Locally and on travis-ci, integration tests (i.e., TestPrj4) were run and compared against reference database dbOutput_TestPrj4_v4.1.0.sqlite3 successfully (i.e., no differences): for instance,
PR #370 (v4.1.0) on Oct 18, 2019
commit 4138dbf99f8f108b77fa2270b61a21f34ca4e4c2 on Dec 19, 2019
passed checks successfully.
However, they started to disagree with the reference database locally and on travis-ci in Jan 2020. I pulled the commits of rSFSW2, rSOILWAT2, and rSW2utils (if needed) to recreate conditions at several points in time until May 2019. These runs no longer reproduce output as in the reference database. I tested on different machines, under different operating systems, and with different compiler versions.
The table aggregation_overall_mean has 109 differences and aggregation_overall_sd has 97 differences. The daily aggregation tables have 0 differences. I checked intermediate RData files (direct rSOILWAT2 output) and cannot produce any deviations.
Example of differences
$aggregation_overall_mean
[1] "Component “WaterExtractionSpell_MeanContinuousDuration_L5_days_mean”: Mean absolute difference: 0.2"
[2] "Component “WaterExtractionSpell_AnnualSummedExtraction_L4_mm_mean”: Mean absolute difference: 0.01"
[3] "Component “WaterExtractionSpell_AnnualSummedExtraction_L5_mm_mean”: Mean absolute difference: 0.01"
[4] "Component “DeepDrainage_DailyMax_doy_mean”: Mean absolute difference: 0.01695991"
[5] "Component “SWP_bottomLayers_DailyMin_doy_mean”: Mean absolute difference: 0.002545719"
[6] "Component “RelRecharge_bottomLayers_DailyMin_doy_mean”: Mean absolute difference: 0.002555868"
[7] "Component “NRCS_COND3_MCS_AnyDry_days_mean”: Mean absolute difference: 0.05882353"
[8] "Component “NRCS_COND7_MCS_AnyMoist_GT180Days_days_mean”: Mean absolute difference: 0.0625"
[9] "Component “NRCS_COND8_MCS_AnyWetConsec_days_mean”: Mean absolute difference: 0.0625"
[10] "Component “NRCS_COND10_MCS_AllDry_days_mean”: Mean absolute difference: 0.0625"
[11] "Component “WetDegreeDays_SWPcrit1500kPa_topLayers_Cdays_mean”: Mean absolute difference: 0.2322581"
[12] "Component “WetDegreeDays_SWPcrit1500kPa_bottomLayers_Cdays_mean”: Mean absolute difference: 0.3409677"
[13] "Component “WetDegreeDays_SWPcrit1500kPa_anyLayer_Cdays_mean”: Mean absolute difference: 0.3409677"
[14] "Component “WetDegreeDays_SWPcrit3000kPa_topLayers_Cdays_mean”: Mean absolute difference: 0.07016129"
[15] "Component “WetDegreeDays_SWPcrit3000kPa_bottomLayers_Cdays_mean”: Mean absolute difference: 0.1045161"
[16] "Component “WetDegreeDays_SWPcrit3000kPa_anyLayer_Cdays_mean”: Mean absolute difference: 0.07016129"
[17] "Component “WetDegreeDays_SWPcrit3900kPa_topLayers_Cdays_mean”: Mean absolute difference: 0.4620968"
[18] "Component “WetDegreeDays_SWPcrit3900kPa_bottomLayers_Cdays_mean”: Mean absolute difference: 0.03693548"
[19] "Component “ThermalDrySoilPeriods_SWPcrit1500kPa_NSadj_topLayers_Start_LongestContinuous_days_mean”: Mean absolute difference: 1.032376"
Locally and on travis-ci, integration tests (i.e.,
TestPrj4
) were run and compared against reference databasedbOutput_TestPrj4_v4.1.0.sqlite3
successfully (i.e., no differences): for instance,However, they started to disagree with the reference database locally and on travis-ci in Jan 2020. I pulled the commits of rSFSW2, rSOILWAT2, and rSW2utils (if needed) to recreate conditions at several points in time until May 2019. These runs no longer reproduce output as in the reference database. I tested on different machines, under different operating systems, and with different compiler versions.
The table
aggregation_overall_mean
has 109 differences andaggregation_overall_sd
has 97 differences. The daily aggregation tables have 0 differences. I checked intermediateRData
files (directrSOILWAT2
output) and cannot produce any deviations.Example of differences