Barry Smith <[email protected]> writes:

> and why doesn't it list the directory it is run from at the bottom like the 
> other "errors" do?

Because the guy that committed runex19_draw did not include any code to
do that.

https://bitbucket.org/petsc/petsc/commits/8eef25e527892391887028685be59b9025c343d6

>
> Possible problem with with ex7_2, diffs above 
> =========================================
> lid velocity = 0.0204082, prandtl # = 1, grashof # = 1
> [0]PETSC ERROR: --------------------- Error Message 
> ------------------------------------
> [0]PETSC ERROR: Object is in wrong state!
> [0]PETSC ERROR: You called KSPSetDM() but did not use 
> DMKSPSetComputeOperators() or KSPSetDMActive(dm,PETSC_FALSE);!

PCFIELDSPLIT is forwarding a fully-functional DM to the inner solver,
but Galerkin has not been set.  I'll merge these if there are no objections.

commit 19ae98dbbafc9c48a2c756250106d26ed65ac43b
Author: Jed Brown <[email protected]>
Date:   Wed Sep 18 17:45:16 2013 -0500

    SNES runex19_draw: use -fieldsplit_x_velocity_pc_mg_galerkin
    
    PCFieldSplit now sends a fully-functional DM into the split, but PCMG
    does not know whether to use it to create an operator.  Since an
    assembly function has not been provided for the split, we'll have to use
    Galerkin.
    
    Reported-by: Barry Smith <[email protected]>

diff --git a/src/snes/examples/tutorials/makefile 
b/src/snes/examples/tutorials/makefile
index ee9a513..ac03600 100644
--- a/src/snes/examples/tutorials/makefile
+++ b/src/snes/examples/tutorials/makefile
@@ -792,7 +792,7 @@ runex19_composite_gs_newton: #test additive composite SNES
         ${RM} -f ex19_composite_gs_newton.tmp
 
 runex19_draw:
-       -@${MPIEXEC} -n 1 ./ex19 -pc_type fieldsplit -snes_view draw 
-fieldsplit_x_velocity_pc_type mg -fieldsplit_x_velocity_pc_mg_levels 2 
-da_refine 1 -fieldsplit_x_velocity_mg_coarse_pc_type svd
+       -@${MPIEXEC} -n 1 ./ex19 -pc_type fieldsplit -snes_view draw 
-fieldsplit_x_velocity_pc_type mg -fieldsplit_x_velocity_pc_mg_galerkin 
-fieldsplit_x_velocity_pc_mg_levels 2 -da_refine 1 
-fieldsplit_x_velocity_mg_coarse_pc_type svd
 
 runex18:
        -@${MPIEXEC} -n 1 ./ex18 -pc_type mg -ksp_type fgmres -da_refine 2 
-pc_mg_galerkin -snes_view > ex18_1.tmp 2>&1; \

commit 97d80a904c5ead878597add8a40e26965dc087f0
Author: Jed Brown <[email protected]>
Date:   Wed Sep 18 17:50:36 2013 -0500

    SNES runex19_draw: check output and report errors
    
    Reported-by: Barry Smith <[email protected]>

diff --git a/src/snes/examples/tutorials/makefile 
b/src/snes/examples/tutorials/makefile
index ac03600..d7ff5c9 100644
--- a/src/snes/examples/tutorials/makefile
+++ b/src/snes/examples/tutorials/makefile
@@ -792,7 +792,9 @@ runex19_composite_gs_newton: #test additive composite SNES
         ${RM} -f ex19_composite_gs_newton.tmp
 
 runex19_draw:
-       -@${MPIEXEC} -n 1 ./ex19 -pc_type fieldsplit -snes_view draw 
-fieldsplit_x_velocity_pc_type mg -fieldsplit_x_velocity_pc_mg_galerkin 
-fieldsplit_x_velocity_pc_mg_levels 2 -da_refine 1 
-fieldsplit_x_velocity_mg_coarse_pc_type svd
+       -@${MPIEXEC} -n 1 ./ex19 -pc_type fieldsplit -snes_view draw 
-fieldsplit_x_velocity_pc_type mg -fieldsplit_x_velocity_pc_mg_galerkin 
-fieldsplit_x_velocity_pc_mg_levels 2 -da_refine 1 
-fieldsplit_x_velocity_mg_coarse_pc_type svd > ex19_draw.tmp 2>&1; \
+          ${DIFF} output/ex19_draw.out ex19_draw.tmp || echo  ${PWD} 
"\nPossible problem with ex19_draw, diffs above 
\n========================================="; \
+          ${RM} ex19_draw.tmp
 
 runex18:
        -@${MPIEXEC} -n 1 ./ex18 -pc_type mg -ksp_type fgmres -da_refine 2 
-pc_mg_galerkin -snes_view > ex18_1.tmp 2>&1; \
diff --git a/src/snes/examples/tutorials/output/ex19_draw.out 
b/src/snes/examples/tutorials/output/ex19_draw.out
new file mode 100644
index 0000000..327cd46
--- /dev/null
+++ b/src/snes/examples/tutorials/output/ex19_draw.out
@@ -0,0 +1,2 @@
+lid velocity = 0.0204082, prandtl # = 1, grashof # = 1
+Number of SNES iterations = 2

Attachment: pgptFl01lWfXz.pgp
Description: PGP signature

Reply via email to