Three accepted options do nothing or abort later: incflo.test_tracer_conservation (consumer commented out; 8 decks set it), incflo.ic_p (sets m_p000, which nothing reads), amr.plt_divu / "divu" (documented, aborts "xxxxx TODO" at the first plotfile) #264

Severity: low · Category: correctness
Locations: src/setup/init.cpp:70, src/incflo_advance.cpp:89-94, src/setup/init.cpp:146, src/setup/set_background_pressure.cpp:7, src/incflo.H:513, src/setup/init.cpp:383-386, src/setup/init.cpp:426-429, src/utilities/io.cpp:675-678, Docs/sphinx_documentation/source/InputsPlotFiles.rst:69
Based on commit 46de3367 (line numbers refer to that tree).

The defect

Three inputs are parsed by ReadParameters/ReadIOParameters and then have no live consumer:

1. incflo.test_tracer_conservation (init.cpp:70 -> m_test_tracer_conservation). Its only use is the
   commented-out block at the end of Advance (incflo_advance.cpp:89-94):

   // // This sums over all levels
   // if (m_test_tracer_conservation) {
   //     Real sum = volumeWeightedSum(get_tracer_new_const(),0,geom,ref_ratio);

   Eight shipped decks set it to true (test_{2d,3d}/benchmark.tracer_advection,
   benchmark.tracer_adv_diff, benchmark.tracer_adv_diff_cn, benchmark.bds_tracer_advection; the three
   test_3d/test_redistribution/benchmark.periodic_box_{x,y,z} decks set it to false) -- they are
   the tracer-advection regression tests, i.e. exactly the runs the diagnostic was written for -- and get no
   output. amrex::volumeWeightedSum(Vector<MultiFab const*>, int, Vector<Geometry> const&, Vector<IntVect> const&, bool)
   exists in the pinned AMReX (Src/Base/AMReX_MultiFabUtil.H:552, EB-volume weighted), so the block was
   disabled for no remaining reason.
2. incflo.ic_p (init.cpp:146 -> m_ic_p). set_background_pressure copies it into m_p000
   (set_background_pressure.cpp:7, "background pressure at node (0,0,0)", incflo.H:513) and no code reads
   m_p000 (grep -rn m_p000 src finds only those two lines). 41 shipped decks set incflo.ic_p (all to 0.0,
   so no shipped result changes); a user setting a non-zero initial pressure gets p = 0.
3. amr.plt_divu / the plot variable divu. ReadIOParameters accepts plt_divu (init.cpp:383-386) and
   amr.plotVariables = ... divu (426-429) and WritePlotVariables then does
   amrex::Abort("plt_divu: xxxxx TODO") (io.cpp:675-678). The option is documented as working
   (InputsPlotFiles.rst:69, "Save velocity divergence to plot file"). Because plotfiles are written after
   the whole initialisation (EB generation, initial projection, initial iterations), the abort comes at the
   end of InitData, not at read time; issue 001 (steady_state) and PR fix issue 177 178 179 #212's KE_int fix are the same
   pattern.

Why it matters

Options that are accepted and silently do nothing (1, 2) hide the fact that the tracer-conservation check
the regression decks request is not being made, and that an initial pressure cannot be prescribed; option 3
costs a full initialisation before failing with a message that reads like an internal note. No wrong numbers
in shipped runs (all ic_p values are 0).

How to reach it

Any build. test_2d/benchmark.tracer_advection (sets incflo.test_tracer_conservation = true): no
"Sum tracer" line is ever printed. test_no_eb_2d/benchmark.taylor_green_vortices with
amr.plot_per_exact=-1 amr.plot_int=1 amr.plt_divu=1: amrex::Abort::0::plt_divu: xxxxx TODO after
InitialIterations, at the step-0 plotfile (confirmed with a DEBUG 2D build). Any deck with
incflo.ic_p = 5: p in plt00000 is 0.

Suggested fix

Re-enable the conservation print with the AMReX helper, and refuse divu when the plot list is read (the
m_p000 member should either be wired into the pressure initialisation or removed together with ic_p;
that is a design decision, so no diff for item 2). Diff checked with git apply --check
(audit/notes/T9-scratch/034.diff) and compiled/run with a DEBUG 2D no-EB build: bouss_bubble_god with
incflo.test_tracer_conservation=1 prints Sum tracer volume wgt2 = 0.2 0.004697265625 after each step;
amr.plotVariables="velx divu" aborts in ReadIOParameters.

--- a/src/incflo_advance.cpp
+++ b/src/incflo_advance.cpp
@@ -86,12 +86,12 @@
     particleData.Redistribute();
 #endif
 
-    // // This sums over all levels
-    // if (m_test_tracer_conservation) {
-    //     Real sum = volumeWeightedSum(get_tracer_new_const(),0,geom,ref_ratio);
-    //     amrex::Print() << "Sum tracer volume wgt2 = " << m_cur_time+m_dt << " " <<
-    //                        sum << "\n";
-    // }
+    // This sums over all levels (and weights by the EB volume fraction)
+    if (m_test_tracer_conservation && m_advect_tracer) {
+        Real sum = amrex::volumeWeightedSum(get_tracer_new_const(),0,geom,ref_ratio);
+        amrex::Print() << "Sum tracer volume wgt2 = " << m_cur_time+m_dt << " " <<
+                           sum << "\n";
+    }
 
     // Stop timing current time step
     Real end_step = static_cast<Real>(ParallelDescriptor::second()) - strt_step;
--- a/src/setup/init.cpp
+++ b/src/setup/init.cpp
@@ -428,6 +428,14 @@
         pp.queryarr("plotVariables", m_plotVars);
     }
 
+    // "divu" is accepted by the parser but WritePlotVariables only aborts on
+    // it; refuse it here rather than at the first plotfile.
+    for (auto const& v : m_plotVars) {
+        if (v == "divu") {
+            amrex::Abort("plotfile variable 'divu' (amr.plt_divu) is not implemented");
+        }
+    }
+
     // Small plot options
     pp.query("smallplot_file"      , m_smallplot_file);
     pp.query("smallplot_int"       , m_smallplot_int);