From mboxrd@z Thu Jan  1 00:00:00 1970
Content-Type: multipart/mixed; boundary="===============6929306058164970317=="
MIME-Version: 1.0
From: Max Filippov <jcmvbkbc@gmail.com>
To: elfutils-devel@lists.fedorahosted.org
Subject: Re: Versioned symbols in elfutils and uClibc
Date: Wed, 29 Apr 2015 03:29:39 +0300
Message-ID: <CAMo8BfJj+RxPr=Df0WBZj2EPUPZhi8WcwRWsbSbU3HVtvG_2fg@mail.gmail.com>
In-Reply-To: 20150428173405.GC3161@blokker.redhat.com

--===============6929306058164970317==
Content-Type: text/plain; charset="utf-8"
MIME-Version: 1.0
Content-Transfer-Encoding: quoted-printable

On Tue, Apr 28, 2015 at 8:34 PM, Mark Wielaard <mjw@redhat.com> wrote:
> On Mon, Apr 27, 2015 at 05:12:01AM +0300, Max Filippov wrote:
>> So, I've switched to the pristine sources from git and forward-ported
>> series pointed to by Anthony and now the results are the following:
>>
>> # TOTAL: 134
>> # PASS:  116
>> # SKIP:  7
>> # XFAIL: 0
>> # FAIL:  11
>> # XPASS: 0
>> # ERROR: 0
>>
>> The log is attached. I'll take a look at the segfaults.
>
>> FAIL: run-strip-reloc.sh
>> =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
>>
>> *** failure --reloc-debug-sections not smaller /home/jcmvbkbc/ws/tensili=
ca/elfutils/elfutils-0.161/src/strip.o
>> FAIL run-strip-reloc.sh (exit status: 1)
>>
>> FAIL: run-elflint-self.sh
>> =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D
>>
>> invalid machine flags: 0x300
>> section [ 7] '.rela.dyn': relocation 0: invalid type
>> [...]
>> section [17] '.dynamic': entry 19: unknown tag
>> section [17] '.dynamic': entry 20: unknown tag
>> section [35] '.symtab': _GLOBAL_OFFSET_TABLE_ symbol refers to section [=
 9] '.text'
>
> These kind of failures often indicate that the backend code couldn't
> be loaded for the ELF file. Which architecture is this? What is the
> ELF EM machine type of the file? Does ebl openbackend () find the correct
> shared library to dlopen?

It's Tensilica Xtensa, EM_XTENSA, 94.
openbackend couldn't dlopen libebl_xtensa.so because
there's no such file.

>> FAIL: run-backtrace-native.sh
>> =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D
>>
>> /home/jcmvbkbc/ws/tensilica/elfutils/elfutils-0.161/tests/backtrace: bac=
ktrace.c: 346: exec_dump: Assertion `((((__extension__ (((union { __typeof(=
status) __in; int __i; }) { .__in =3D (status) }).__i))) & 0xff) =3D=3D 0x7=
f)' failed.
>> ./test-subr.sh: line 84:  2886 Aborted                 LD_LIBRARY_PATH=
=3D"${built_library_path}${LD_LIBRARY_PATH:+:}$LD_LIBRARY_PATH" $VALGRIND_C=
MD "$@"
>> backtrace-child: no main
>> FAIL run-backtrace-native.sh (exit status: 1)
>
> That is strange. waitpid () returned something we didn't expect.

I'll take a look.

>> FAIL: run-stack-d-test.sh
>> =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D
>>
>> --- stack.out
>> +++ -
>> @@ -1,5 +1,5 @@
>>  PID 13654 - core
>>  TID 13654:
>> -#0  0x00000000004006c8 _Z2fui
>> +#0  0x00000000004006c8 fu(int)
>>  #1  0x00000000004004c5 main
>>  /home/jcmvbkbc/ws/tensilica/elfutils/elfutils-0.161/src/stack: tid 1365=
4: shown max number of frames (2, use -n 0 for unlimited)
>> FAIL run-stack-d-test.sh (exit status: 1)
>
> This might be a testsuite bug.
> I assume the configure check for __cxa_demangle failed because you
> don't have libstdc++ installed. Then some tools will not be able to
> demangle (C++) symbols. We should fix the test to SKIP in that case.

Correct. After installing libstdc++ and rebuilding these two tests pass.

>>  #1  0x00000000004006c8 foobar
>>  #2  0x00000000004006c8 bar
>>  #3  0x00000000004006c8 foo
>> -#4  0x00000000004006c8 _Z2fui
>> +#4  0x00000000004006c8 fu(int)
>>  #5  0x00000000004004c5 main
>>  /home/jcmvbkbc/ws/tensilica/elfutils/elfutils-0.161/src/stack: tid 1365=
4: shown max number of frames (6, use -n 0 for unlimited)
>> FAIL run-stack-i-test.sh (exit status: 1)
>
> Likewise.

-- =

Thanks.
-- Max

--===============6929306058164970317==--