[OE-core] gdb failure on simplest hello world program -- oe-selftest

[ChenQi]

On 12/29/2018 01:09 AM, Khem Raj wrote:
>
>
> On 12/28/18 8:31 AM, richard.purdie at linuxfoundation.org wrote:
>> On Fri, 2018-12-28 at 17:49 +0800, ChenQi wrote:
>>> On 12/26/2018 07:06 PM, richard.purdie at linuxfoundation.org wrote:
>>>> On Wed, 2018-12-26 at 10:09 +0800, ChenQi wrote:
>>>>> Hi All,
>>>>>
>>>>> 'oe-selftest -r package.PackageTests.test_gdb_hardlink_debug'
>>>>> fails.
>>>>> I did the tests on both my Fedora27 and ubuntu16.04 servers. They
>>>>> both
>>>>> failed.
>>>>> I can also reproduce this issue manually.
>>>>>
>>>>> I did some investigation. No clue how to fix this it. So I want
>>>>> to seek
>>>>> help from community.
>>>>>
>>>>> Below is some output from oe-selftest.
>>>>> [snip]
>>>>> 2018-12-26 09:27:58,440 - oe-selftest - INFO -
>>>>> test_gdb_hardlink_debug
>>>>> (package.PackageTests)
>>>>> 2018-12-26 09:37:55,333 - oe-selftest - INFO - gdbtest
>>>>> /usr/bin/hello1
>>>>> 2018-12-26 09:37:59,904 - oe-selftest - ERROR - GDB result:
>>>>> 1: Reading symbols from /usr/bin/hello1...Reading symbols from
>>>>> /usr/bin/.debug/hello4...done.
>>>>> done.
>>>>> (gdb) Breakpoint 1 at 0x1060: file /usr/include/bits/stdio2.h,
>>>>> line 107.
>>>>> (gdb) Starting program: /usr/bin/hello1
>>>>>
>>>>> Breakpoint 1, main () at /usr/include/bits/stdio2.h:107
>>>>> 107    /usr/include/bits/stdio2.h: No such file or directory.
>>>>> (gdb) Continuing.
>>>>> Hello World!
>>>>> [Inferior 1 (process 312) exited normally]
>>>>> (gdb)
>>>>> 2018-12-26 09:38:00,907 - oe-selftest - INFO -  ... FAIL
>>>>> [snip]
>>>>>
>>>>> The hello world problem is as follows.
>>>>> $ cat
>>>>> meta-selftest/recipes-test/selftest-hardlink/selftest-
>>>>> hardlink/hello.c
>>>>> #include <stdio.h>
>>>>>
>>>>> int main() {
>>>>>       printf("Hello World!\n");
>>>>> }
>>>>>
>>>>> Some extra information:
>>>>> 1. Remove '-O2' or '-D_FORTIFY_SOURCE=2', and it works correctly.
>>>>> 2. Adding some system call like daemon(0,0) before printf also
>>>>> works.
>>>>> 3. Reverting poky repo to
>>>>> 1c048479af7cc98e30f9a33d1f3bef03711544aa,
>>>>> which adds this selftest case, does not work. It still fails.
>>>>> 4. Hongxu confirms that things work on his host.
>>>>> 5. I also don't see failure about this test case on yocto's
>>>>> autobuilder.
>>>>>
>>>>> Could somebody help to provide some suggestions?
>>>>
>>>> This is strange and there must be some difference between your
>>>> configuration and what what we test on the autobuilder or on
>>>> Hongxu's
>>>> setup.
>>>   Hi Richard,
>>>
>>> I have talked with Hongxu and we can now both confirm that there does
>>> exist some problem.
>>>
>>>> Which DISTRO and MACHINE are you using?
>>>   DISTRO is poky
>>> MACHINE is qemux86.
>>>
>>> It seems that if MACHINE is qemux86-64, the test case could pass. But
>>> that's just because the test case only checks 'Breakpoint 1, main ()
>>> at hello.c:4'. See the console output below.
>>> (gdb) b main
>>> Breakpoint 1 at 0x1050: file /usr/include/bits/stdio2.h, line 107.
>>> (gdb) r
>>> Starting program: /usr/bin/hello1
>>>
>>> Breakpoint 1, main () at hello.c:4
>>> 4        printf("Hello World!\n");
>>> (gdb) c
>>> Continuing.
>>> Hello World!
>>> [Inferior 1 (process 156) exited normally]
>>> (gdb) q
>>>
>>> Even in this case, the 'b main' does not give correct result. The
>>> test case passes because 'Breakpoint 1, main () at hello.c:4' is
>>> there.
>>>
>>> On qemux86, we have:
>>> (gdb) Breakpoint 1 at 0x1060: file /usr/include/bits/stdio2.h, line
>>> 107.
>>> (gdb) Starting program: /usr/bin/hello1
>>>
>>> Breakpoint 1, main () at /usr/include/bits/stdio2.h:107
>>> 107    /usr/include/bits/stdio2.h: No such file or directory.
>>> (gdb) Continuing.
>>> Hello World!
>>> [Inferior 1 (process 156) exited normally]
>>> (gdb)
>>>
>>>> I think the key to resolving this is will be to understanding how
>>>> to
>>>> reproduce it...
>>>   Two ways to reproduce the problem:
>>>
>>> I) reproduce it via oe-selftest
>>> 1) . oe-init-build-env build-selftest
>>> 2) SANITY_TESTED_DISTROS = ""
>>> 3) oe-selftest -r package.PackageTests.test_gdb_hardlink_debug
>>>
>>> II) reproduce it manually
>>> 1) add meta-selftest to bblayers.conf
>>> 2) IMAGE_INSTALL_append = " selftest-hardlink selftest-hardlink-dbg
>>> selftest-hardlink-gdb"
>>> 3) bitbake core-image-minimal
>>> 4) On target:
>>>       gdb /usr/bin/hello1
>>>       (gdb) b main
>>
>> I confirmed I could also reproduce it. We've just been lucky with the
>> configuration the autobuilder tests.
>>
>> I think we need to improve the test case. Does installing the libc
>> headers/dev package so that stdio2.h is there help?
>
> This is not exported API so libc headers may not work but installing 
> libc dev package should help here.
>

Hi Richard and Khem,

Unfortunately installing libc-dev package does not help.

On qemux86:
1: Reading symbols from /usr/bin/hello1...Reading symbols from 
/usr/bin/.debug/hello4...done.
done.
*(gdb) Breakpoint 1 at 0x1060: file /usr/include/bits/stdio2.h, line 107.*
(gdb) Starting program: /usr/bin/hello1

*Breakpoint 1, main () at /usr/include/bits/stdio2.h:107**
**107      return __printf_chk (__USE_FORTIFY_LEVEL - 1, __fmt, 
__va_arg_pack ());*
(gdb) Continuing.
Hello World!
[Inferior 1 (process 309) exited normally]
(gdb)
2018-12-29 10:58:11,538 - oe-selftest - INFO -  ... FAIL

For qemux86-64, things are the same.

Best Regards,
Chen Qi

>>
>> Cheers,
>>
>> Richard
>>
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openembedded.org/pipermail/openembedded-core/attachments/20181229/df9dea48/attachment.html>