On Cygwin (don't know if this affects other systems), the test script's
http server seems to fail randomly. None of the other servers seem to
have this difficulty. I typically see about 4 of the tests (out of 72)
fail, though it's never the same set of 4.

I don't know why (though I'm willing to help track it down if someone
can provide guidance), however I hacked in a quick fix that seems to
work fine. It simply watches for failures, and tries them one more time,
after restarting the server. Here's an example of the new standard
(non-short) output, and I'm attaching the diff.

test 2...[HTTP GET with user and password]
curl returned 52, 0 was expected
 error FAILED
 ... trying one more time ...
test 2...[HTTP GET with user and password]
 data OK protocol OK

With this hack in place, I usually get 100%. Though once (out of about 5
tries) I got a repeat failure.

Thanks,
--Kevin
 <<runtests.diff>>