Re: opannotate throwing objdump: No such file error

[Eric S. <sch...@gm...>]

Maynard Johnson <may...@us...> writes:

> On 07/22/2011 9:43 PM, Eric Schulte wrote:
>> Maynard Johnson<may...@us...>  writes:
>>
>>> Eric Schulte wrote:
>>>> Maynard Johnson<may...@us...>  writes:
>>>>
>>>>> Eric Schulte wrote:
>>>>>>>> The current version of oprofile just gives a warning for vdso stuff
>>>>>>>> -- e.g, warning: [vdso] (tgid:56455
>>>>>>>> range:0xfff8e6c0000-0xfff8e6e0000) could not be found.
>>>>>>>>
>>>>>>>> But since you're passing in an image specification (coverage.neg),
>>>>>>>> opannotate shouldn't even be trying to objdump anything else (I
>>>>>>>> wouldn't think).  What version of oprofile are you running?  Can you
>>>>>>>> download the latest 0.9.7-RC2
>>>>>>>
>>>>>>
>>>>>> I'm getting the same error
>>>>>>
>>>>>> objdump: '[vdso] (tgid:8508 range:0x909000-0x90a000)': No such file
>>>>>
>>>>> If you are trying to annotate the entire profile (without passing an
>>>>> image specification), then opannotate *will* try to objdump every
>>>>> binary for which samples were taken.  For "vdso" stuff, there are no
>>>>> physical binaries, so this error is expected.
>>>>
>>>> I am passing it an image, specifically...
>>>>
>>>>    opannotate -a coverage.pos
>>>
>>> Oh, yeah -- forgot that you already said that.  Juggling too many things this week.
>>>
>>
>> I know the feeling :)
>>
>>>
>>>>
>>>> where coverage.pos is a binary which does appear to have samples in
>>>> /var/lib/oprofile/samples/
>>>>
>>>>> The opannotate continues on and should give you a complete annotated
>>>>> output for all binaries it can find.
>>>>>
>>>>
>>>> not what I'm seeing, instead I get a number of errors like the above on
>>>> STDERR and the attached file on STDOUT (all comment lines, no actual
>>>> source or annotation).
>>>
>>> I can't reproduce this problem.  Apparently not a new problem, since
>>> you saw it with 0.9.6, too (which is going on two years old).  So
>>> evidently it's something unique to your environment that we're
>>> tripping over.  Please try running with '--verbose=all'.  It's
>>> possible the verbose output may not be helpful.
>>
>> A short shell transcript is attached [1] showing first the output of the
>> file command demonstrating that the executable exists and is a 32-bit
>> binary, then showing the verbose output of opannotate -a.
>>
>>> Can you do some debugging and try to narrow this down?
>>
>> I'm afraid I wouldn't know where to start.  I can say that I have
>> generated this new instance of the exception on an old 32-bit laptop (a
>> different machine than used previous).  This new machine is also debian
>> based (ubuntu) rather than fedora based as the previous one was.
>>
>>> Are you getting the "objdump:<blah>': No such file" error for any
>>> binaries that you *know* exist?
>>>
>>
>> I am confident that this binary exists.  Would it be helpful if I send
> When you say "this binary", I assume you mean 'coverage.pos'.  But the
> output you sent me does *not* show an error message for that binary.
>
> Can you please send the output of 'opreport -l coverage.pos'.
>

Yes, the output for "opreport -l coverage.pos" is attached.

Hope this helps -- Eric