Skip to content

tracker-extract: Delete files with unknown graphs

From the time they are processed to the time their content is extracted by tracker-extract, there is a window of opportunity that the file might change substantially while its nfo:FileDataObject is already present in a content graph.

So this file would exist and be on a content graph, thus be a candidate for tracker-extract-3. Extraction would be attempted and failed, and deletion of the extractor hash would be attempted. The only thing that might not quite work is that the file mimetype no longer corresponds to any extractor hash, so this deletion might implicitly fail and do nothing, and the file be left there for it to trigger neverending activity from tracker-extract-3 in this fruitless loop.

Handle this situation where the hash is unknown, resorting to the deletion of the nfo:FileDataObject from the content graph, and fixing this situation.

Merge request reports