Re: quasiofficial Fedora kernel build with VTA debuginfo

[Mark Wielaard <mjw at redhat dot com>]

On Wed, 2009-10-14 at 16:44 -0700, Roland McGrath wrote:
> http://kojipkgs.fedoraproject.org/packages/kernel/2.6.31.4/83.fc12/
> 
> Future F>=12 builds will all have it.

Thanks. I did some (make installcheck) tests against it:

Host: Linux rawhide.wildebeest.org 2.6.31.4-83.fc12.x86_64 #1 SMP Wed
Oct 14 17:55:23 EDT 2009 x86_64 x86_64 x86_64 GNU/Linux
Snapshot: version 1.0/0.143 commit release-1.0-108-gb393f6f + changes
GCC: 4.4.1 [gcc (GCC) 4.4.1 20091008 (Red Hat 4.4.1-20)]
Distro: Fedora release 11.92 (Rawhide)

FAIL: empty-struct resolve-pass compilation failed
FAIL: gtod (201)
FAIL: onoffprobe (timeout)
FAIL: 32_BIT_UTRACE_SYSCALL_ARGS shutdown (eof)
FAIL: buildok/linuxmib-all-probes.stp
FAIL: buildok/seventeen.stp
FAIL: transok/tval-opt.stp
FAIL: 32-bit net1

# of expected passes        1137
# of unexpected failures       8
# of unexpected successes      9
# of expected failures       209
# of unknown successes         3
# of known failures            3
# of untested testcases       16
# of unsupported tests         4

Compared with an earlier non-vta compiled kernel:

Host: Linux rawhide.wildebeest.org 2.6.31.1-56.fc12.x86_64 #1 SMP Tue
Sep 29 16:16:22 EDT 2009 x86_64 x86_64 x86_64 GNU/Linux
Snapshot: version 1.0/0.143 commit release-1.0-108-gb393f6f + changes
GCC: 4.4.1 [gcc (GCC) 4.4.1 20091008 (Red Hat 4.4.1-20)]
Distro: Fedora release 11.92 (Rawhide)

FAIL: empty-struct resolve-pass compilation failed
FAIL: onoffprobe (timeout)
FAIL: systemtap.base/optim_voidstmt.stp shutdown (eof)
FAIL: buildok/linuxmib-all-probes.stp
FAIL: buildok/seventeen.stp
FAIL: 32-bit net1

# of expected passes        1139
# of unexpected failures       6
# of unexpected successes      9
# of expected failures       209
# of unknown successes         2
# of known failures            4
# of untested testcases       16
# of unsupported tests         4

The two regressions are:

FAIL: 32_BIT_UTRACE_SYSCALL_ARGS shutdown (eof)
FAIL: transok/tval-opt.stp

Compared to a systemtap 1.0 testrun (on the vta-enabled-build kernel):

Host: Linux rawhide.wildebeest.org 2.6.31.4-83.fc12.x86_64 #1 SMP Wed
Oct 14 17:55:23 EDT 2009 x86_64 x86_64 x86_64 GNU/Linux
Snapshot: version 1.0/0.143 commit release-1.0-0-g6a7aa1a + changes
GCC: 4.4.1 [gcc (GCC) 4.4.1 20091008 (Red Hat 4.4.1-20)]
Distro: Fedora release 11.92 (Rawhide)

FAIL: empty-struct resolve-pass compilation failed
FAIL: gtod (2)
FAIL: onoffprobe (timeout)
FAIL: systemtap.base/optim_voidstmt.stp shutdown (eof)
FAIL: sdt additional_flags=-pedantic kprobe
FAIL: systemtap.base/vta-test.stp
FAIL: systemtap.examples/network/tcpipstat build
FAIL: systemtap.examples/network/tcpipstat run
FAIL: buildok/linuxmib-all-probes.stp
FAIL: buildok/seventeen.stp
FAIL: transok/tval-opt.stp
FAIL: 32-bit net1

# of expected passes        1112
# of unexpected failures      12
# of unexpected successes      8
# of expected failures       208
# of unknown successes         3
# of known failures            3
# of untested testcases       14
# of unsupported tests         4

Which shows some more failures for issues apparently solved in recent
systemtap git. It also shows the tval-opt.stp regression, but not the
32_BIT_UTRACE_SYSCALL_ARGS for some reason. So we might want to track
down those two regressions and then try to spin a new release if this
vta-enabled-build kernel hits distros soon.

Cheers,

Mark