Re: Systemtap snap:b6371390 test on kernel 2.6.30-rc3-git2

[Mark Wielaard <mjw@redhat.com>]

Hi Pavan,

On Mon, 2009-04-27 at 17:17 +0530, Pavan Naregundi wrote:
> Results of systemtap snap:b6371390 test on kernel 2.6.30-rc3-git2
> Arch: ppc64
>
> Please contact me, for any other details.

Thanks for testing this combination. Which elfutils version did you use?

> Unexpected failures
> =============================
> FAIL: LOCAL1 (0)

I have seen this fail sporadically, is it failing always for you?

> FAIL: STRUCT1 (0)

Could you post the relevant part of the testsuite/systemtap.log?

> FAIL: backtrace (1 0)
> FAIL: backtrace-unwindsyms (1 0)

This is http://sourceware.org/bugzilla/show_bug.cgi?id=6961
"backtrace from non-pt_regs probe context"

> FAIL: systemtap.base/cast.stp
> FAIL: debugpath-bad (eof)
> FAIL: debugpath-good (eof)

Could you post the relevant parts of the testsuite/systemtap.log for the
above failures?

> FAIL: gtod (9)

This is probably http://sourceware.org/bugzilla/show_bug.cgi?id=5094
"gtod.exp fails on ppc64/i386"

> FAIL: MAXACTIVE01 compilation
> FAIL: MAXACTIVE02 compilation
> FAIL: conditional probes (0)
> FAIL: OVERLOAD1 compilation
> FAIL: OVERLOAD2 compilation
> FAIL: OVERLOAD3 compilation

Could you post the relevant parts of the testsuite/systemtap.log for the
above failures?

> FAIL: compiling sdt.c ""
> FAIL: compiling sdt.c additional_flags=-std=gnu89
> FAIL: compiling sdt.c additional_flags=-ansi
> FAIL: compiling sdt.c additional_flags=-pedantic
> FAIL: compiling sdt.c additional_flags=-ansi additional_flags=-pedantic
> FAIL: compiling sdt.c additional_flags=-O2
> FAIL: compiling sdt.c additional_flags="-O3"
> FAIL: static_uprobes compiling C -g

That probably means the sdt.h header doesn't compile on ppc.
Could you see if testsuite/systemtap.log gives more hints about what is
wrong? What gcc version are you using?

> FAIL: stmtvars - .function
> FAIL: system_func (0,0,0)

Could you post the relevant parts of the testsuite/systemtap.log for the
above failures?

> FAIL: backtrace of yyy_func2 (1)
> FAIL: print_stack of yyy_func2 (1)
> FAIL: backtrace of yyy_func3 (1)
> FAIL: print_stack of yyy_func3 (1)
> FAIL: backtrace of yyy_func4 (1)
> FAIL: print_stack of yyy_func4 (1)

These might need support for the dwarf unwinder on ppc to provide more
accurate backtraces.
http://sourceware.org/ml/systemtap/2009-q2/msg00307.html

> FAIL: integer function arguments -- numeric
> FAIL: unsigned function arguments -- numeric
> FAIL: long function arguments -- numeric
> FAIL: int64 function arguments -- numeric
> FAIL: char function arguments -- numeric
> FAIL: string function arguments -- numeric
> FAIL: integer function arguments -- numeric --kelf --ignore-dwarf
> FAIL: unsigned function arguments -- numeric --kelf --ignore-dwarf
> FAIL: long function arguments -- numeric --kelf --ignore-dwarf
> FAIL: int64 function arguments -- numeric --kelf --ignore-dwarf
> FAIL: char function arguments -- numeric --kelf --ignore-dwarf
> FAIL: string function arguments -- numeric --kelf --ignore-dwarf

Could you post the relevant parts of the testsuite/systemtap.log for the
above failures?

> FAIL: systemtap.examples/general/para-callgraph build
> FAIL: systemtap.examples/general/para-callgraph run

Likewise.

> FAIL: buildok/maxactive01.stp
> FAIL: buildok/signal-all-probes.stp
> FAIL: buildok/thirteen.stp
> FAIL: semok/thirtytwo.stp
> FAIL: semok/twentyeight.stp
> FAIL: semok/twentyfour.stp
> FAIL: semok/twentynine.stp
> FAIL: semok/twentyseven.stp
> FAIL: systemtap.printf/bin6.stp
> FAIL: systemtap.stress/current.stp shutdown (eof)


Some of these might be solved with an updated elfutils (0.141). At least
that was the case for s390.

> FAIL: 64-bit acct
> FAIL: 64-bit net1
> FAIL: 64-bit readwrite
> FAIL: 64-bit signal
> FAIL: 32-bit acct
> FAIL: 32-bit net1
> FAIL: 32-bit readwrite
> FAIL: 32-bit signal
> FAIL: 32-bit statfs

The net1 failures are
http://sourceware.org/bugzilla/show_bug.cgi?id=6991
"accept system call missed on 2.6.27"

For the others look in the systemtap.log file and/or try running the
tests in testsuite/systemtap.syscall by hand. There is a README there
explaining how the tests work and how to use the tcl framework to see
how things should match.

Thanks,

Mark