[Users] ET_2014_05 fails to build on Linux Mint 17
Dumsani Ndzinisa
g14n8326 at campus.ru.ac.za
Thu Jun 26 10:19:16 CDT 2014
On 26/06/2014 17:11, Dumsani Ndzinisa wrote:
> Hi Ian,
>
> Ah I'm sorry I didn't realize that my reply wasn't directed to
> the group. I guess this all has to do with me being very *new*
> on almost everything here! I will append my previous response
> here.
>
> Many thanks to you!!! Commenting out hwloc from the thornlist
> has done the trick. My ET has compiled successfully now. Being
> the newbie that I am, please do expect me to get back again and
> bother you will plenty of elementary questions.
>
> Thanks again.
>
>
> My previous response which was communicated off the group...here.......
>
>> Hi Ian,
>>
>> Yes, I do have a local installation of the hwloc library. And I think
>> you've really hit the
>> nail on the head in your diagnosis. When I went through the make
>> file, and also looked
>> at the error message my understanding was that Cactus is trying to
>> bring in numa
>> when linking, and it looks for it in the wrong place. Running the command
>> dpkg --get-selections | grep hwloc, this is the output I get:
>> hwloc-nox install
>> libhwloc-dev:amd64 install
>> libhwloc5:amd64 install
>>
>> I'll implement the suggestions you gave to me, try to build again and
>> get back here
>> with some feedback.
>
>
>
> On 26/06/2014 12:53, Ian Hinder wrote:
>>
>> On 26 Jun 2014, at 12:51, Dumsani Ndzinisa <g14n8326 at campus.ru.ac.za
>> <mailto:g14n8326 at campus.ru.ac.za>> wrote:
>>
>>> Hi Ian,
>>>
>>> Yes, I do have a local installation of the hwloc library. And I
>>> think you've really hit the
>>> nail on the head in your diagnosis. When I went through the make
>>> file, and also looked
>>> at the error message my understanding was that Cactus is trying to
>>> bring in numa
>>> when linking, and it looks for it in the wrong place. Running the
>>> command
>>> dpkg --get-selections | grep hwloc, this is the output I get:
>>> hwloc-nox install
>>> libhwloc-dev:amd64 install
>>> libhwloc5:amd64 install
>>>
>>> I'll implement the suggestions you gave to me, try to build again
>>> and get back here
>>> with some feedback.
>>
>> Hi Dumsani,
>>
>> Glad to be of help! By the way, you replied just to me, not to the
>> mailing list.
>>
>>>
>>>
>>> On 26/06/2014 12:22, Ian Hinder wrote:
>>>> On 26 Jun 2014, at 11:56, Dumsani Ndzinisa
>>>> <g14n8326 at campus.ru.ac.za <mailto:g14n8326 at campus.ru.ac.za>> wrote:
>>>>
>>>>> Hi everyone,
>>>>>
>>>>> I have a fresh checkout of the latest stable release of the
>>>>> Einstein Toolkit
>>>>> (ET_2014_05) which I'm trying to build on my laptop. The laptop is
>>>>> running on Linux Mint 17, and I have adapted configuration options
>>>>> from
>>>>> the bundled "ubuntu.cfg" file. On my machine, I have gcc version
>>>>> 4.8.2,
>>>>> g++ version 4.8.2, and gfortran 4.8.2 as well.
>>>>>
>>>>> However, when building the toolkit (using the thornlist
>>>>> "einsteintoolkit.th <http://einsteintoolkit.th>)
>>>>> the build always fails no matter what I try. For instance, on my
>>>>> very first
>>>>> attempt, I didn't use any of my locally installed libraries but opted
>>>>> for the
>>>>> ones bundled with the toolkit. There was one error message in that
>>>>> case
>>>>> pointing to the PAPI library having failed to get configured. I
>>>>> then decided
>>>>> to comment out this thorn from the thornlist (in the .th file).
>>>>> From there
>>>>> on, still building with the bundled external libraries, I have been
>>>>> persistently
>>>>> getting an error message (in the linking stage) to the effect a
>>>>> certain
>>>>> library (lnuma) could not be found. Below is the error message
>>>>> that gets
>>>>> returned onto the screen:
>>>>>
>>>>> /usr/bin/ld: cannot find -lnuma
>>>>> collect2: error: ld returned 1 exit status
>>>>> /home/dumsani/Cactus/lib/make/make.configuration:147: recipe for
>>>>> target
>>>>> '/home/dumsani/Cactus/exe/cactus_test-ET6' failed
>>>>> make[1]: *** [/home/dumsani/Cactus/exe/cactus_test-ET6] Error 1
>>>>> Makefile:254: recipe for target 'test-ET6' failed
>>>>> make: *** [test-ET6] Error 2
>>>>>
>>>>> I would appreciate if anyone with some idea came to my rescue here.
>>>>> Cactus and the ET are still very new to me. I've been stuck on
>>>>> this for
>>>>> almost a week now. And I am of the idea that it shouldn't be taking me
>>>>> that long. Together with my advisor, we have made a checkout of Cactus
>>>>> (Llama thornlist) earlier and he did oriente me to the process
>>>>> until we
>>>>> had a successful build (we got the executable) for the llama
>>>>> thornlist. So,
>>>>> now the idea is for me to get some more practice with Cactus/ET by
>>>>> checking out the full toolkit, configuring, building and then
>>>>> running some
>>>>> example simulations provided in the toolkit.
>>>> Hi,
>>>>
>>>> Can you try commenting out the thorn "hwloc" from your thornlist?
>>>> Grepping through the source for "numa", the only use of this
>>>> library seems to be hwloc. hwloc is a thorn which assists in
>>>> binding software threads to physical computational cores and
>>>> processes to processor sockets for efficiency, but it is not
>>>> necessary for the toolkit to run.
>>>>
>>>> There might be something wrong with the hwloc configure script
>>>> (Cactus/arrangements/ExternalLibraries/hwloc/configure.sh). The
>>>> relevant lines appear to be
>>>>
>>>>> # Add libnuma manually, if necessary
>>>>> if grep -q '[-]lnuma' ${HWLOC_LIB_DIR}/libhwloc.la 2>/dev/null; then
>>>>> if ! echo '' ${HWLOC_LIBS} '' | grep -q ' numa '; then
>>>>> HWLOC_LIBS="${HWLOC_LIBS} numa"
>>>>> fi
>>>>> fi
>>>> This seems to be attempting to add the numa library to the link
>>>> line if it is found in libhwloc.la. The error message indicates
>>>> that numa has been added to the link line, but the library is not
>>>> available on the link path.
>>>>
>>>> The most recent change to hwloc, in April (i.e. before the
>>>> release), is:
>>>>
>>>> "Correct detecting whether -lnuma is necessary"
>>>> (http://git.barrywardell.net/?p=arrangements/ExternalLibraries/hwloc.git;a=commitdiff;h=4a3b694661f69df842c12a1e188e42f546dd92e2)
>>>> authoreschnett <eschnett at 152c557c-6d84-4bc7-9b1a-dfca721279c7>
>>>> Sat, 19 Apr 2014 23:07:32 +0200 (21:07 +0000)
>>>> committereschnett <eschnett at 152c557c-6d84-4bc7-9b1a-dfca721279c7>
>>>> Sat, 19 Apr 2014 23:07:32 +0200 (21:07 +0000)
>>>>
>>>> Probably Erik has some insight into this.
>>>>
>>>> We have an automated build and test process running under Ubuntu
>>>> 12.04, and this is working fine. It also builds hwloc from source,
>>>> rather than relying on the Ubuntu version. Do you happen to have
>>>> hwloc installed? Maybe it is conflicting with the self-built version?
>>>>
>>>> dpkg -l hwloc
>>>>
>>>> We should probably upgrade the build and test system to Ubuntu
>>>> 14.04, as that is now the latest stable release. Maybe something
>>>> in Ubuntu has changed with the hwloc library.
>>>>
>>>
>>
>> --
>> Ian Hinder
>> http://numrel.aei.mpg.de/people/hinder
>>
>
>
>
> _______________________________________________
> Users mailing list
> Users at einsteintoolkit.org
> http://lists.einsteintoolkit.org/mailman/listinfo/users
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.einsteintoolkit.org/pipermail/users/attachments/20140626/1827df70/attachment.html
More information about the Users
mailing list