Tristan Van Berkom [Sat, 8 Jun 2013 14:12:14 +0000 (10:12 -0400)]
Bug 701838 - Work around a semantic API break in GObject
Recent versions of GObject fire critical messages if properties are
installed outside of class initializers (causing all of EDS tests to
fail). This patch adds some class data and initializes the class in
the dynamic subclass initializer.
Note that the class data which should normally be freed in the class
finalizer is not freed, since it is a static type the finalizer will
never run (and GLib also takes the liberty of firing critical messages
for specifying a class finalizer).
Marián Čavojský [Wed, 5 Jun 2013 18:40:16 +0000 (20:40 +0200)]
Updated slovak translation
Milan Crha [Wed, 5 Jun 2013 13:57:31 +0000 (15:57 +0200)]
Bug #699448 - Crash under e_book_backend_ldap_stop_view()
Matej Urbančič [Mon, 3 Jun 2013 19:40:16 +0000 (21:40 +0200)]
Updated Slovenian translation
Marek Černocký [Mon, 3 Jun 2013 12:37:35 +0000 (14:37 +0200)]
Updated Czech translation
Milan Crha [Mon, 3 Jun 2013 11:22:25 +0000 (13:22 +0200)]
Be able to debug SMTP provider separately
One can use CAMEL_DEBUG=smtp to debug only SMTP camel provider.
Fran Diéguez [Sat, 1 Jun 2013 21:35:47 +0000 (23:35 +0200)]
Updated Galician translations
Milan Crha [Fri, 31 May 2013 16:47:06 +0000 (18:47 +0200)]
Bug #672837 - Crash in camel_pop3_engine_command_free
Matthew Barnes [Fri, 31 May 2013 11:17:07 +0000 (07:17 -0400)]
Add camel_imapx_conn_manager_ref_store().
Replaces camel_imapx_conn_manager_get_store(), which is not thread-safe.
Also, track the CamelStore internally with a GWeakRef instead of a weak
pointer, which is also not thread-safe.
Matthew Barnes [Thu, 30 May 2013 19:00:58 +0000 (15:00 -0400)]
Add camel_imapx_command_add_qresync_parameter().
This splits a big wad of extra logic in imapx_select() out into a nice
and tidy utility function.
One important change, however, is the message sequence match data stops
at the lesser message count reported by the IMAP server vs. our summary.
This is to work around a frequent Zimbra IMAP failure where listing a
message sequence number in a QRESYNC parameter which exceeds the true
message count results in a so-called "parse error" from Zimbra:
A00010 SELECT INBOX (QRESYNC (... (1,5,10 ...)))
* 9 EXISTS
...
A00010 BAD parse error: invalid message sequence number: 1,5,10
This scenario can happen if messages have been expunged from the
folder, but our summary has not yet been updated to reflect that.
Not sure if this is a Zimbra bug. I certainly would not call it a
"parse error", but RFC 5162 is not clear on this detail.
Matthew Barnes [Thu, 30 May 2013 17:20:02 +0000 (13:20 -0400)]
Move imapx_get_uid_from_index() to camel-imapx-utils.
Rename it camel_imapx_dup_uid_from_summary_index() and take a
CamelFolder instead of a CamelFolderSummary.
Daniel Mustieles [Thu, 30 May 2013 15:43:27 +0000 (17:43 +0200)]
Updated Spanish translation
Tristan Van Berkom [Fri, 10 May 2013 08:23:16 +0000 (17:23 +0900)]
test-client-custom-summary.c: Added tests for regex queries
Tristan Van Berkom [Tue, 23 Apr 2013 08:36:47 +0000 (17:36 +0900)]
EBookBackendSqliteDB: Support regex queries
E_BOOK_QUERY_REGEX_NORMAL is supported in summary queries if the
query field is summarized, E_BOOK_QUERY_REGEX_RAW and normalized
but non-summarized queries are delegated to EBookBackendSexp for
matching.
Tristan Van Berkom [Tue, 23 Apr 2013 08:35:54 +0000 (17:35 +0900)]
EBookBackendSexp: Support regex query types.
Tristan Van Berkom [Tue, 23 Apr 2013 08:34:54 +0000 (17:34 +0900)]
EBookQuery: Added regex query API (bug 701258)
Added E_BOOK_QUERY_REGEX_NORMAL and E_BOOK_QUERY_REGEX_RAW
ManojKumar Giri [Wed, 29 May 2013 16:06:55 +0000 (21:36 +0530)]
Updated Odia Translation.
ManojKumar Giri [Wed, 29 May 2013 15:58:37 +0000 (21:28 +0530)]
Updated Odia Translation.
Matthew Barnes [Sun, 26 May 2013 15:03:47 +0000 (11:03 -0400)]
Post-release version bump.
Matthew Barnes [Sun, 26 May 2013 14:51:05 +0000 (10:51 -0400)]
NEWS update for 3.9.2 release.
Matthew Barnes [Fri, 24 May 2013 17:19:47 +0000 (13:19 -0400)]
libedataserver: Fix some documentation typos.
Matthew Barnes [Tue, 21 May 2013 18:54:33 +0000 (14:54 -0400)]
Improve error handling in imapx_command_fetch_message_done().
Spotted a few execution paths where we claim success but have no
stream object for the fetched message.
Matthew Barnes [Tue, 21 May 2013 18:32:44 +0000 (14:32 -0400)]
camel_imapx_command_set_error_if_failed() minor cleanup.
Matthew Barnes [Tue, 21 May 2013 02:48:40 +0000 (22:48 -0400)]
Make CamelNNTPStore a bit more thread-safe.
* Make CamelNNTPStore's public members private.
* Guard the private CamelNNTPStore members with a mutex.
* Add thread-safe accessors:
camel_nntp_store_ref_cache()
camel_nntp_store_ref_stream()
camel_nntp_store_ref_summary()
camel_nntp_store_get_current_group()
camel_nntp_store_dup_current_group()
camel_nntp_store_set_current_group()
camel_nntp_store_add_capabilities()
camel_nntp_store_has_capabilities()
camel_nntp_store_remove_capabilities()
Matthew Barnes [Tue, 21 May 2013 17:26:21 +0000 (13:26 -0400)]
CamelNNTPStore: Remove unused "password_reprompt" flag.
Matthew Barnes [Tue, 21 May 2013 17:25:48 +0000 (13:25 -0400)]
CamelNNTPStore: Remove unused "posting_allowed" flag.
Matthew Barnes [Tue, 21 May 2013 17:25:07 +0000 (13:25 -0400)]
CamelNNTPStore: Remove unused "extensions" flags.
Matthew Barnes [Tue, 21 May 2013 00:57:01 +0000 (20:57 -0400)]
CamelNNTPStore: Remove CamelMemStream member.
It's just used for assembling command strings. Use a GString instead,
and for heaven sake don't make it a public member! That's not thread-
safe at all.
Matthew Barnes [Mon, 20 May 2013 21:43:46 +0000 (17:43 -0400)]
Make CamelPOP3Store a bit more thread-safe.
* Make CamelPOP3Store's cache and engine members private.
* Guard the private cache and engine members with a mutex.
* Add thread-safe accessors which return a new reference:
camel_pop3_store_ref_cache()
camel_pop3_store_ref_engine()
TODO: CamelPOP3Engine itself should also be made thread-safe, but that
will be considerably more work. This commit merely prevents the
CamelPOP3Engine from being finalized while still in use.
David Woodhouse [Mon, 20 May 2013 13:14:24 +0000 (14:14 +0100)]
Bug 699597 - Autocomplete using GAL is slow
With large addressbooks, using an inner join which produces hundreds of
thousands of duplicate results, and then filtering them back down to what
we really wanted with 'DISTINCT' is really inefficient. I can't even tell
you *how* inefficient, because it didn't complete a single query in the
hours that I left it running. Switch it to a more sensible outer join and
it completes in a second or so on my 216000-entry address book.
Kjartan Maraas [Mon, 20 May 2013 12:42:29 +0000 (14:42 +0200)]
Updated Norwegian bokmål translation
Milan Crha [Thu, 16 May 2013 14:15:05 +0000 (16:15 +0200)]
CamelGpgContext: Report decrypt error only if no data was decrypted
When there is a message which is both signed and encrypted in once,
the gpg checks also signature of the sender, and if the sender's key
is not available between user's keys, or it's not trusted, or
the signature is otherwise incorrect, then the gpg returns failure,
which caused no data returned, though the decrypt itself worked
flawlessly.
Milan Crha [Thu, 16 May 2013 09:27:35 +0000 (11:27 +0200)]
test-intervaltree.c: Include toplevel libical header file
Shankar Prasad [Tue, 14 May 2013 09:43:46 +0000 (15:13 +0530)]
updated kn translations
Shankar Prasad [Tue, 14 May 2013 09:16:10 +0000 (14:46 +0530)]
updated kn translations
Milan Crha [Mon, 13 May 2013 17:08:03 +0000 (19:08 +0200)]
Camel MH provider can deadlock on folder creation
This was reported downstream as [1]. The MH provider can deadlock
during folder creation, because it tries to get a new CamelFolder
instance for a folder it is currently creating, which makes
the CamelObjectBag (of CamelFolder-s) wait for itself.
[1] https://bugzilla.redhat.com/show_bug.cgi?id=854017
Milan Crha [Mon, 13 May 2013 14:35:46 +0000 (16:35 +0200)]
[IMAPX] Set real Trash/Junk folder type as expected
Thus it's shown with a special icon in a folder tree.
Milan Crha [Mon, 13 May 2013 14:28:44 +0000 (16:28 +0200)]
Bug #699811 - Moved IMAPx messages reappear
This affected users whom had configured a real Trash folder.
David Woodhouse [Fri, 10 May 2013 20:13:35 +0000 (21:13 +0100)]
Bug 699909 — Do not use toggle_refs in CamelObjectBag
They are broken and not thread-safe. Use GWeakRef instead to fix the
original problem that they were introduced to fix (bug 638810).
Daniel Mustieles [Thu, 9 May 2013 15:30:54 +0000 (17:30 +0200)]
Updated Spanish translation
Matthew Barnes [Wed, 8 May 2013 17:37:54 +0000 (13:37 -0400)]
IMAPX: Fix error message when moving with UID MOVE.
Separate commit since this introduces a new translatable string.
Matthew Barnes [Wed, 8 May 2013 17:30:18 +0000 (13:30 -0400)]
IMAPX: Support IMAP MOVE extension.
http://tools.ietf.org/rfc/rfc6851
Matthew Barnes [Wed, 8 May 2013 16:20:11 +0000 (12:20 -0400)]
IMAPX: Add macros for checking server capabilities.
These just read better and are more compact.
CAMEL_IMAPX_HAVE_CAPABILITY(cinfo, capa)
CAMEL_IMAPX_LACK_CAPABILITY(cinfo, capa)
Both macros evaluate to FALSE if 'cinfo' is NULL.
Matthew Barnes [Wed, 8 May 2013 14:48:03 +0000 (10:48 -0400)]
Bug 697859 - Google: "Other Fax" is saved as "Business Fax"
The gd:phoneNumber @rel value "http://schemas.google.com/g/2005#fax"
has no clear mapping to an EContact field.
Both it and "http://schemas.google.com/g/2005#other_fax" end up in a
vCard as TEL;TYPE=fax. But because of the conversion table ordering,
TEL;TYPE=fax was being saved as "#fax" instead of "#other_fax".
Then when reading the contact back from Google, I guess some fallback
logic somewhere was converting "#fax" to "TEL;TYPE="work,fax". Never
found exactly where that was happening. Maybe in the Contact Editor?
This commit reorders the conversion table to prefer "#other_fax" over
"#fax". There's still a conflict where if a Google contact specifies
both "#fax" and "#other_fax" one of the fax numbers may be lost. But
honestly, who has that many fax machines nowadays?
Matthew Barnes [Wed, 8 May 2013 14:38:18 +0000 (10:38 -0400)]
google: Remove environment variable check for debugging.
GLib suppresses debug-level log messages by default now, so just defer
to the G_MESSAGES_DEBUG mechanism to activate debug messages, like so:
$ export G_MESSAGES_DEBUG=e-book-backend-google
$ evolution-addressbook-factory
Matthew Barnes [Wed, 8 May 2013 13:14:26 +0000 (09:14 -0400)]
Coding style and whitespace cleanup.
Matthew Barnes [Tue, 7 May 2013 17:51:37 +0000 (13:51 -0400)]
ECalBackendWeather: e_weather_source_new() is failable.
Backend logic was failing to take this into account and crashing if the
weather location code in the ESource was invalid.
I'm not sure if EWeatherSource is even worth keeping around, but I'll
come back to that later. Not doing an all-out rewrite at the moment.
Matthew Barnes [Tue, 7 May 2013 16:18:47 +0000 (12:18 -0400)]
EWeatherSource cleanups.
Matthew Barnes [Mon, 6 May 2013 14:02:41 +0000 (10:02 -0400)]
Bug 699024 - [google/yahoo] Don't override username
When mail child sources are added to a google or yahoo collection, avoid
overriding an existing user name setting. The IMAP or (especially) SMTP
configuration may have been modified to use a non-Google/Yahoo! server.
Matthew Barnes [Sat, 4 May 2013 13:44:00 +0000 (09:44 -0400)]
ECalClientView: Make D-Bus signal handlers thread-safe.
Seeing crashes now where handler functions for D-Bus signals like
"objects-added" are receiving finalized ECalClientView instances.
Possibly a side-effect of recent changes to GMainContext usage.
Use the trick of passing an allocated GWeakRef along with a suitable
GClosureNotify to g_signal_connect_data(). The handler function will
just return quietly if it can't obtain a strong client view reference.
Matthew Barnes [Sat, 4 May 2013 14:32:36 +0000 (10:32 -0400)]
EBookClientView: Make D-Bus signal handlers thread-safe.
Seeing crashes now where handler functions for D-Bus signals like
"objects-added" are receiving finalized EBookClientView instances.
Possibly a side-effect of recent changes to GMainContext usage.
Use the trick of passing an allocated GWeakRef along with a suitable
GClosureNotify to g_signal_connect_data(). The handler function will
just return quietly if it can't obtain a strong client view reference.
Matthew Barnes [Fri, 3 May 2013 17:14:53 +0000 (13:14 -0400)]
Bug 699223 - Emit view signals from context where view was started
This applies to both EBookClientView and ECalClientView.
Incoming D-Bus signals were being propagated to the EClient's main loop
context, but there may not be a running main loop on that thread, so the
main loop sources were never being dispatched and the view signals never
emitted.
Instead use the main loop context which was the thread-default at the
time the view was started, which is often the so-called "main" thread
where a main loop is more likely to be running.
Matthew Barnes [Thu, 2 May 2013 23:34:37 +0000 (19:34 -0400)]
IMAPX: Recognize RFC 5530 response codes.
Matthew Barnes [Thu, 2 May 2013 21:39:42 +0000 (17:39 -0400)]
Bug 699160 - IMAPX: Always flush debugging messages
Had a debugging scenario where key information was not written to a log
file until another IMAP command was issued. Caused a bit of confusion.
This commit ensures that won't happen again.
Matthew Barnes [Thu, 2 May 2013 21:38:25 +0000 (17:38 -0400)]
camel_imapx_debug() cleanups.
Dan Williams [Thu, 2 May 2013 20:36:22 +0000 (15:36 -0500)]
Make camel_mime_message_dump() dump message contents
Previously unimplemented. Dumps in hex and ASCII:
class: CamelMimeMessage
mime-type: text/plain; charset="UTF-8"
content class: CamelDataWrapper
content mime-type: text/plain; charset="UTF-8"
content len 967
4f 6e 20 54 68 75 2c 20 32 30 31 33 2d 30 35 2d On Thu, 2013-05-
30 32 20 61 74 20 31 34 3a 31 37 20 2d 30 35 30 02 at 14:17 -050
30 2c 20 44 61 6e 20 57 69 6c 6c 69 61 6d 73 20 0, Dan Williams
77 72 6f 74 65 3a 0a 3e 20 4f 6e 20 54 68 75 2c wrote:.> On Thu,
20 32 30 31 33 2d 30 35 2d 30 32 20 61 74 20 31 2013-05-02 at 1
31 3a 35 33 20 2d 30 37 30 30 2c 20 4c 69 6e 75 1:53 -0700, Linu
73 20 54 6f 72 76 61 6c 64 73 20 77 72 6f 74 65 s Torvalds wrote
3a 0a 3e 20 3e 20 4f 6e 20 54 68 75 2c 20 4d 61 :.> > On Thu, Ma
79 20 32 2c 20 32 30 31 33 20 61 74 20 31 31 3a y 2, 2013 at 11:
(02:09:38 PM) mbarnes: dcbw: nice, I forgot about camel_mime_message_dump() -- go ahead and commit that
Matthew Barnes [Thu, 2 May 2013 18:48:28 +0000 (14:48 -0400)]
Bug 699461 - Accept any 2xx HTTP code as successful DELETE
Matthew Barnes [Thu, 2 May 2013 12:55:40 +0000 (08:55 -0400)]
ESource: Fix a compiler warning.
Matthew Barnes [Sun, 28 Apr 2013 01:36:46 +0000 (21:36 -0400)]
Bug 698725 - ESourceRegistry: Handle service restarts gracefully
Check GDBusObjectManagerClient's "name-owner" property when handling
"object-added" and "object-removed" signals. If the property is NULL,
implement a strategy that swaps out the ESource's old proxy object for
an equivalent replacement such that applications should remain unaware
of the service restart.
This is important to handle gracefully not only for crash recovery, but
also when the registry service receives a Reload() method invocation or
a SIGHUP, both of which cause it to temporarily release its well-known
bus name and reload key files.
Matthew Barnes [Sat, 27 Apr 2013 15:18:22 +0000 (11:18 -0400)]
e_source_ref_dbus_object: Use the property lock.
ESourceRegistry is going to swap out the GDBusObject as part of the
recovery strategy for evolution-source-registry restarts, so we need
to make sure e_source_ref_dbus_object() is properly thread-safe.
Matthew Barnes [Tue, 30 Apr 2013 17:52:19 +0000 (13:52 -0400)]
ESourceRegistry: object_path_table cleanups.
Matthew Barnes [Tue, 30 Apr 2013 17:45:44 +0000 (13:45 -0400)]
Add source_registry_dbus_object_dup_uid().
Handy internal utility for ESourceRegistry.
This also happens to fix an EDBusSource reference leak in
source_registry_create_sources_object_added_cb().
Matej Urbančič [Wed, 1 May 2013 18:32:29 +0000 (20:32 +0200)]
Updated Slovenian translation
Matthew Barnes [Sun, 28 Apr 2013 16:19:25 +0000 (12:19 -0400)]
Post-release version bump.
Matthew Barnes [Sun, 28 Apr 2013 15:35:23 +0000 (11:35 -0400)]
NEWS update for 3.9.1 release.
Matthew Barnes [Sun, 28 Apr 2013 14:57:27 +0000 (10:57 -0400)]
Fix reference leak in book_client_view_emit_objects_modified().
This was causing test-client-photo-is-uri to fail.
Matthew Barnes [Sat, 27 Apr 2013 13:28:13 +0000 (09:28 -0400)]
EGoaClient nitpicks.
Matthew Barnes [Sat, 27 Apr 2013 12:47:18 +0000 (08:47 -0400)]
Bug 698726 - Handle goa-daemon crashes/restarts gracefully
EGoaClient is an improved GoaClient that handles daemon crashes and
restarts gracefully so E-D-S accounts are not destroyed unnecessarily.
Marek Černocký [Fri, 26 Apr 2013 21:53:24 +0000 (23:53 +0200)]
Updated Czech translation
Shantha kumar [Thu, 25 Apr 2013 12:47:08 +0000 (18:17 +0530)]
Tamil Translation Updated
Dimitris Spingos [Thu, 25 Apr 2013 06:24:30 +0000 (09:24 +0300)]
Updated Greek translation
Matthew Barnes [Wed, 24 Apr 2013 14:09:09 +0000 (10:09 -0400)]
Improve error handling in imapx_server_get_message().
If a GET_MESSAGE job for the same UID is already in progress, wait for
the job to finish, but don't assume it was successful. Try retrieving
the message from cache, but if that fails then we have to re-fetch the
message from the IMAP server.
Tristan Van Berkom [Wed, 24 Apr 2013 08:49:23 +0000 (17:49 +0900)]
ECalClient test: test-client-get-revision not broken anymore
Not sure how this is passing again, but it is.
Milan Crha [Wed, 24 Apr 2013 06:53:41 +0000 (08:53 +0200)]
Do not consider PGP and S/MIME subparts as attachments
Tristan Van Berkom [Tue, 23 Apr 2013 10:30:51 +0000 (19:30 +0900)]
test-client-custom-summary.c: Optimize test.
Reuse the same two addressbooks for every test, so that only the
first 2 tests populate the addressbooks with contacts and the remaining
tests only create a client and execute a query.
Also add all test contacts with a single batch command and dont fetch
the test contacts, also removed the #if 0'd case which is the only
case which required the local contact copies.
Tristan Van Berkom [Fri, 19 Apr 2013 10:48:08 +0000 (19:48 +0900)]
test-server-utils: Allow reuse of existing addressbook.
If committing the new scratch source ends with an error that
the addressbook/calendar already exists, then just use the existing
source instead of bailing out.
This is needed for tests which reuse the addressbook from a previous test.
Conflicts:
tests/test-server-utils/e-test-server-utils.c
Daniel Mustieles [Mon, 22 Apr 2013 15:48:15 +0000 (17:48 +0200)]
Updated Spanish translation
Carles Ferrando [Fri, 19 Apr 2013 22:29:32 +0000 (00:29 +0200)]
[l10n] Updated Catalan (Valencian) translation
Gil Forcada [Fri, 19 Apr 2013 22:28:29 +0000 (00:28 +0200)]
[l10n] Update Catalan translation
Matthew Barnes [Thu, 18 Apr 2013 14:45:34 +0000 (10:45 -0400)]
ECalClientView: Remove unnecessary fetching of ECalClient.
Left over from the D-Bus error stripping work.
Matthew Barnes [Thu, 18 Apr 2013 14:44:20 +0000 (10:44 -0400)]
EBookClientView: Remove unnecessary fetching of EBookClient.
Left over from the D-Bus error stripping work.
Milan Crha [Thu, 18 Apr 2013 12:14:48 +0000 (14:14 +0200)]
Bug #698261 - Skip invalid flags in imapx_parse_flags()
Previous patch had incorrectly implemented skip of the broken flag,
which fixes this change.
Milan Crha [Thu, 18 Apr 2013 10:27:16 +0000 (12:27 +0200)]
Bug #698261 - Skip invalid flags in imapx_parse_flags()
Milan Crha [Thu, 18 Apr 2013 06:35:18 +0000 (08:35 +0200)]
CamelStoreSettings::filter-inbox: Change default to TRUE
Matthew Barnes [Wed, 17 Apr 2013 14:37:13 +0000 (10:37 -0400)]
LC_ADDRESS is a GNU extension. Treat it as such.
Unconditional use of the GNU extension LC_ADDRESS causes compiler
errors on non-GNU systems.
In test cases, use LC_MESSAGES if LC_ADDRESS is unavailable.
Also fail gracefully with a log message if setlocale() fails.
Matthew Barnes [Wed, 17 Apr 2013 10:48:06 +0000 (06:48 -0400)]
ESource: Move "remember-password" to Authentication extension.
It's a more suitable place since the option is only relevant to
EAuthenticationSession.
Milan Crha [Wed, 17 Apr 2013 09:24:59 +0000 (11:24 +0200)]
Bug #691939 - Password prompt has "save password" always checked
Matthew Barnes [Tue, 16 Apr 2013 17:31:02 +0000 (13:31 -0400)]
Remove backward-compatibility cruft for goa-1.0 < 3.8.
In particular, GOA's Google provider uses OAuth 2.0 now, so we can
drop the OAuth 1.0a support in EGDataGoaAuthorizer and with it the
last lingering goa-1.0 dependency outside of the GOA module.
Matthew Barnes [Tue, 16 Apr 2013 15:42:38 +0000 (11:42 -0400)]
Bump minimum goa-1.0 requirement to 3.8.
Still an optional dependency. This is to purge backward-compat cruft.
Milan Crha [Tue, 16 Apr 2013 14:51:50 +0000 (16:51 +0200)]
Switch NULL to empty string in e_util_utf8_{data_}make_valid()
It's used on various places which expect a valid string, but which
doesn't check for NULL values of the string, which can cause a crash,
like xtian found with e_data_cal_respond_get_object(), which can have
passed in a NULL object. This change is also to avoid NULL-checking
on too many places.
Tristan Van Berkom [Tue, 16 Apr 2013 08:18:53 +0000 (17:18 +0900)]
EBookBackendSqliteDB: Avoid summary introspection on initial creation.
The summary introspection does not introspect whether regular prefix
indexes are created, which is acceptable since it does not effect queries.
However since we introspect before creating the indexes, we lose the
configured index information which is needed to initially create those
indexes.
This patch simply checks if the table exists when initially creating
the summary and avoids the introspection if the table is being created
for the first time.
Matthew Barnes [Mon, 15 Apr 2013 22:54:41 +0000 (18:54 -0400)]
Automatically strip D-Bus info from error messages.
I mistakenly assumed code generated by the gdbus-codegen tool would
automatically strip off the D-Bus error name from error messages, but
that apparently is not possible according to my discussion with David
Zeuthen in [1], since it would break g_dbus_error_is_remote_error().
Distinguishing between local and remote errors strikes me as a rarely
needed corner case, and is certainly not worth the extra step imposed
on applications when dealing with something as fundamental as GError.
This commit prevents the D-Bus error name from leaking through in any
Evolution-Data-Server client-facing library functions.
Also rename any stack-allocated GError pointers to 'local_error' to
help distinguish them from GError "out" parameters which are always
named 'error'. This is just to improve variable name consistency.
[1] https://bugzilla.gnome.org/697819
Matthew Barnes [Mon, 15 Apr 2013 11:42:26 +0000 (07:42 -0400)]
Bug 695578 - Trailing space in translatable string
Matthew Barnes [Sun, 14 Apr 2013 14:31:20 +0000 (10:31 -0400)]
Bug 697705 - e_cal_client_remove_objects_sync() crashes on empty UID
Matthew Barnes [Sun, 14 Apr 2013 13:24:04 +0000 (09:24 -0400)]
Test suite cleanups.
Colin Walters [Fri, 12 Apr 2013 13:23:53 +0000 (09:23 -0400)]
build: Fix srcdir != builddir from git
We can't do (cd $(srcdir); ...) and inside reference $(top_srcdir)
because that variable uses a *relative* path. Thus we copy the
approach from gnome-shell of explicitly using addprefix to append the
source directory.
Milan Crha [Fri, 12 Apr 2013 08:38:55 +0000 (10:38 +0200)]
Bug #696757 - Do not use dynamic keys in g_object_set_data()
Milan Crha [Fri, 12 Apr 2013 07:44:42 +0000 (09:44 +0200)]
Bug #681837 - Make enum type registration thread safe
Matthew Barnes [Thu, 11 Apr 2013 15:17:58 +0000 (11:17 -0400)]
Bug 697734 - Memory leak in local calendar backend
We were leaking an icalproperty while seeding the "revision" D-Bus
property during instance initialization, because the backend does
not yet have an icalcomponent to hold the icalproperty.
This commit avoids the leak during instance initialization, but also
makes sure the "revision" property gets updated once the backend does
have an icalcomponent.
Matthew Barnes [Thu, 11 Apr 2013 13:45:24 +0000 (09:45 -0400)]
ECalBackendFile: Indentation cleanups.
Tabs, not spaces.