SIGTSTP propagation for 0.80 ?

David Golden dgolden at cp.dias.ie
Mon Sep 19 19:42:13 EDT 2005


> As a related issue, I've seen some comments that qsig -s TSTP does the right 
> thing, but I cannot reproduce this myself. 

N.B. That was with the STOP-propagation-on-TSTP patch I sent for 0.80 (conceptually
similar to an earlier patch by Sébastien Georget for 0.78, though not
really sharing code as there were quite big changes to mpiexec itself 
in between 0.78 and 0.80).

If you were to make "qsig -s suspend" send SIGSTOP to all tasks, not just
the mother superior's, that would likely eliminate the need for the patch to
mpiexec for reasons of allowing suspension (but it might still be desirable for
mpiexec to do things on TSTP to deal with job-control ("ctrl-z") in interactive 
jobs): I would say my and guess Sébastien's patches were written under
the assumption signals would only go to tasks on the mother superior.




More information about the mpiexec mailing list