JDK 9 Build 111 seems to miss some locale data, Lucene tests fail with Farsi and Thai language

[Robert Muir]

On Sat, Mar 26, 2016 at 7:56 AM, Uwe Schindler <uschindler at apache.org> wrote:
>
> (1) Thai's locale does not have working dictionary-based BreakIterator available. The following "check" in Lucene for this fails, because it cannot detect a boundary correctly:

Something sneakier is happening. Months ago when the first jigsaw EA
came out, I hit this same problem, and just like now, could not
reproduce it standalone.

Here is what the lucene test is doing: http://pastebin.com/5YUhjiAa

So Thai works in standalone, but sometimes fails in our tests? Maybe
something else like a compiler issue or depending on other stuff the
JVM has done. I played with it this morning in various ways but cannot
make a simple standalone test that fails!