[pulseaudio-discuss] [PATCH] echo-cancel: Avoid segfaults due to invalid master sink or source

[Georg Chini]

On 07.05.2017 11:16, Tanu Kaskinen wrote:
> On Sat, 2017-05-06 at 22:15 +0200, Georg Chini wrote:
>> On 06.05.2017 22:06, Tanu Kaskinen wrote:
>>> I made a strange interpretation, because what you really meant seemed
>>> even more crazy. I forgot that module-switch-on-connect only moves
>>> streams if they are routed to the default sink, and because of that
>>> missing piece your explanation made no sense to me at all. I understand
>>> the logic now.
>>>
>>> Implementing your proposal doesn't seem that simple: what if the user
>>> has set a filter sink as the default sink? Would you change the default
>>>    sink just for the duration of the filter sink move?
>>>
>> I already implemented it and it is quite simple (using part of your
>> patch).
>>
>> The default sink will be set to the new sink by switch-on-connect
>> anyway (even the configured default sink), so the previous user
>> choice is irrelevant. I think however, that the simple logic of
>> switch-on-connect should at least be changed so that in the case
>> that the previous default sink was a filter then the streams should
>> not be moved away from it (unless the new sink also is a filter).
>> But this would be a matter of a different patch.
>>
>> BTW did you see my other remark about switch-on-connect? It is
>> somewhat hidden in another topic.
> I forgot to reply to the last part... No, I don't think I saw that
> other remark. What was that remark?
>
I just copy&paste what I wrote in my previous mail:

Another similar point, while reading the code of 
module-switch-on-connect, i
found a possible issue:
When a new sink turns up, pa_core_update_default_sink() is called from
sink_put(). After that, PA_CORE_HOOK_SINK_PUT is fired. Within the hook
in module switch-on-connect, it is checked, if the new sink is the default
sink. If yes, the hook does nothing. Doesn't that mean that if the new
sink has highest priority (and no user default is set), the streams will
not be moved away from the old default sink, because the old default
is lost when pa_core_update_default_sink() is called before the hook?

The same has already been the case before the recent changes, so again
I'm not sure if I am missing something ...