Re: [Libunwind-devel] [PATCH] testsuite evaluation/fixes for ARM

[David Mosberger-Tang]

Bruna,

I checked your patches into the GIT tree (please speak up if anyone
has a problem with these patches...).

I think it would be useful to summarize your test-results for ARM in
the README (** Expected results on ARM...).  Do you want to send me a
patch for that?

Thanks,

  --david

On 4/17/08, Bruna Moreira <address@hidden> wrote:
> Hi all,
>
>  We have done a quick evaluation of current libunwind status on ARM
>  using the internal testsuite. Libunwind from development git tree [1]
>  (last change 5 Feb 2008). The following tests have failed or received
>  SIGSEGV on ARM. Detailed descriptions about errors can be found below:
>
>  * Gtest-bt and Ltest-bt
>     Issue: The stack size passed to sigaltstack() is too small for ARM
>  thus causing segmentation fault due to stack overflow.
>     Feature affected: None (testcase bug)
>
>  * Gtest-exc and Ltest-exc
>     Issue: unw_resume() is not implemented
>     Feature affected: Resume execution at a different stack frame
>  (useful for e.g. exception handling)
>
>  * Gtest-resume-sig and Ltest-resume-sig
>     Issue: Currently it PASS but their results depend on unw_resume()
>  which is not implemented on ARM, so they should FAIL
>     Feature affected: Resume execution from a signal handler
>
>  * Gtest-dyn1 and Ltest-dyn1:
>     Issue: Size code definition of dynamic function (template()) on
>  testcase is too big for ARM architecture so memcpy() reads invalid
>  memory. After match the sizes, the test will still fail because
>  unw_is_signal_frame() is not implemented on ARM
>     Feature affected: Stack unwinding of dynamic (at run-time) generated code
>
>  * run-check-namespace:
>     Issue: Some "bashisms" in check-namespace.sh.in causes the script
>  to fail to run on N8XX's busybox shell. After fixing those, the test
>  still fails because the ARM version of libunwind have _Unwind_*
>  interfaces disabled
>     Feature affected: break exceptions completely for ARM [2]
>
>  * test-ptrace:
>     Issue: Some register maps are missing for ARM in
>  _UPT_reg_offset.c. After add them, we have the current implementation
>  is not able to retrieve the function name of the attached program,
>  probably because the offset of the dynamic linked list where these
>  informations are found is different from others archs. The function
>  get_list_addr() is returning -UNW_ENOINFO (just implemented for IA64)
>     Feature affected: Remote backtraces on ARM
>
>  * run-ptrace-mapper and run-ptrace-misc:
>     Issue: Needs test-ptrace working. Result invalid
>     Feature affected: Not tested
>
>  * test-setjmp
>     Issue: The test is in infinite loop. There are some issues there:
>  definition of JB_SP is incorrect for ARM on file
>  include/tdep-arm/jmpbuf.h. Implementation of longjmp() in libunwind
>  API is not recovering the SP register value
>     Feature affected: Internal support for non-local gotos
>
>
>  Please, see attached the following patches:
>  - libunwind_arm_fixes.patch: add some missing bits on ARM platform.
>  - libunwind_testsuite_fixes.patch: fix some testsuite issues on ARM
>  (Gtest-bt, Gtest-dyn1 and  check-name-space.in).
>
>
>  Any comments/suggestions are welcome!
>
>
>  BR,
>  Bruna.
>
>  References:
>  [1] The official libunwind git repository. URL:
>  
> http://git.kernel.org/gitweb.cgi?p=libs/libunwind/libunwind.git;a=commit;h=3842dac7333e42aa44531eda34ba55200b99ccf8
>  [2] libunwind ARM port. URL:
>  http://thread.gmane.org/gmane.comp.lib.unwind.devel/270/focus=276
>
>
>
>  --
>  Bruna Moreira
>  Instituto Nokia de Tecnologia
>  Manaus - Brazil
>
> _______________________________________________
>  Libunwind-devel mailing list
>  address@hidden
>  http://lists.nongnu.org/mailman/listinfo/libunwind-devel
>
>
>


-- 
Mosberger Consulting LLC, http://www.mosberger-consulting.com/