False matches in Java class files

[kahatlen]

If you index the OpenGrok code base and then search for "bject" in the full text search, you'll get a match in testdata/sources/java/Main.class. However, there is no "bject" token in that file. It's probably "Object" that has been garbled.

This is most likely a regression. Tested the same search on src.opensolaris.org, which runs 0.9, and it returned 0 matches.

[kahatlen]

Pull request #11 moved this problem around. When indexing the OpenGrok source tree using head, the "bject" token is gone. The problem is still there, but now other tokens are garbled instead. For example, try searching full text for "genric" or "ookupswitch", and you'll get matches in bcel-5.2.jar. There are no such tokens in that jar file, as far as I can see. It's probably the tokens "generic" and "lookupswitch" that have been garbled.

[tarzanek]

problem seems to be in in TagFilter, input xref from bcel is ok

[tarzanek]

ok, now I am 100% sure, tagfilter is broken
I will replace it with HTMLStripCharFilter from lucene ...