[TIKA-1183] tika-app --server --detect fails with "mark/reset not supported"

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 1.5
• Fix Version/s: 1.5
• Component/s: cli, detector
• Assignee: Unassigned
• Reporter: Nick Burch
• Created: 17/Oct/13 21:54
• Updated: 25/Mar/14 16:21
• Resolved: 22/Oct/13 21:43

Description

If you try running the Tika App in detect server mode, eg with "--server --detect", then your detection will fail with exceptions like:

java.io.IOException: mark/reset not supported
at java.io.InputStream.reset(InputStream.java:347)
at org.apache.tika.parser.microsoft.POIFSContainerDetector.detect(POIFSContainerDetector.java:158)
at org.apache.tika.detect.CompositeDetector.detect(CompositeDetector.java:61)
at org.apache.tika.cli.TikaCLI$10.process(TikaCLI.java:252)
at org.apache.tika.cli.TikaCLI$TikaServer$1.run(TikaCLI.java:803)

I'm not sure if this is a problem with the detector, or if it's a problem with the Tika CLI calling the wrong detector / not doing any buffering

Since detection works best when the stream is rewindable, so that container detectors can work, I think we should probably change the Tika CLI Server to buffer into a TikaInputStream for detection. Not sure if there's a downside to this though?