GitLab

The existing tests served to check that correct metadata was
extracted. But if an extract module produced invalid SPARQL that
would not be detected, until now.

Fixes tracker#196

When called with `--file --output-format=sparql`, it now correctly
separates the FileDataObject and the InformationElement resources.

While this is an API break, the API was added recently and I suspect
it's not used yet.

The old representation was a problem because the isStoredAs attribute,
which we use to relate files with their contents, was being used to
relate the bookmark to the website.

Old output:

    INSERT DATA {
      <urn:link:https:%2F%2Fen.wikipedia.org%2Fwiki%2FTracker_(search_software)> a nie:DataObject , nfo:Website , nfo:FileDataObject ;
        nie:url "https://en.wikipedia.org/wiki/Tracker_(search_software)" .
      _:1 nie:title "Wikipedia Tracker" ;
        nie:isStoredAs <urn:link:https:%2F%2Fen.wikipedia.org%2Fwiki%2FTracker_(search_software)> .
    };

New output:

    INSERT DATA {
      <file:///home/sam/src/tracker-miners/tests/functional-tests/test-extraction-data/desktop/link-wikipedia-tracker.desktop> a nfo:FileDataObject .
      <https://en.wikipedia.org/wiki/Tracker_(search_software)> a nie:DataObject , nie:Website ;
        nie:url "https://en.wikipedia.org/wiki/Tracker_(search_software)" .
      _:1 a nfo:Bookmark ;
        nie:isStoredAs <file:///home/sam/src/tracker-miners/tests/functional-tests/test-extraction-data/desktop/link-wikipedia-tracker.desktop> ;
        nie:title "Wikipedia Tracker" ;
        nie:bookmarks <https://en.wikipedia.org/wiki/Tracker_(search_software)> .
    };

See !181
for original MR.

This occured in the TestCase module, the "No test" is a placeholder
docstring which serves no purpose. By moving init later, we can pass
the name of the real test function, and there's no output.

Extraction and other fixes

Closes #141

See merge request !273

The tracker-extract service is no longer deemed a "miner", but its
status is relevant to "tracker3 status" and "tracker3 daemon -f".
As tracker-extract is really an "implementation detail", proxy this
status via tracker-miner-fs, so these tools don't mislead if only
tracker-extract is active.

This is wrongly set on the nie:InformationElement, not on the
nfo:FileDataObject where it belongs. Causes SPARQL insertion errors for
files with chromaprint tags.

We were even testing for it wrong in the audio/flac-musicbrainz test,
update the expected output there.

Fixes: #141

This is already useful, and will be used further.

Bubble the error up from TrackerDecorator, and handle it like the other
failures in tracker-extract, so the file is reported, but doesn't sit
there waiting to be picked up for extraction again.

Now that we have "tracker3 status" and "tracker3 status $pattern", we
can patiently wait for user bug reports without melting the poles.

In some error paths, we'll want to put the SPARQL there.

We now don't have a log handler that redirects g_message()s anywhere,
so they get to stdout/journal. This spew is most times unnecessary to
look, so make it g_debug() so it requires G_MESSAGES_DEBUG=Tracker in
place.

If one tries to "tracker3 reset -s" multiple times, the second time

(tracker reset:56140): Tracker-CRITICAL **: 12:35:42.437: Location does not have a Tracker DB: Error opening directory '/home/carlos/.cache/tracker3/files/errors': No such file or directory

will happen. Try to avoid this silly situation.

If "tracker3 status" is run while tracker-extract-3 is running, it
looks enough of a miner to poke at it, but we get:

(tracker status:53148): Tracker-CRITICAL **: 12:26:14.447: No D-Bus proxy found for miner 'org.freedesktop.Tracker3.Miner.Extract'

Since the service is not listed in our special directory. Make
TrackerMinerManager skip over unlisted matches, so we don't issue
this warning.

However, might be nice to carry the status/timing information from
tracker-extract-3 to "tracker3 status". Probably carried over by
tracker-miner-fs-3.

The tracker_common_enums_header element is not used (directly) in
tracker-miners.

ci: Add coverage stage

See merge request !272

Like we have in the Tracker repo. Although the numbers are less
spectacular.

malloc_trim() is not available in musl libc

Closes #135

See merge request !271

So detect it and use it conditionally.

Closes #135

libtracker-miners-common: Remove unused file utils function

See merge request !270

build: Use specific Meson options to enable debug flags

See merge request !269

doap: Remove mailing list and point to developer-forum

See merge request !266

Based on gtk@eaef24c5

Note that TRACKER_DEBUG is disabled in most distro builds as they
pass `--buildtype=plain`. I recommend distros add `-DG_ENABLE_DEBUG`
to Tracker Miners' CFLAGS as the overhead is minimal and it helps users
to provide useful bug reports.

See also: tracker!316

ci: Correctly show SHA1 of tracker.git being built

See merge request !268

Fix stop words location when building against system-installed tracker

See merge request !267

It's /usr/share/tracker3, not /usr/share/tracker.

See https://discourse.gnome.org/t/doap-and-discourse-instead-of-a-mailing-list/4267
and GNOME/tracker!315

Fixes for Coverity warnings

See merge request !264

Check for errors, but ignore ENOENT, since it's a likely path (we delete
error report files for successful extraction without checks). Spotted by
Coverity.

CID: #361432

Spotted by Coverity

CID: #162188

As done most usually. Spotted by Coverity.

CID: #354741