[asterisk-dev] [Code Review] 2638: res_parking: Gut parker snapshots and replace them with ParkerDialString

[jrose]

> On June 26, 2013, 10:05 p.m., Matt Jordan wrote:
> > /trunk/res/parking/parking_applications.c, lines 371-398
> > <https://reviewboard.asterisk.org/r/2638/diff/1/?file=39929#file39929line371>
> >
> >     Looking at this now, this all feels rather dangerous.
> >     
> >     ast_strdup is allocating memory on the heap. We don't check that these allocations don't fail, and we don't return an error when they do. This means consumers of this functions aren't provided a success/failure, and they don't know if each of these are pointing to valid memory, pointing to garbage, or pointing to NULL.
> >     
> >     Currently, because we don't return on error, they *should* all be NULL or valid when done, but that still means that callers have to check each value themselves. This feels like a lot of off-nominal conditions that could be handled better.
> >     
> >     This should return a configuration struct containing the items from the datastore. That struct should be allocated by this function, and this function should either return NULL on error, or the configuration struct on success. Consumers of the function should free the memory allocated by this function when they are done with the configuration information.
> >     
> >

*BOOM* changed the function to return a clone of the data from the datastore. So far it seems to work properly.


- jrose


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviewboard.asterisk.org/r/2638/#review8979
-----------------------------------------------------------


On June 20, 2013, 9:44 p.m., jrose wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviewboard.asterisk.org/r/2638/
> -----------------------------------------------------------
> 
> (Updated June 20, 2013, 9:44 p.m.)
> 
> 
> Review request for Asterisk Developers, Matt Jordan and rmudgett.
> 
> 
> Bugs: ASTERISK-21877
>     https://issues.asterisk.org/jira/browse/ASTERISK-21877
> 
> 
> Repository: Asterisk
> 
> 
> Description
> -------
> 
> We decided parker snapshots were misleading and unuseful for all but a few uses of parked call messages and events. As such, they've been replaced with just having parker dial strings for handling call control on timeout.
> 
> If a channel directly parks another channel, then setting the park dial string is fairly trivial.  We just use the copy the name of the channel that parked the call and then flatten it while creating the parked user. If the call was transferred to the park application somehow, then it must be read from either the BLINDTRANSFER variable or from the newly added ATTENDEDTRANSFER variable which is basically just the BLINDTRANSFER variable for attended transfers.
> 
> This effectively wipes out one of the bug bug comments mentioned in ASTERISK-21877 regarding not having a snapshot of who parked the call on so-called 'deep parks' where a park application is deeper than the first priority of the extension a call is transferred to. The other bug bug comment remains at large, but mmichelson has a patch for bridge roles in the works which should make fixing that trivial.
> 
> 
> Diffs
> -----
> 
>   /trunk/bridges/bridge_builtin_features.c 392357 
>   /trunk/include/asterisk/parking.h 392357 
>   /trunk/main/bridging.c 392357 
>   /trunk/main/cel.c 392357 
>   /trunk/main/features.c 392357 
>   /trunk/main/parking.c 392357 
>   /trunk/res/parking/parking_applications.c 392357 
>   /trunk/res/parking/parking_bridge.c 392357 
>   /trunk/res/parking/parking_bridge_features.c 392357 
>   /trunk/res/parking/parking_controller.c 392357 
>   /trunk/res/parking/parking_manager.c 392357 
>   /trunk/res/parking/parking_ui.c 392357 
>   /trunk/res/parking/res_parking.h 392357 
> 
> Diff: https://reviewboard.asterisk.org/r/2638/diff/
> 
> 
> Testing
> -------
> 
> All kinds of different transfers (DTMF blind, DTMF attended, SIP blind, SIP attended, SIP attended with hangups before park, etc).
> 
> The only noteworthy case where the parker was anything other than who transferred the call was the following:
> 
> SIP/A calls SIP/B
> SIP/A performs attended transfer to extension f
> 
> f,1,NoOp
> f,2,Wait(10)
> f,3,Park()
> 
> SIP/A hangs up during the wait
> 
> This causes the park dial string for SIP/B's parked user to be SIP/B unfortunately, but this is more of a transfer related issue than a parking related issue.  It's also one that Richard says was also the case in previous versions, so this isn't a change in behavior.
> 
> This could be addressed by setting the ATTENDEDTRANSFER variable when the masquerade occurs to put B in the PBX where A leaves off.
> 
> 
> Thanks,
> 
> jrose
> 
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.digium.com/pipermail/asterisk-dev/attachments/20130627/d8942609/attachment-0001.htm>