On Mi, 2011-09-14 at 17:44 +0200, Patrick Ohly wrote:
I'm currently integrating some more fixes into the 1.2 branch
not in 18.104.22.168 (add<->add conflict (BMC #22783), command line and
source passwords (BMC #21311, #22937), fixing some EDS API issues with
regards to multiple detached recurrences (BMC #22940), don't check
"backend" unless it is needed, backend API changes, updated testing).
More details once I have written an updated NEWS entry tomorrow.
I can either publish a 22.214.171.124 release candidate this Friday before
going on a 10 day vacation or call it 1.2 right away.
Are there opinions either way? Anyone willing to give a release
candidate some testing, in particular with an eye towards the more
recent changes? Anyone urgently waiting for the final 1.2?
Okay, call me a chicken, but I am going down the 126.96.36.199 pre-release
path instead of releasing 1.2 right away.
The list of changes is pretty long:
binaries: now compatible with Debian Testing/libnotify.so.4 (BMC
libnotify is not linked directly into syncevo-dbus-server in the
binaries. Instead libnotify.so.1 till .so.4
(current Debian Testing) are opened opened dynamically and the
necessary functions are looked up via dlsym(). Not finding the
libraries or the functions silently disables this notification
* calendar sync: better handling for add<->add conflicts (partly fixes BMC #22783)
When both sides of a sync have added the same event, the sync must
determine which one is more recent instead of blindly overwriting
always the same side. Such conflicts are typically rare except for
enterprise scenarios where meeting invitiations are processed
automatically by a groupware (Exchange, Google Calendar/Mail, ...)
and then the attendee status is updated on one side.
SyncEvolution now does the necessary age comparison and preserves the more
recent data for most properties. In some properties the data from both
sides is preserved by concatenating the text (description, location, ...).
It remains to be seen whether that is really desirable. Also, sync statistics
are slightly off: the incoming item is counted as "added" even though it
gets turned into an update.
* item operations: authentication problem for WebDAV when using keyring (BMC #21311)
The password still wasn't looked up in the keyring when using
* "databasePassword" source property: lookup failure in keyring (BMC #22937)
The databasePassword also wasn't looked up at all when doing item operations
via the command line.
When configuring sources for an HTTP server, the config name typically
is just the context (@foo). When using the config in the HTTP server,
the config name is the peer inside that context (client@foo). Because
the GNOME keyring lookup keys for the "databasePassword" (more
specifically, the object name) contained the full config name which
was different in both cases, looking up the saved password failed.
The solution is to normalize the config name (to accomodate for
different ways of spelling it) and use only the context, with @ as
before. This will break existing setups where the object name in the
keyring (incorrectly) includes the full config name. In that case just
configure the source again to set the password anew.
* Evolution Calendar: fixed detached recurrence support (BMC #22940)
When manipulating a meeting series with more than one detached
recurrence certain sequences of operations could incorrectly fail
with "UID already exists".
* iCalendar 2.0: must set VALUE in EXDATE (part of BMC #22940)
EXDATE has a VALUE parameter, which wasn't defined in the XML
profile. Didn't seem to matter at all in practice, but wasn't
* GTK sync-ui: wrap sync service descriptions (BMC #7199)
Descriptions of different sync services are not fully visible unless
word-wrapping gets enabled.
* source configs: don't check "backend" unless it is needed
When using a config which has sources with a backend type set which is
not currently available, an error was thrown even if those sources
weren't even part of the current operation (for example, syncing
another source which is currently supported).
* config migration: avoid name conflicts and auto syncing of old configs (BMC #22691)
When (auto-)migrating a config, it was possible that a name for the
peer, say foo.old, was chosen for the renamed config although there
was already such a config, for example foo.old in ~/.sync4j. Besides
being confusing for users, this also led to a bug in the code where it
copied from the older config with the foo.old name.
The main problem fixed is the disabling of auto syncing
in the old config. Otherwise it was still used by syncevo-dbus-server
for syncing, which triggered another auto-migration, ad infinitum...
* auto syncing: must check whether enabled when looking at unknown URLs (part of BMC
"syncURL = insert your URL here" with "autoSync = 0" did lead to
sync attempts although it wasn't enabled. A check for "auto syncing
enabled" was missing for the "unknown transport" case.
* CalDAV/CardDAV + local storage: avoid empty properties
The main motivation for this change is that a recent Apple Calendar
server rejects vCards with empty BDAY property. Another reason is that
keeping the data as small as possible is desirable by itself.
Sending an empty property serves as a hint for the peer that the
property is supported. This is not necessary when storing an item in a
backend. Therefore this commit disables empty properties for all
backends which do not themselves set the m_backendRule Synthesis info
* Apple CardDAV: apply PHOTO import/export scripts by default
A recent Apple Calendar server (correctly) rejects the invalid
PHOTO;TYPE=unknown: property in a vCard. This internal representation
must be cleared before serializing the field list.
* for developers: modified backend API
- ClientTestConfig modernized
- InsertItemResult::m_merged turned from boolean to enum
* testing and compilation changes; for example, the minimum version of
libsynthesis is now checked at configure time instead of failing at
runtime due to missing features in the Synthesis engine
Best Regards, Patrick Ohly
The content of this message is my personal opinion only and although
I am an employee of Intel, the statements I make here in no way
represent Intel's position on the issue, nor am I authorized to speak
on behalf of Intel on this matter.