GitLab

Reading converts from date-only to local time to UTC. Writing must do
the same conversion. Conversion from UTC to local time was missing,
therefore the date may have gotten shifted by a day (depending on the
local time zone).

This commit intentionally does not cache the local time zone (like
reading does), because this is a bit ugly and the assumption is that
writing occurs a lot less often than reading.

Closes: https://bugzilla.gnome.org/show_bug.cgi?id=686693

core: Remove trailing whitespace in folks/backend.vala

This commit makes folks build without zeitgeist again.

Always set CHECK_VERBOSE when running tests under gdb so that output (from
gdb) isn’t redirected to a log file. Apparently, headless gdb is hard to
use.

Closes: https://bugzilla.gnome.org/show_bug.cgi?id=685250

eds: Add create-remove-stores unit test.

telepathy: Add empty set_persona_stores to telepathy backend.

keyfile: Add set_persona_stores to keyfile backend.

eds: Implement set_persona_stores in eds backend.

eds: Implement new API in eds backend

We should be comparing a set of strings, not a set of AbstractFieldDetails.
This fixes a crash in the link-personas-via-local-ids Tracker test.

The E.Sources for testing were being created by calling `cp` in the wrapper
script running around the entire test suite. However, each test case was
tearing down the E.Sources by deleting them. This meant that no E.Source
existed for anything except the first test case in a given test suite.

The fix is to move the file copying inside the test suite’s setup method.

fixes 855803a6Signed-off-by: Marc-Antoine Perennou <Marc-Antoine@Perennou.com>

Fix the debug message in the Individual constructor.

Closes: https://bugzilla.gnome.org/show_bug.cgi?id=684954

Counter works for last 30 days instead of all time since it can
mess up calculations easily.
Reduce query size to Zeitgeist since events are filtered in the
Persona class.

Closes: https://bugzilla.gnome.org/show_bug.cgi?id=672709 (again)

This reverts commit af500a4b and applies
a different fix. The problem was that persona changes could get away without
being notified when disconnecting due to an account being disabled. This
is because, after storing the cache, an additional check is performed for
whether the account was enabled. If it is, the cache is loaded (and persona
changes notified). If not, the cache is not loaded, *and persona changes
were not notified*.

This situation could occur when a Telepathy account is disabled:
 1. _notify_connection_cb() is called, but the TpAccount::enabled property
    is still true.
 2. While in store_cache(), the TpAccount::enabled property goes false.
 3. The check before load_cache() would fail, so load_cache() would never
    be called.

Similarly, it could occur when a Telepathy account is removed, with
TpAccountManager:account-removed being emitted while in store_cache().

Closes: https://bugzilla.gnome.org/show_bug.cgi?id=683390