[asterisk-dev] [Code Review] 4108: Weak Proxy Objects

[Corey Farrell]

> On April 10, 2015, 7:05 p.m., rmudgett wrote:
> > /trunk/main/astobj2.c, lines 492-494
> > <https://reviewboard.asterisk.org/r/4108/diff/5-6/?file=71780#file71780line492>
> >
> >     The comment doesn't make sense.  How is destructor_fn supposed to access values in the weak proxy?  The real object is no longer linked with the weakproxy.

The following is allowed and a reasonable thing to do:
weakproxy->name = ast_strdup(name);
realobj->name = weakproxy->name;

The idea is that weakproxy has to own storage of any shared fields since realobj could be destroyed at any time.  So if the destructor uses realobj->name, it is accessing memory that is owned by the weakproxy.


- Corey


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviewboard.asterisk.org/r/4108/#review15185
-----------------------------------------------------------


On April 3, 2015, 12:58 p.m., Corey Farrell wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviewboard.asterisk.org/r/4108/
> -----------------------------------------------------------
> 
> (Updated April 3, 2015, 12:58 p.m.)
> 
> 
> Review request for Asterisk Developers, George Joseph and rmudgett.
> 
> 
> Bugs: ASTERISK-24936
>     https://issues.asterisk.org/jira/browse/ASTERISK-24936
> 
> 
> Repository: Asterisk
> 
> 
> Description
> -------
> 
> This implements "weak" references.  The weakproxy object is a real ao2 with normal reference counting of its own.  When a weakproxy is pointed to a normal object they hold references to each other.  The normal object is automatically freed when a single reference remains (the weakproxy).  The weakproxy also supports subscriptions that will notify callbacks when it does not point to any real object.
> 
> 
> Diffs
> -----
> 
>   /trunk/tests/test_astobj2_weaken.c PRE-CREATION 
>   /trunk/main/astobj2.c 433964 
>   /trunk/include/asterisk/astobj2.h 433964 
> 
> Diff: https://reviewboard.asterisk.org/r/4108/diff/
> 
> 
> Testing
> -------
> 
> Ran the included test with REF_DEBUG enabled under valgrind.  No reference leaks or improper memory access.  Though this does not test for races, I don't know of an automated way to do that.
> 
> 
> Thanks,
> 
> Corey Farrell
> 
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.digium.com/pipermail/asterisk-dev/attachments/20150410/5b503a0e/attachment.html>