RHEL6 testsuite results

[David Smith <dsmith at redhat dot com>]

I've been going over testsuite results on RHEL6
(2.6.32-358.6.1.el6.x86_64), and here's what I'm seeing. I've cleaned up
a lot of problems, both in systemtap and in the testsuite itself. Here's
where we stand now:

====
		=== systemtap Summary ===

# of expected passes		3498
# of unexpected failures	94
# of unexpected successes	1
# of expected failures		289
# of unknown successes		1
# of known failures		28
# of untested testcases		81
# of unsupported tests		4
====

94 failures isn't too bad for the systemtap testsuite, especially since
80 failures come from one testcase. Note that since dyninst isn't
available on RHEL6, that code wasn't tested.

Here's a detailed look at the failures.


*** 1 systemtap.base/bz10078.exp failure

FAIL: bz10078 -p5 (2)

Note that this testcase is misnamed - this is a testcase for PR10078
(uretprobes: longjmp fix broke funcs returning structs), not for
BZ10078. This bug was originally filed against the original version of
uprobes (RHEL5-era), but this is also a problem with uprobes2 (which
RHEL6 uses). Basically, if a function returns a struct, process.return
probes don't fire.


*** systemtap.base/utrace_p5.exp and systemtap.unprivileged_myproc.exp
failures

FAIL: UTRACE_P5_01_cmd shutdown (eof)
FAIL: unprivileged myproc: --privilege=stapusr process.thread.end

These failures are odd. The 'UTRACE_P5_01_cmd' failure is for a
process.end probe that doesn't fire. The myproc failure is for a
process.thread.end probe that doesn't fire. One of the odd things is
that the utrace_p5.exp testcase has a subtest for process.thread.end
which passes successfully.


*** systemtap.examples/check.exp failure:

FAIL: systemtap.examples/profiling/perf run

This one is related to PR15727. Basically, some of the hardware perf
probes aren't supported, depending on your processor or virtual
processor. It would be possible to kfail this failure, if we could
figure out what to check for.


*** 80 systemtap.exelib/exelib.exp failures:

FAIL:
ustack-uprobesgcc-O3-m32-debug-uprobeslibgcc-O3-m32-debug_uprobeslibgcc-O3-m32-debug
main (0)
FAIL:
ustack-uprobesgcc-O3-m32-debug-uprobeslibgcc-O3-m32-debug_uprobeslibgcc-O3-m32-debug
main_func (2)
... list truncated ...
FAIL:
ustack-uprobesgcc-O3-m32-pie-sep-debug-uprobeslibgcc-O3-m32-prelink-sep-debug-after_uprobeslibgcc-O3-m32-prelink-sep-debug-after
lib_main (0)
FAIL:
ustack-uprobesgcc-O3-m32-pie-sep-debug-uprobeslibgcc-O3-m32-prelink-sep-debug-after_uprobeslibgcc-O3-m32-prelink-sep-debug-after
lib_func (2)

All 80 of these come from the same problem, that 32-bit-on-64-bit exes
have truncated backtraces. I've filed PR15757 on this issue.


*** systemtap.pass1-4/semok.exp

FAIL: semok/thirtynine.stp

This is a failure of finding parameters to an inlined function.


*** systemtap.printf/sharedbuf.exp

FAIL: buffer sharing (1, 0)

This is a problem with module symbol versioning. I've filed PR15775 on
this issue.


*** systemtap.syscall/nd_syscall.exp and systemtap.syscall/syscall.exp
failures

FAIL: 32-bit clock nd_syscall
FAIL: 32-bit rt_signal nd_syscall
FAIL: 32-bit timer nd_syscall
FAIL: 32-bit clock syscall
FAIL: 32-bit rt_signal syscall
FAIL: 32-bit timer syscall

These failures are examples of bug PR15219 (syscall.exp failures on
RHEL5, RHEL6, and rawhide). Basically we have trouble when compat
wrappers copy things from user memory to kernel memory, then we try to
access that kernel memory as user memory.


*** systemtap.unprivileged/unprivileged_embedded_C.exp failures

FAIL: unprivileged embedded C: no embedded C: --unprivileged:
user_string_warn (long)
FAIL: unprivileged embedded C: no embedded C: --privilege=stapusr:
fastcall()

I believe these are related to PR15625 (unprivileged_embedded_C test
confused by stub registers.stp).

-- 
David Smith
dsmith@redhat.com
Red Hat
http://www.redhat.com
256.217.0141 (direct)
256.837.0057 (fax)