[Gpe-list] Progress on gpesyncd -- but problems with libeventdb

[Graham Cobb]

I am making progress on changes to gpesyncd to make it compatible with the new 
gpe-calendar.  I have a new gpesyncd which "sort of" works for contacts and 
events (I haven't looked at todos, yet).  This uses a copy of the 
gpe-calender import/export code -- this should really move into one of the 
libraries (libgpevtype?) so it is being fixed and enhanced in one place.

Import of contacts and events mostly works -- I can now receive all my events 
and contacts on my Nokia 770 from Outlook.  The are, however, significant 
bugs (including crashes) in the export code, particularly with recurrences 
and alarms.  Once I have got gpesyncd itself working better I will start 
logging bug reports on the export code.

There are a couple of more serious obstacles in problems with the event DB 
library:

1) It is not possible to tell if an event with a particular UID exists or not.  
If event_db_find_by_uid is called with any UID an event will be returned even 
if there is no such event in the database.   This "phantom" event has no 
useful data.

This is not a major problem for gpesyncd as the user should use UIDLIST to see 
the list of valid UIDs and only ask for them.  But it can cause confusion and 
it seems like event_db_find_by_uid should really return a NULL if the UID 
does not exist (which is what the comments in the code say).

2) There is no event DB API to return all the event sources in the database.  
It is only possible to ask for event instances within a particular time 
range.  The maximum time can be unspecified but in that case any indefinitely 
repeating event causes an infinite loop trying to generate all the instances!  

Even if a time range is specified, recurring events are returned multiple 
times (corresponding to their instances).  There is no API to provide the 
event source from the instances and it is up to the caller to de-duplicate 
(which I do not do at the moment).

I realise that for calendar purposes event instances are the useful things.  
However, for synchronisation purposes, event sources are the interesting 
things -- they (along with their recurrence rules and exceptions) are the 
objects which get synchronised, not the instances themselves.

It would be useful if the event DB API provided a call to return all the event 
sources in the database, without expanding them into their instances.

Any volunteers to sort out these problems with the event DB?

Graham