Boost logo

Boost :

From: Jeff Garland (jeff_at_[hidden])
Date: 2004-09-19 15:50:06


On Sun, 19 Sep 2004 10:29:35 -0500, Aleksey Gurtovoy wrote
> > Some of the date-time failures with gcc2.95 appear to be a runtime
> > configuration issue:
> >
> > Run output []:
> >
> >
../bin/boost/libs/date_time/test/testconstrained_value.test/gcc-2.95.3-stlport-4.5.3-linux/debug/testconstrained_value:
> > error while loading shared libraries: libstdc++-libc6.3-2.so.3: cannot open
> > shared object file: No such file or directory
>
> That was my guess as well; Martin had the following to say, though:
>
> I don't think it is a configuration issue here.
> The library does exist at its default location:
>
> > ll /usr/local/gcc-2.95.3/lib/
> total 4055
> drwxr-xr-x 3 m m 280 May 27 2003 ./
> drwxr-xr-x 9 m root 240 May 27 2003 ../
> drwxr-xr-x 3 m m 88 May 27 2003 gcc-lib/
> -rw-r--r-- 1 m m 370370 May 27 2003 libiberty.a
> -rw-r--r-- 1 m m 2462378 May 27 2003 libstdc++-3-libc6.3-2-2.10.0.a
> -r-xr-xr-x 1 m m 1307235 May 27 2003 libstdc++-3-libc6.3-2-2.10.0.so*
> lrwxrwxrwx 1 m m 30 May 27 2003 libstdc++-libc6.3-
> 2.a.3 -> libstdc++-3-libc6.3-2-2.10.0.a lrwxrwxrwx 1 m m
> 31 May 27 2003 libstdc++-libc6.3-2.so.3 -> libstdc++-3-
> libc6.3-2-2.10.0.so*
>
> There was a report on the jamboost list about similar
> problems at another site. I can't recall whether a bug
> was found or fixed.
>
> Any pointers on how to track this down will be greatly appreciated.

If you look at the date_time/test/Jamfile you will see that the group of tests
that is failing to run is from this group:

   test-suite date_time_core
     :
    [ run testint_adapter.cpp ]
    [ run testtime_resolution_traits.cpp ]
    [ run testwrapping_int.cpp ]
    [ run testconstrained_value.cpp ]
    [ run testgregorian_calendar.cpp ]
    [ run testgeneric_period.cpp ]
   ;

not the sets:

   test-suite date_time_gregorian
     :
   [ run gregorian/testdate.cpp <lib>../build/boost_date_time : : :
$(DATE_TIME_PROPERTIES) ]

 ...etc...

You'll note that additional lib dependency. Pure speculation is that adding
the lib dependency somehow sets things up so that the path to the lib can be
found (perhaps the gcc rpath option which builds the library paths into the
executable?). Note that we have done testing with similar setups here on
Linux and don't have a problem, so this is all just wild guesses...

Jeff


Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk