[asterisk-dev] [Code Review] 3759: chan_sip: upgrade registry and mwi object to ao2

[Corey Farrell]

> On July 21, 2014, 1:14 p.m., Mark Michelson wrote:
> > /trunk/channels/chan_sip.c, lines 3522-3534
> > <https://reviewboard.asterisk.org/r/3759/diff/3/?file=64859#file64859line3522>
> >
> >     Heh, it's a bit late for this now, but it would have probably saved you a bunch of searching and replacing to just change the definition of registry_unref() and registry_addref() to use ao2 in their implementations.

I knew this, but decided to put the work in to use the best ao2 function for each use.  I do not like the use of intermediary functions or macro's that simply link to ao2 operations.


> On July 21, 2014, 1:14 p.m., Mark Michelson wrote:
> > /trunk/channels/chan_sip.c, lines 6543-6547
> > <https://reviewboard.asterisk.org/r/3759/diff/3/?file=64859#file64859line6543>
> >
> >     I agree with this comment. Since mwi->call was created via sip_alloc(), destruction of it should be accomplished with dialog_unref()

I've removed this comment and will post a separate review to fix this in 1.8+.

OTOH I think this code is actually unreachable, we have a circular link - mwi->call holds a reference to mwi->call->mwi.  I ran testsuite - tests/channels/SIP/subscribe before and after this patch, and mwi leaks.


> On July 21, 2014, 1:14 p.m., Mark Michelson wrote:
> > /trunk/channels/chan_sip.c, lines 33324-33336
> > <https://reviewboard.asterisk.org/r/3759/diff/3/?file=64859#file64859line33324>
> >
> >     You sometimes search for sip_registrys using OBJ_KEY, but your hash and cmp callbacks do not account for a search key being passed to the callback instead of a sip_registry.
> >     
> >     Use the templates on this wiki page https://wiki.asterisk.org/wiki/pages/viewpage.action?pageId=25919686 to define your hash and comparison callbacks and you should be good to go.

I actually only ever used OBJ_KEY.  I've fixed the hash/cmp callbacks to follow the template, supporting OBJ_SEARCH_OBJECT in case it's used in the future.  I've also replaced the use deprecated OBJ_KEY with OBJ_SEARCH_KEY.

When I originally ran the testsuite this was testing for reg1->name, which was a char[80] defined as the first field.  Since &reg1->name == reg1, it didn't cause a problem.  Now that reg1->configvalue is a stringfield, I think this would have caused a segfault, though I didn't get one with the testsuite.


- Corey


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviewboard.asterisk.org/r/3759/#review12784
-----------------------------------------------------------


On July 21, 2014, 11:52 p.m., Corey Farrell wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviewboard.asterisk.org/r/3759/
> -----------------------------------------------------------
> 
> (Updated July 21, 2014, 11:52 p.m.)
> 
> 
> Review request for Asterisk Developers.
> 
> 
> Bugs: ASTERISK-24067
>     https://issues.asterisk.org/jira/browse/ASTERISK-24067
> 
> 
> Repository: Asterisk
> 
> 
> Description
> -------
> 
> Upgrade all ASTOBJ objects in chan_sip to ao2.
> 
> 
> Diffs
> -----
> 
>   /trunk/channels/sip/include/sip.h 419127 
>   /trunk/channels/chan_sip.c 419127 
> 
> Diff: https://reviewboard.asterisk.org/r/3759/diff/
> 
> 
> Testing
> -------
> 
> Full testsuite run.
> 
> 
> Thanks,
> 
> Corey Farrell
> 
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.digium.com/pipermail/asterisk-dev/attachments/20140722/abcb66a6/attachment-0001.html>