Appveyor CI: Fixing Regression Tests

[bemcdonnell]

Lateral Inflow Bug:

A few months ago we discovered what we thought was a redundant line in the lateral inflow (see #61). Turns out that this change resulted in us failing two regression tests. Lateral inflows were off by 0.001. This has been reverted back in the next PR coming in today.

System Inflow Totalizers:

This change will actually be somewhat disruptive. Before, when SWMM had a negative inflow through an outfall, it was not counted as a positive inflow to the system. This change is the following:

--- a/src/routing.c
+++ b/src/routing.c
@@ -804,7 +804,7 @@ void removeOutflows(double tStep)
         // --- update mass balance with flow and mass leaving the system
         //     through outfalls and flooded interior nodes
         q = node_getSystemOutflow(i, &isFlooded);
+        if ( q > 0.0 ) // <-----------------------------------------------
-        if ( q != 0.0 )
         {
             massbal_addOutflowFlow(q, isFlooded);
             for ( p = 0; p < Nobjects[POLLUT]; p++ )
@@ -813,7 +813,7 @@ void removeOutflows(double tStep)
                 massbal_addOutflowQual(p, w, isFlooded);
             }
         }
+        else massbal_addInflowFlow(EXTERNAL_INFLOW, -q);// <-------------------

         // --- update mass balance with mass leaving system through negative
         //     lateral inflows (lateral flow was previously accounted for)

This change requires us update the benchmarks for user3

[dickinsonre]

Good catch @bemcdonnell

[michaeltryby]

Looks like test "User 3" fails due to this change in the flow totalizer. So I guess we need to update the benchmark. We need to develop a process for doing this so we can keep track of how the benchmark changes over time and be able to reproduce benchmark results. Any thoughts?

[dickinsonre]

How about just a note in the note section of the input file showing the old flow equalizer total and explaining the change in the code. Or, how about a a more multi objective test that compares all of the flow totals in the links of the two model output files.