The attached patch adds the required clirr-maven-plugin configuration.
It currently reports the following problems:
[ERROR] org.apache.tika.io.ByteArrayOutputStream: Class org.apache.tika.io.ByteArrayOutputStream removed
[ERROR] org.apache.tika.io.IOUtils: Method 'public java.io.InputStream toBufferedInputStream(java.io.InputStream)' has been removed
[ERROR] org.apache.tika.metadata.MSOffice: Changed type of field LAST_PRINTED from java.lang.String to org.apache.tika.metadata.Property
[ERROR] org.apache.tika.metadata.MSOffice: Changed type of field LAST_SAVED from java.lang.String to org.apache.tika.metadata.Property
[ERROR] org.apache.tika.sax.SecureContentHandler: Parameter 2 of 'public SecureContentHandler(org.xml.sax.ContentHandler, org.apache.tika.io.CountingInputStream)' has changed its type to org.apache.tika.io.TikaInputStream
The first two are from revision 1125422 for TIKA-375 to get rid of unused code.
The second two are from revision 1100053 for
TIKA-656 to get properly typed metadata keys.
The last one is from revision 1124385 for
TIKA-645 to avoid extra layers of stream wrapping.
All of these could we worked around fairly easily to restore full backwards compatibility to Tika 0.9. The question is whether we want to do so, especially since the jump from 0.x to 1.x offers a clean point for getting rid of old baggage in the API.
If we don't want to fix these issues, then we should commit this change only after 1.0 is released, and use the 1.0 release as the reference point for future compatibility checks.