On Wed, 23 Apr 2003, Irv Elshoff wrote:
> The problem is NOT stderr, but more general. If LAM wants to run user
> dot profiles it should provide a hint, hook or whatever so that
> subsequent code can anticipate LAM (e.g., by piping stderr to /dev/null
> in an exec).
Ah -- I see. So you're more interested in being able to "short circuit"
parts of the target dot files to avoid running unnecessary things when LAM
is being booted.
Can you give a few examples?
> >1. Add an option to all the "boot" commands (...) that makes LAM ignore
> >anything on stderr and not treat it as an error.
>
> That would probably be nice, but is not sufficient to solve my problem.
>
> >2. Add an ability ... to specify a command line option to insert either
> >a standard environment variable that the remote rsh/ssh target can see
> >in their environment.
>
> As for the 7.0 release, I'm sorry I'm too late, but adding an
> appropriately named exported standard envar to the boot commands (e.g.,
> LAM_MPI) seems innocent. Again, the value is not important (maybe
It's not a question of being innocent -- the issue is twofold:
a) can we support such an environment variable in all boot environments
(LAM 7.0 will support rsh/ssh, Globus, bproc, tm). I think that bproc and
tm are no-brainers, and with a little cleverness, rsh/ssh should be
do-able. Globus is the question -- we'll need to look into that.
b) we're well into the release process of 7.0, and new features -- however
small or large -- are not allowed so that we can have controlled testing
and whatnot. Sorry! :-(
--
{+} Jeff Squyres
{+} jsquyres_at_[hidden]
{+} http://www.lam-mpi.org/
|
