Hi Dave,
On 1/03/2010, at 10:38 AM, case wrote:
> On Mon, Mar 01, 2010, Ben Roberts wrote:
>
>> This morning when I rebuilt and tested AT, some of the nab tests
>> failed with the following:
>>
>> ld: duplicate symbol _second in /opt/intel/Compiler/11.1/076/
>> Frameworks/mkl/lib/em64t/libmkl_intel_lp64.a(_time_mkl_second.o)
>> and /
>> Applications/Amber11/lib/libnab.a(time.o)
>
> Do you have any notion of why this is showing up now and not earlier?
>
> Do you think it is specific to Frameworks? I don't see any problem
> with MKL
> under Linux. My code is also searching libmkl_intel_lp64.a.
>
> Aha: on linux (icc) if I do this:
>
> cd amber11/lib; nm libnab.a | grep second
>
> I see that libnab.a is defining "second"; however, if I do the same
> on a Mac (with gcc), libnab is exporting "_second". And MKL is also
> exporting "_second". What compiler are you using?? And what do you
> get
> on your system with the above nm command?
>
> Don't do anything right now to the code; I think we can eliminate all
> references to second() pretty easily. But it would be nice to track
> down
> exactly what combination is failing.
>
>
First, as to why now and not earlier: I hadn't done a full cvs
checkout for several days, during which time a lot changed. Partly, I
was waiting for things to settle down. I knew people had changed
something to do with second, and figured that the chances were good
that this breakage was related to that.
Secondly, the nm command reports this:
[roberts.qtpclnt070 lib]$ nm libnab.a | grep second
U _seconds
U _seconds
U _seconds
U _second
0000000000000000 T _seconds
00000000000000d0 T _second
I'm not entirely sure what all that means. "man nm" tells me that U
means "undefined" and T means "in the text section". But that doesn't
illuminate me greatly, since I'm not yet familiar with such
programming technical terms.
I've been using the Intel compiler, icc 11.1.076, and its associated
MKL libraries.
- Ben
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Mon Mar 01 2010 - 10:30:03 PST
