Hello,
If you still want to define as many point sources as points in the mesh, there is nothing to change in difsou.f, as this is already done, so your lines :
do N=1,NPOIN !2014-11-24
TSCEXP%ADR(ITRAC)%P%R(N)=TSCEXP%ADR(ITRAC)%P%R(N)+1 !2014-11-24
enddo !2014-11-24
should be removed, TSCEXP is built just after. However there should be no segmentation fault for this which looks correct in Fortran, except that 1 should be 1.D0. The problem would be rather how you managed to define as many point sources as points. For example there is in declarations_telemac2d a maximum of sources set to 100, called MAXTRA, that triggers the size of a number of arrays. So telemac-2d should be recompiled with a new maximum corresponding to your number of points. This could be a reason for a segmentation fault.
Regards,
JMH