[asterisk-dev] [Code Review] Make reloading cdr_pgsql.so not screw up catastrophically.

[Tilghman Lesher]

-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviewboard.asterisk.org/r/1711/#review5388
-----------------------------------------------------------



/branches/1.8/cdr/cdr_pgsql.c
<https://reviewboard.asterisk.org/r/1711/#comment9944>

    A better approach would be simply to hold the same lock which is held while posting CDRs.  Unregistering means that we could lose some CDRs while the reload is occurring.


- Tilghman


On Feb. 2, 2012, 3:34 p.m., jrose wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviewboard.asterisk.org/r/1711/
> -----------------------------------------------------------
> 
> (Updated Feb. 2, 2012, 3:34 p.m.)
> 
> 
> Review request for Asterisk Developers.
> 
> 
> Summary
> -------
> 
> Without this patch, attempting cli command:
> module reload cdr_pgsql.so
> 
> will cause the columns list to get doubled up (because they aren't cleared) which will make insertions fail since the insert statement will have all of its elements repeated.
> 
> This patch does two things.  Well, three.  First it unregisters cdr_pgsql from cdr.c so that cdr logs won't try to use this cdr storage method while the reload process is happening (re-registration occurs at the end of the reload).  Second, it clears the columns list during reload so that it doesn't end up doubling up the entires.  As an irrelevant third, it updates the copyright status in the documentation to 2012 from 2006 or so.
> 
> 
> This addresses bug ASTERISK-19216.
>     https://issues.asterisk.org/jira/browse/ASTERISK-19216
> 
> 
> Diffs
> -----
> 
>   /branches/1.8/cdr/cdr_pgsql.c 353844 
> 
> Diff: https://reviewboard.asterisk.org/r/1711/diff
> 
> 
> Testing
> -------
> 
> Tested with and without patch. With patch, the CDRs get logged correctly after reloading without the errors mentioned in the bug report and also the reload doesn't spit out a "this cdr backend has already been registered in cdr.c" (paraphrasing) type error.
> 
> 
> Thanks,
> 
> jrose
> 
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.digium.com/pipermail/asterisk-dev/attachments/20120202/62d106b6/attachment-0001.htm>