[LSST|dm-users 527] Re: Fwd: Error installing meas_base during software install
darko at aob.rs
darko at aob.rs
Sat Dec 20 14:12:00 PST 2014
Hi folks.
meas_base and meas_algorithms are not able to do tests on OpenSuse13.1
with gcc version 4.8.1 20130909 [gcc-4_8-branch revision 202388] (SUSE
Linux)
Reason is to agresive optimization (-O3). -O2 doesn't work either, but I
can report success with -O1. So the recipe is as follows
when
eups distrib install -t v10_0 lsst_distrib
crashes on meas_base - open a new terminal, go to you install directory,
do loadLSST.bash (or whatever).
change directory to
$Instal_directory/EupsBuildDir/Linux64/meas_base-10.0+2/
do
export SCONSFLAGS='opt=1'
end do all the setups from build.sh - namely
setup --just --type=build apr 1.3.3.lsst2
setup --just --type=build cfitsio 3360.lsst1
setup --just --type=build doxygen 1.8.5
setup --just --type=build eigen 3.2.0
setup --just --type=build fftw 3.3.3
setup --just --type=build gsl 1.16.lsst1
setup --just --type=build minuit2 5.28.00
setup --just --type=build mysql 5.1.65.lsst1
setup --just --type=build mysqlclient 5.1.73.lsst1
setup --just --type=build python 0.0.2
setup --just --type=build sqlite 3080200
setup --just --type=build swig 3.0.2.lsst1
setup --just --type=build xpa 2.1.15.lsst1
setup --just --type=build apr_util 1.3.4.lsst2
setup --just --type=build boost 1.55.0.1.lsst2
setup --just --type=build mysqlpython 1.2.3+13
setup --just --type=build numpy 0.0.1+2
setup --just --type=build scons 2.3.0+1
setup --just --type=build wcslib 4.14+7
setup --just --type=build activemqcpp 3.5.0+15
setup --just --type=build astrometry_net 0.50.1+3
setup --just --type=build matplotlib 0.0.1+2
setup --just --type=build pyfits 3.2.4.lsst1
setup --just --type=build scisql 0.3.4+2
setup --just --type=build sconsUtils 10.0
setup --just --type=build astrometry_net_data 10.0+1
setup --just --type=build base 10.0+2
setup --just --type=build geom 10.0+1
setup --just --type=build lsst 10.0
setup --just --type=build lsst_thirdparty 10.0+2
setup --just --type=build ndarray 10.0+1
setup --just --type=build pex_exceptions 10.0+1
setup --just --type=build utils 10.0+1
setup --just --type=build daf_base 10.0+2
setup --just --type=build pex_logging 10.0+1
setup --just --type=build pex_policy 10.0+1
setup --just --type=build daf_persistence 10.0+2
setup --just --type=build pex_config 10.0+1
setup --just --type=build afw 10.0+3
setup --just --type=build cat 10.0+2
setup --just --type=build lsst_libs 10.0+2
setup --just --type=build ctrl_events 10.0+4
setup --just --type=build ctrl_provenance 10.0+5
setup --just --type=build shapelet 10.0+2
setup --just --type=build skymap 10.0+2
setup --just --type=build skypix 10.0+2
setup --just --type=build daf_butlerUtils 10.0+3
setup --just --type=build pex_harness 10.0+7
setup --just --type=build ctrl_orca 10.0+8
setup --just --type=build obs_test 10.0+2
setup --just --type=build ctrl_execute 10.0+8
setup --just --type=build pipe_base 10.0+2
setup --just --type=build testing_endtoend 10.0+7
setup --just --type=build coadd_utils 10.0+3
setup --just --type=build ctrl_platform_gordon 10.0+8
setup --just --type=build ctrl_platform_lsst 10.0+8
setup --just --type=build coadd_chisquared 10.0+3
after that change directory to
$Instal_directory/EupsBuildDir/Linux64/meas_base-10.0+2/meas_base-10.0+2
and do
setup --type=build -k -r
./ups/eupspkg VERBOSE=0 FLAVOR=Linux64 config
./ups/eupspkg VERBOSE=0 FLAVOR=Linux64 build
./ups/eupspkg VERBOSE=0 FLAVOR=Linux64 install
If everything goes fine then from your original terminal/window do
eups declare meas_base 10.0+2 -r $Instal_directory/Linux64/meas_base/10.0+2 -t current
eups declare meas_base 10.0+2 -r $Instal_directory/Linux64/meas_base/10.0+2 -t v10_0
restart instalation
Now, meas_algorithms will probably die as well (same recipe):
In terminal you were using to manualy build go to
$Instal_directory/EupsBuildDir/Linux64/meas_algorithms-10.0+2
do just one new setup (others are already done)
setup --just --type=build meas_base 10.0+2
change directory to
$Instal_directory/EupsBuildDir/Linux64/meas_algorithms-10.0+2/meas_algorithms-10.0+2
do
setup --type=build -k -r
./ups/eupspkg VERBOSE=0 FLAVOR=Linux64 config
./ups/eupspkg VERBOSE=0 FLAVOR=Linux64 build
./ups/eupspkg VERBOSE=0 FLAVOR=Linux64 install
Now from the original terminal do eups declare step
eups declare meas_base 10.0+2 -r $Instal_directory/Linux64/meas_algorithms/10.0+2 -t current
eups declare meas_base 10.0+2 -r $Instal_directory/Linux64/meas_algorithms/10.0+2 -t v10_0
and restart installation (this should go smoothly till the end)
Hope this helps in sorting out problems with meas_base and meas_algorithms
on other distributions as well.
Someone should check why this happens and maybe offer more elegant
solution...
Cheers
D.
On Tue, 9 Dec 2014, Mario Juric wrote:
> Hi Mike,
> CentOS 6 is supported (it's out default/reference platform). What problem are you seeing building on CentOS?
> --
> Sent from my Android device with K-9 Mail. Please excuse my brevity.
>
> On December 8, 2014 3:55:51 PM PST, Mike Lund <mike.lund at gmail.com> wrote:
> Thanks for all that information, at the present I have the poor luck of the other machine I've got access to running CentOS 6.6, which is also not supported at this time, so I'll be
> watching for support being extended to Ubuntu 14.04 or CentOS 6.6 so that I can complete and installation and resume working with MAF again.
>
> I'm definitely also willing to try out any other distribution methods that are being explored as a guinea pig of sorts, although from looking at CernVM-FS, it seems like Fabio
> wouldn't have much need of that for that implementation.
>
> - Mike Lund
>
> On Mon, Dec 8, 2014 at 1:27 PM, Dominique Boutigny <boutigny at in2p3.fr> wrote:
> Hi Mike,
>
> I am also using a Ubuntu 14.04 system and I am having the same problem as you. So, what I am doing is to build the stack on another platform (I use Scientic Linux / Redhat)
> and simply make a tar ball that I deploy on my system. This is working very well, you only have to fix a couple of things for eups.
> Here is a copy of the recipe I put in the IN2P3 wiki : http://lsst.in2p3.fr/wiki/index.php/Calcul (you can create an account for yourself, it is not moderated).
>
> Hope this helps
>
> Dominique
>
> PS : Fabio Hernandez at CC-IN2P3 is also working on a system to deploy the LSST stack using CernVM-FS. This is very convenient. You can contact him directly if you wish.
>
> INTRODUCTION
>
> It is possible to replicate a working stack installation into another directory structure or even on another computer without recompiling the packages. Here is how to proceed,
> taking the example of the stack v9_2 installed at CC-IN2P3 :
>
> cd /sps/lsst/Library/stack_v9_2
> tar -cvzf stack.tar.gz *
>
> cp stack.tar.gz some_directory (on the same or on another server)
>
> tar -xvzf stack.tar.gz
>
> At this stage we have the stack directory structure but in order to make it work, we need to fix the following :
>
> o Cleanup any LSST or EUPS environment variable (it may be safer to start a completely new shell)
> o Change LSST_HOME in loadLSST.sh
> o Change EUPS_DIR and EUPS_PATH in eups/bin/setups.sh
>
> Then one can now re-initialize the stack environment :
>
> source loadLSST.sh
>
> The eups database should now be ok.
>
> [MODIFIER]CAVEATS
>
> [modifier]Fixing eups for some manually declared packages
>
> If some packages have been manually declared to eups without specifying a path relative to EUPS_PATH, one have to undeclare those packages in order to re-declare them with the
> right path. For instance :
>
> unsetup obs_cfht my_obs_cfht
> eups undeclare obs_cfht my_obs_cfht
> eups declare -r some_directory/my_packages/obs_cfht obs_cfht my_obs_cfht
> setup obs_cfht my_obs_cfht
>
> [modifier]Fixing ipython
>
> If the stack release comes with the anaconda python distribution, the firs line of the ipython wrap-up script should be fixed to point to the right python path.
>
> [MODIFIER]REMARK
>
> This method is working to install the stack on Ubuntu 14.04 by copying the stack from a SL6.5 distribution
>
> On 12/08/2014 11:17 AM, Mike Lund wrote:
> Dominique, thanks for the quick answer on this.
>
> To follow this up with a probably naive question, are there any options to get this working other than the stack build? If it effects anything, my needs for the software
> are primarily just for writing new metrics for the MAF. I also know that Ubuntu 14.10 is out now, and that is something that I could upgrade to if the issues with gcc
> are limited to just 14.04 and not later versions.
> - Mike Lund
>
> On Sat, Dec 6, 2014 at 1:12 PM, Dominique Boutigny <boutigny at in2p3.fr> wrote:
> Hi Mike,
>
> Until recently, the stack build was incompatible with Ubuntu 14.04 due to some gcc subtleties. So, unless this was fixed in the mean time, this is likely the
> reason of the errors that you are seeing.
>
> Best,
>
> Dominique
> On 12/6/2014 11:03 AM, Mike Lund wrote:
> Hello, Lynne Jones recommended I send this question to here and thanks in advance for any help.
>
> I'm trying to install the LSST software, and my install is failing when trying to meas_base, and the screen output is saying that it's running python test
> scripts and failing 15 tests. I'm trying to install this on Ubuntu 14.04.1 LTS, and the last few lines output to terminal are the following:
>
> scons: *** [checkTestStatus] Error 1
> scons: building terminated because of errors.
> + exit -4
> eups distrib: Failed to build meas_base-master-g5cbca8985d.eupspkg: Command:
> source /home/mike/lsst/eups/bin/setups.sh; export EUPS_PATH=/home/mike/lsst; (/home/mike/lsst/EupsBuildDir/Linux64/meas_base-master-g5cbca8985d/build.sh) >>
> /home/mike/lsst/EupsBuildDir/Linux64/meas_base-master-g5cbca8985d/build.log 2>&1 4>/home/mike/lsst/EupsBuildDir/Linux64/meas_base-master-g5cbca8985d/build.msg
> exited with code 252
>
> I've also attached the build.log for this.
>
> - Mike Lund
>
>
>
> _______________________________________________
> dm-users mailing list
> dm-users at lists.lsst.org
> https://lists.lsst.org/mailman/listinfo/dm-users
>
>
> --
> Dominique Boutigny - CNRS / CC-IN2P3
> Now at SLAC National Accelerator Laboratory
>
> Mail : boutigny at in2p3.fr - boutigny at slac.stanford.edu
> Office : +1 650-926-5759 - Cellular : +1 774-232-0912
>
> _______________________________________________
> dm-users mailing list
> dm-users at lists.lsst.org
> https://lists.lsst.org/mailman/listinfo/dm-users
>
>
>
> --
> Dominique Boutigny - CNRS / CC-IN2P3
> Now at SLAC National Accelerator Laboratory
>
> Kavli Building - Room 221
> Mail : boutigny at in2p3.fr - boutigny at slac.stanford.edu
> Office : +1 650-926-5759 - Cellular : +1 774-232-0912
>
>
> _________________________________________________________________________________________________________________________________________________________________________________________________
> dm-users mailing list
> dm-users at lists.lsst.org
> https://lists.lsst.org/mailman/listinfo/dm-users
>
>
>
More information about the dm-users
mailing list