Note: I believe the numberformat06,12 and one of the string tests fail on certain JDK 1.1x implementations because of subtle bugs in the JDK implementations, not in Xalan. (You could put it another way and say that the tests could be rewritten to use different data sets, but...) So some of the differences may be known.
Not sure about the idkey set tho - that sounds suspicous. (Unless it's a wierd mismatch with the gold files in that the start of random id number assignment is different depending on some JDK method) - Shane ---- you [EMAIL PROTECTED] wrote ---- BTW, I just tested the currently checked-in code by compiling it under a more recent JDK and attemting to run it under IBM's JDK 1.1.8. A fairly large number of tests fail in that environment -- much of the "idkey" set, numberformat06 and 16, and string132. I haven't isolated the problems yet, but this bodes not well for the assertion that code built on the new platforms will run on the older ones.
