[bug]: bridge_channel.c: Stream topology change amplification with multiple layers of Local channels about asterisk HOT 12 CLOSED

Have you tested with the inclusion of the patch from #403?

from asterisk.

Or if you revert 315eb551dbd18ecd424a2f32179d4c1f6f6edd26 does it resolve it?

from asterisk.

I applied the patch from #403 and it resolved my minimally reproducible example but not the actual issue that I stumbled into in the first place, so now I'll have to come up with a new, more complex example.

This happens with all channel technologies, not just PJSIP, so I think it's safe to say it is not resolved by either of the above.

I've narrowed down now I can make the issue reproduce or not reproduce based on replacing same => n,Dial(Local/s@dialme/n,,G(split)) with something suitably more complex. I'm trying to narrow that down to something minimal and will report back.

from asterisk.

Okay, I figured out how to get it to still reproduce with #403.

Just add another layer of dialing a Local channel. So add:

[dialme2]
exten => s,1,Dial(Local/${EXTEN}@dialme/n,,g)
	same => n,Hangup()

And then change the Dial to:

same => n,Dial(Local/s@dialme2/n,,G(split))

from asterisk.

Try reverting the commit @jcolp mentioned above. We've had another report of this (and can reproduce it) and narrowed it down to that commit from November 2022.

from asterisk.

@InterLinked1 We were going to cut release candidates today so if you can test the revert and it fixes your issue, we can include the revert in the RCs.

from asterisk.

@InterLinked1 We were going to cut release candidates today so if you can test the revert and it fixes your issue, we can include the revert in the RCs.

This seems to be somewhat subtle. The example I showed actually didn't reproduce with DAHDI so maybe you guys are right that it's that PJSIP thing.

However, I reverted lines 1938-1946 in res_pjsip_session.c, and the issue still happens with that modified reproducible example. Didn't bother with the test_media_add lines.

from asterisk.

The res_pjsip_session change is not the huge change, the change to channel.c is.

from asterisk.

I'm not following. Did you revert the changes in channel.c from 315eb55?

from asterisk.

The res_pjsip_session change is not the huge change, the change to channel.c is.

Ah, I'm dumb. I saw that but it looked like the same code from #403 so I assumed I already had it and #403 was just reverting that bit. Same code but different file.

Yes, this resolves the new example and the original issue - thanks for the catch!

from asterisk.

If "core & res_pjsip: Improve topology change handling." is reverted by #532.
Do we still need #403?

from asterisk.

It's not strictly required, but does no harm in still being present.

from asterisk.