setenv MPI_FLAGS eladebug - NOT working

[Richard Felkins_1]

One of our developers is having this problem.

We're still not at 100% capability with our tools on the Alpha because ladebug 4.69 does not listen to the set MPI_DEBUG_CONT=1
as it's supposed to, per the HP MPI 2.0 User's Guide.

Can anyoneshed some light on this?

P.S. NOTE that if we instead 'setenv MPI_FLAGS egdb' to use the gdb debugger, gdb *does* listen to that set command and proceed, but there are other aspects of gdb that we are not happy with...


-------------------detailed problem description:-------------------

If we do:

setenv MPI_FLAGS eladebug
mpirun -np 2

it fires off 2 ladebug xterms as it should, both stopping in MPI_INIT. You are supposed
to be able to then:

[set breakpoints as desired]
set MPI_DEBUG_CONT=1
continue

in each xterm to get it to proceed, but it doesn't like that 'set' command. I also tried it as an environment variable with no luck--we're just stuck in MPI_INIT.

[Michael Schulte zur Sur]

Hi,

are you using csh?

greetings,

Michael

[Richard Felkins_1]

>I have a response on the forum, they are asking if you are using 'csh'?
>
Yes--

also, I have tried

setenv MPI_DEBUG_CONT 1

once inside the xterms and that doesn't work either. The processes
won't budge past MPI_INIT. But doing

set MPI_DEBUG_CONT=1

using gdb *does* kick it loose. Ladebug says there is
no such variable when I try that.