Test result of utrace on PPC64(based on FC6 Test 3 release).

[Yong Zheng <zhengyo at cn dot ibm dot com>]

hi all,

Based on the FC6 Test 3, we do one thorough test of utrace on PPC64. The
test still consists of the following two parts and we summarize the test
results as followings, wishing it can be helpful for you to get to know
the progress of utrace on PPC64.

1. Gdb's test on FC6 Test 3 release.
The followings are some packages relevant with this test:

* kernel: 2.6.17-1.2630.fc6 (shipped with FC6 Test 3)
* gcc: 4.1.1 20060828 (Red Hat 4.1.1-20)(shipped with FC6 Test 3)
* gdb-6.5: down load from gdb's website.

In order to know the progress of utrace on PPC64 more easily, we list
all test results including on FC5(kernel: 2.6.17-1.2145_FC5, this kernel
is with the original ptrace and does not apply utrace's patch), on FC6
T2(kernel: 2.6.17-1.2517_fc6, the original ptrace in the kernel is
replaced by utrace) and on FC6 T3.

Table1 Test results for 32on64on64
Testcase               FC5          FC6 T2        FC6 T3   
expected passes       10673          10660         10654
unexpected failures   339(3.18%)     356(3.34%)    355(3.332%)
expected failures     40             40            40
known failures        59             59            59
unresolved testcases  15             14            13
untested testcases    5              4             8
unsupported tests     11             11            11

Table2 Test results for 64on64on64
Testcase               FC5          FC6 T2      FC6 T3   
expected passes        10684        10665        10637
unexpected failures    392(3.67%)   409(3.835%)  421(3.958%)
expected failures      44           44           44
known failures         60           60           60
unresolved testcases   4            4            4
untested testcases     1            1            4
unsupported tests      12           14           12

Notes:
1)32on64on64 means 32-bit cases on 64-bit gdb and on 64-bit kernel.
2)"ulimit -c unlimited" is run at all tests except the 64on64on64 test
based on FC6 Test2.(sorry for this mistake)
3)On all 64on64on64 tests, CFLAGS_FOR_TARGET='-m64' is passed as one
RUNTESTFLAGS option when running gdb's testcases, which can make all
testcases into 64-bit mode except assembly programs.
4)On all 64on64on64 tests, ASFLAGS_FOR_TARGET='-a64 -mppc64' and
LDFLAGS_FOR_TARGET='-m elf64ppc' are passed as RUNTESTFLAGS options just
for the test based on FC6 T3.(on tests based on FC5 and FC6 T2, we
didn't realize this problem. sorry again for this mistake)

The figure in Table1 shows the progress of utrace, the unexpected
failure rate has a mild decrease(from 3.334% to 3.332%). As for
64on64on64, the unexpected failure rate should have some decrease if
considering item 3) and item 4) in the above Notes. 

2. utrace's testsuites on FC6 Test 3 release.

In this part, we run the test cases from roland's website: ntrace-0.0.6
and crash-suspend. 

The case crash-suspend seems to work well, but after the crashed process
has stopped, we fail when attaching to it by gdb, read the following
error messages:

Attaching to
program: /home/zhengyong/working/utrace_test/crash-suspend/crash,
process 3069
../../gdb/linux-nat.c:1057: internal-error: linux_nat_attach: Assertion 
`pid == GET_PID (inferior_ptid) && WIFSTOPPED (status) && WSTOPSIG (status) == SIGSTOP' failed.
A problem internal to GDB has been detected,
further debugging may prove unreliable.
Quit this debugging session? (y or n)

We traced down the error and find WSTOPSIG (status) return 0 but not SIGSTOP, 
which makes the gdb fail to attach the process stopped by utrace. We're not sure about that the 
utrace may not set the process's stop status properly in kernel.

About the ntrace packages, there're still some failures, read the following for details:

* testsuite's result:

  1: insmod bad parameter                         ok
  2: module initializer failure                   ok
  3: basic event reporting                        FAILED (t_basic.at:52)
  4: attach on clone                              ok
  5: implicit detach on reap                      FAILED (t_reap.at:33)
  6: vfork-done event                             FAILED (t_vfork.at:37)
  7: quiesce                                      ok
  8: read first general register                  ok
  9: single-step out of system call               FAILED (t_step.at:37)

* testsuite-biarch's result:

  1: insmod bad parameter                         ok
  2: module initializer failure                   ok
  3: basic event reporting                        FAILED (t_basic.at:52)
  4: attach on clone                              ok
  5: implicit detach on reap                      FAILED (t_reap.at:33)
  6: vfork-done event                             FAILED (t_vfork.at:37)
  7: quiesce                                      ok
  8: read first general register                  ok
  9: single-step out of system call               FAILED (t_step.at:37)


If the test logs may be helpful, pls let us know and we are willing to send them to you.

Best regards
Yong Zheng