If the string being processed is detected as UTF16, the "utf8"
variable would already contain a malloc()'ed string, so the
GString would be leaked altogether.

Then, ensure the string is freed if already converted from UTF16,
and spare the utf8 validation too, since we already know the string
is valid utf8.

the "keyw" variable ends up pointing to chunks within the string
that must be freed, so keep a pointer to the original string
so it can be freed afterwards.

Those loops bail out because the variable word hasn't been found to
contain non-blank characters, but it still may point out to strings
like "", so ensure that memory is freed.

gst_discoverer_discover_uri() may return a non-NULL error and
a GstDiscovererInfo pointer, so ensure it is freed in case of
error. The info was also being leaked on successful extraction,
that is fixed too.

In case of error, the filename was being leaked.

If a tasks was iterated over multiple extraction modules, the task
ended up inserted on priv->running_tasks more than once, but only
removed once when the task finished, leaving dangling pointers in
priv->running_tasks.

If no extraction module was found for a file, the corresponding
TrackerExtractTask was being leaked.

Set the only reference to the decorator info as the GTask return
value, so it is definitely destroyed when the GTask itself is
unref()'ed.

The values returned by g_variant_get() must be freed.

>0 processed items are needed in order to estimate time, otherwise division by
0 happens.

This can be fixed up to commit 2c1745ec

Given we're reusing these struct sizes many times, it makes sense to use
GSlice.

It is now unused, and its role replaced by TrackerExtractDecorator

This is to cover all basic classes that tracker-extract needs
to do piggyback indexing.

TrackerExtractDecorator is put in use, replacing TrackerController as
the master of TrackerExtract. .service and .desktop files are added as
appropriate so tracker-extract is recognized as a miner, handled by
tracker-control, etc...

This is a tracker-extract implementation of TrackerDecorator, that
takes care of fetching files to be inspected, and forward these
to the passed TrackerExtract object. Upon completion, the information
is returned back to TrackerDecorator control so it is eventually
committed.

This prevents a crash when reporting the statistics later, as there is a
NULL key that is expected to be a GModule.

This is a file-specific implementation of TrackerDecorator, that besides
checks on initialization, it checks for previously known items not yet
extracted on newly added volumes.

This object is a TrackerMiner abstract implementation that instead of
actively looking for information, it passively receives notifications
through the tracker-store GraphUpdated DBus signal, and offers the
possibility to fill in further information of newly added resources
or older resources missing inspection by this miner.

By default, this object receives information from all resources with
tracker:notify true, the class-names property can tweak this so it
only acts on events from a certain rdf:type.

In order to know which elements are missing inspection, the miner
relies on a nie:DataSource specific to each TrackerDecorator
implementation. On startup, TrackerDecorator will check for the
eligible elements missing that datasource to ensure those are checked.
The data source can be set through the data-source property.

This nie:DataSource will be added to resource by tracker-extract,
in order to mark files already inspected, and recognize the files
that are still unprocessed if resuming is needed.

This way, files will be recognized as promptly as possible as being
of the correct type, even though tracker-extract hasn't gone through
it yet. This information could also be used in tracker-extract itself
in order to figure out the best extractor.

Tracker-extract is going to be meant to listen for GraphUpdated
signals, so no direct communication between tracker-miner-fs and
tracker-extract would be necessary.

FallbackRdfTypes (with an extra 's') is now a string list, so all rdf:types
that apply to a mimetype can be specified.

Glib >= 2.30 already hard depend on libffi.

To debug a functional test failure you can now replace the 'rm' at the
bottom of 'test-runner.sh' with 'echo', and then call:

    tracker-sandbox.py -i /tmp/tracker-test-XXXX/tracker-tests --shell

It's not a good idea to set XDG_CONFIG_HOME, XDG_DATA_HOME, etc. to
the same directory, because some software might store a file
with the same name in two of those directories. DConf does it.

Andrew den Exter authored Nov 22, 2013 and Philip Van Hoof committed Jan 09, 2014

Use the much more comprehensive avformat_stream_info to probe
data from packetized data rather than directly trying to partially
decode frames.