Updating from 0303 to 0505 has problems

[Quentin Barnes <qbarnes at urbana dot css dot mot dot com>]

I wanted to start pushing some of my systemtap runtime changes back
to the mailing list, so I just updated Systemtap on my ARM system
from 20070303 to 20070505 to verify my work.

With 0505 I'm getting dozens of new "eof" failures such as:

	FAIL: systemtap.base/add.stp shutdown (eof)

This test worked as expected in 0303, but with 0505 so it's failing
with "eof".  I've gone over all my ARM patches and I don't see
anything I did wrong or needs to be changed with the move.  I don't
see any other new failures (so far), so I think everything else is
working.  Here's a longer version of the failing run:
====
....
Pass 4: compiled C into "stap_e26f197cff9d02147ee429af631d0efb_473.ko" in 71350usr/23200sys/98020real ms.
Pass 5: starting run.
systemtap starting probe
PASS: systemtap.base/add.stp startup
PASS: systemtap.base/add.stp load generation
poll warning: Interrupted system call
systemtap ending probe
systemtap test success
Pass 5: run completed in 400usr/1940sys/3481real ms.
FAIL: systemtap.base/add.stp shutdown (eof)
testcase /usr/src/systemtap-20070505/testsuite/systemtap.base/add.exp completed in 122 seconds
====

The common theme in the eof failures is the "poll warning: Interrupted
system call" diagnostic.  (Nothing unusual in the system log either,
standard systemtap load notification in both cases.)  This poll warning
message seems to come from new code in runtime/staprun/relay.c.  I've
noticed a lot of changes in the relayfs area between 0303 and 0505.

I'm using a 2.6.20.4 kernel compiling with gcc version 4.1.1
(CodeSourcery ARM Sourcery G++ 2006q3-26) on a ARM PXA27x board.
On the exact same system, kernel, and compiler, I can still run
0303 tests and they pass fine, so it should have to do with 0505.
(Okay, maybe an unexercised problem in one of the other items
is possible too.)

Anybody else see a problem like this?  Anything to try to help
isolate the cause?

Quentin