXX 100ms ... probably should be configurable */
        	}
        	
        	return ast_senddigit_end(chan, digit, 100);
        }

i.e., send a begin only if the channel's tech wants it?

Another thing is, that ast_senddigit_begin() will use playtones to fake
the digit when called for a channel that has a tech with
send_digit_begin == NULL (e.g., SIP with INFO). Is this a desired
behaviour? I can't think of any reason why it should behave that way.
Maybe it makes sense if the tech has both send_digit_begin and
send_digit_end set to NULL.

Regards,
	Mike
-- 
Dr. Michael Neuhauser                              mailto:mike at firmix.at
Firmix Software GmbH                                  sip:mike at firmix.at
Vienna/Austria/Europe                               tel:+43-1-7890849-30
Linux Development and Services                     http://www.firmix.at/
AST_FLAG_END_DTMF_ONLY