[lsst-dm-stack-users] LSST DM Stack demo

[Tim Jenness]

On Jun 11, 2014, at 22:38 , Mario Juric <majuric at gmail.com<mailto:majuric at gmail.com>> wrote:

On 6/11/14, 20:06 , Tim Jenness wrote:

That segv still worries me though. DYLD_LIBRARY_PATH does seem to be
filled with LSST stuff so it doesn't seem like it's being confused by
other libcfitsio libraries I have.


Well, that's interesting (the bad kind of interesting).

Could you send the contents of DYLD_LIBRARY_PATH?

My guess is that one of the libraries somehow picked up the headers
and/or got linked against the other cfitsio library you have, but at
runtime picks up the library we build (or vice versa).


Very good. You are correct.

I had a quick look at that path:

$ echo $DYLD_LIBRARY_PATH
/Users/timj/work/lsst/DarwinX86/shapelet/8.0.0.0+3/lib:/Users/timj/work/lsst/DarwinX86/meas_extensions_multiShapelet/8.0.0.0+3/lib:/Users/timj/work/lsst/DarwinX86/skymap/8.0.0.0+3/lib:/Users/timj/work/lsst/DarwinX86/coadd_chisquared/8.0.0.0+3/lib:/Users/timj/work/lsst/DarwinX86/coadd_utils/8.0.0.0+3/lib:/Users/timj/work/lsst/DarwinX86/ip_diffim/8.0.0.0+3/lib:/Users/timj/work/lsst/DarwinX86/ip_isr/8.0.0.0+3/lib:/Users/timj/work/lsst/DarwinX86/astrometry_net/0.30+3/lib:/Users/timj/work/lsst/DarwinX86/meas_astrom/8.0.0.0+3/lib:/Users/timj/work/lsst/DarwinX86/meas_algorithms/8.0.0.0+3/lib:/Users/timj/work/lsst/DarwinX86/minuit2/5.22.00+2/lib:/Users/timj/work/lsst/DarwinX86/wcslib/4.14+3/lib:/Users/timj/work/lsst/DarwinX86/fftw/3.3.2+2/lib:/Users/timj/work/lsst/DarwinX86/pex_config/8.0.0.0+1/lib:/Users/timj/work/lsst/DarwinX86/afw/8.0.0.3/lib:/Users/timj/work/lsst/DarwinX86/base/8.0.0.0+1/lib:/Users/timj/work/lsst/DarwinX86/pex_policy/8.0.0.0+1/lib:/star/lib:/star/starjava/lib/i386:/star/starjava/lib/x86_64:/Users/timj/work/lsst/DarwinX86/daf_base/8.0.0.0+1/lib:/Users/timj/work/lsst/DarwinX86/utils/8.0.0.0+1/lib:/Users/timj/work/lsst/DarwinX86/boost/1.55.0.1/lib:/Users/timj/work/lsst/DarwinX86/pex_exceptions/8.0.0.0+1/lib:/Users/timj/work/lsst/DarwinX86/pex_logging/8.0.0.0+1/lib:/Users/timj/work/lsst/DarwinX86/daf_persistence/8.0.0.0+1/lib:/Users/timj/work/lsst/DarwinX86/mysqlclient/5.1.65+3/lib:/Users/timj/work/lsst/DarwinX86/cfitsio/3310+2/lib:/Users/timj/work/lsst/DarwinX86/xpa/2.1.14+2/lib:/Users/timj/work/lsst/DarwinX86/gsl/1.15+2/lib:/Users/timj/work/lsst/DarwinX86/meas_deblender/8.0.0.0+3/lib

and my initial impression was that it was completely rewritten but on closer inspection I see that my original values are sandwiched into the middle of the variable. Removing them fixes the problem. Sorry about that. Maybe your initialisation script can decide whether it is prepending or appending values and whether it should clear previous values before it starts? Seems a bit dangerous to leave it in the middle.

On OS X it is possible to use a relative path in -rpath such that DYLD_LIBRARY_PATH is not required (Apple really don't like people setting it). I have it set because autoconf doesn't quite know how to use relative rpath and, ironically, libcfitsio is the only comment that I need to set it for as that library doesn't use -rpath at all.

For the record I have version v3.36 installed in the other location (so a bit newer than yours although I see that I'm 8 days out of date at the moment). v3.35 introduced explicit version numbering of the shared library indicating binary compatibility.

--
Tim Jenness
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://listserv.lsstcorp.org/pipermail/dm-users/attachments/20140612/38bdcf47/attachment.html>