[asterisk-dev] [Code Review] Calendaring API for Asterisk

Thu Dec 11 22:16:42 CST 2008

-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
http://reviewboard.digium.com/r/58/#review202
-----------------------------------------------------------


/trunk/main/calendar.c
<http://reviewboard.digium.com/r/58/#comment395>

    Here is one thing to consider in your reload handling ...
    
    The way some other modules handle reload is that instead of deleting all objects and reloading all of them, they do this:
    
    1) Mark all objects with a pending deletion flag.
    
    2) Process configuration.  If the object already exists, remove the deletion flag, and update its contents.  If it did not yet exist, create it.
    
    3) Delete all objects that are still marked with the deletion flag.
    
    That way, you don't have period of time where everything disappears, and you end up doing less memory freeing/allocating. 


/trunk/main/calendar.c
<http://reviewboard.digium.com/r/58/#comment396>

    You have a race condition here.
    
    It is possible to end up sleeping longer that you really want to.  Here's how it could go down:
    
    1) You determine how long you should sleep.
    
    2) Something else gets scheduled that is sooner than what you got from #1
    
    3) You lock and then go to sleep.
    
    What you need to do is hold the lock before you determine how long you should sleep.  You will also need to hold the same lock when you add scheduler entries, and also signal the condition when you add entries.
    
    That will avoid this issue.


/trunk/res/res_caldav.c
<http://reviewboard.digium.com/r/58/#comment397>

    Update copyright for just 2008


/trunk/res/res_caldav.c
<http://reviewboard.digium.com/r/58/#comment398>

    You missed unlocking the calendar here and in a few other error handling blocks


/trunk/res/res_caldav.c
<http://reviewboard.digium.com/r/58/#comment399>

    space after for


/trunk/res/res_exchangecal.c
<http://reviewboard.digium.com/r/58/#comment400>

    Use ast_strdup() and check for failure


/trunk/res/res_exchangecal.c
<http://reviewboard.digium.com/r/58/#comment401>

    space after for


/trunk/res/res_exchangecal.c
<http://reviewboard.digium.com/r/58/#comment402>

    space after if


/trunk/res/res_exchangecal.c
<http://reviewboard.digium.com/r/58/#comment403>

    It looks like this cast isn't needed.  Just declare tmp as a const char *


/trunk/res/res_exchangecal.c
<http://reviewboard.digium.com/r/58/#comment404>

    You have a minor bug here.  With strncpy(), you need to subtract 1 from the size of the buffer.  However, just use ast_copy_string() instead.


/trunk/res/res_exchangecal.c
<http://reviewboard.digium.com/r/58/#comment405>

    coding guidelines :)


/trunk/res/res_exchangecal.c
<http://reviewboard.digium.com/r/58/#comment406>

    Missed unlock of the calendar here and in some other error handling blocks


- Russell


On 2008-12-11 10:40:03, Terry Wilson wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> http://reviewboard.digium.com/r/58/
> -----------------------------------------------------------
> 
> (Updated 2008-12-11 10:40:03)
> 
> 
> Review request for Asterisk Developers.
> 
> 
> Summary
> -------
> 
> Calendaring integration for Asterisk--currently supporting iCalendar files (hosted over HTTP/HTTPS), CalDAV, and Microsoft Exchange.  Features available: dialplan functions for querying and writing (if the calendar technology supports it like CalDAV and Exchange do) calendar event information, a device state provider for phone BLF support, and calendar event notification through executing dialplan apps or context/exten.
> 
> The overall structure is that the res_* calendar modules register as a calendar tech to main/calendar.c, and after all modules are loaded, main/calendar.c parses calendar.conf and calls the calendar_load callback for each register calendar tech based on the "type" field for that calendar in calendars.conf.  Currently each calendar runs in its own thread downloading and refreshing the remote calendar data as necessary as to avoid serializing downloads.  
> 
> Eventually it would probably be a good idea for me to implement a thread pool and honor a maximum number of downloads per host:port as well as a global maximum number of simultaneous downloads--perhaps with a min/max refresh value so I could randomize the time for refreshing a bit so as not to be refreshing all of the calendars at once.  Also, there was a request to add support for the dialplan functions to query calendars that aren't in calendar.conf i.e. CALENDAR_QUERY(ical/http://www.google.com/calendar/ical/nkt5atdq7cdbes3ehdfpendpnc%40group.calendar.google.com/public/basic.ics,${EPOCH},$[${EPOCH} + 3600]), etc.
> 
> Currently I don't have any support for querying/setting attendees, mostly because it is a list of results whereas all of the others are individual fields.  Writing gets especially ugly for them because of the current format of CALENDAR_WRITE(calendar,${HASHKEYS(calendar)})=${HASH(calendar)}.  I suppose I could add a ...,${HAHSKEYS(attendees)})=...,${HASH(attendees)} to the end...it is just getting a little ugly.
> 
> 
> Diffs
> -----
> 
>   /trunk/build_tools/menuselect-deps.in 163077 
>   /trunk/configs/calendar.conf.sample PRE-CREATION 
>   /trunk/configure.ac 163077 
>   /trunk/include/asterisk/_private.h 163077 
>   /trunk/include/asterisk/autoconfig.h.in 163077 
>   /trunk/include/asterisk/calendar.h PRE-CREATION 
>   /trunk/main/Makefile 163077 
>   /trunk/main/asterisk.c 163077 
>   /trunk/main/calendar.c PRE-CREATION 
>   /trunk/main/loader.c 163077 
>   /trunk/makeopts.in 163077 
>   /trunk/res/res_caldav.c PRE-CREATION 
>   /trunk/res/res_exchangecal.c PRE-CREATION 
>   /trunk/res/res_icalendar.c PRE-CREATION 
> 
> Diff: http://reviewboard.digium.com/r/58/diff
> 
> 
> Testing
> -------
> 
> I have tested all three calendar modules by adding, deleting, and moving events and verifying that notifications occur.  I have tested writing to Exchange, and Zimbra and Google Calendar through CalDAV.  I have run with MALOC_DEBUG looking for leaks and dropped references to astobj2 objects. 
> 
> 
> Thanks,
> 
> Terry
> 
>