Re: putenv('LD_LIBRARY_PATH',...) does not work

[Javier Fernández]

Hi,

The author of the LD_LIBRARY... question sent me a different description 
of what seems to be the same problem by personal e-mail. I'm a bit 
puzzled, since it was question #3 in a series, I don't know why he 
decided to switch to public questions in Octave-help. Question #1 was 
dated Jan 20 and question #2 was dated Jan 22. This was questions #3 
dated Jan 25.

In the version of q#3 I received earlier this morning, he said

> Then, I can spawn an octave session. But error message "libmpi.so.0 could
> not be found..." appears after using command "MPI_Init". I found out that
> $LD_LIBRARY_PATH is reset by our system because xterm is setgid,
When it is told this way, it seems as if the master Octave was alright 
not only issuing MPI_Init, but also alright issuing MPI_Comm_spawn. The 
problem seems to appear on the _slave_ Octave, when doing MPI_Init 
there. There is _no_ chances of spawning a slave Octave if MPI_Init  
does not work in the master Octave.

I am subscribed to digested-mode Octave-help and I am rather puzzled to 
learn this evening this other description:

> To use MPITB for Octave, I need to be able to start an octave session in a
> xterm windown, i.e. by using command "xterm -e octave". But I found out
> that $LD_LIBRARY_PATH is reset by our system because xterm is setgid
Since I had already replied by personal e-mail, I decided on a first 
thought not to reply to the list, but let the user tell me which one is 
the correct problem description.

On a second thought, if I don't reply, somebody could later say that 
MPITB is not being supported.

So I better reply here just to mention that MPITB users should direct 
MPITB questions directly to me. Most Octave-help readers simply don't 
use MPITB, and are probably not interested in this stuff.

Sorry for the inconveniences, and thanks for your sympathy.

-javier