fix: two hard crash fixes #3382
Conversation
paveltomin
requested review from
rrsettgast,
corbett5,
cssherman,
CusiniM,
dkachuma and
joshua-white
as code owners
| @@ -783,12 +783,14 @@ MultiFluidBase::KernelWrapper:: | |||
| real64 totalMolality = 0.0; | |||
| for( integer ic = 0; ic < numComps; ++ic ) | |||
| { | |||
| GEOS_ERROR_IF( m_componentMolarWeight[ic] < LvArray::NumericLimits< real64 >::epsilon, "Zero component molecular weight" ); | |||
There was a problem hiding this comment.
m_componentMolarWeight is user input so we should validate it in postInputInitialization MultiFluidBase.cpp:184 instead of in this kernel function.
There was a problem hiding this comment.
moved the check there, thanks
paveltomin
added
ci: run integrated tests
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## develop #3382 +/- ##
========================================
Coverage 56.79% 56.79%
========================================
Files 1076 1076
Lines 95739 95742 +3
========================================
+ Hits 54375 54378 +3
Misses 41364 41364 ☔ View full report in Codecov by Sentry. |
|
@corbett5, @cssherman, and/or @rrsettgast could you please review? |
| @@ -54,7 +54,7 @@ EventManager::EventManager( string const & name, | |||
| setDescription( "Start simulation time for the global event loop." ); | |||
|
|
|||
| registerWrapper( viewKeyStruct::maxTimeString(), &m_maxTime ). | |||
| setApplyDefaultValue( std::numeric_limits< real64 >::max() ). | |||
| setApplyDefaultValue( std::numeric_limits< integer >::max() * units::YearSeconds ). | |||
There was a problem hiding this comment.
could even use a smaller number tbh.
There was a problem hiding this comment.
Perhaps we shouldn't use a number based on std::numeric_limits< T >::max(). While the actual value of std::numeric_limits< real64 >::max() in seconds is around 5.7e300 years, this seems like a lot given the age of the universe is 13e9 years, and the age of the earth is 4e9 years. Perhaps a nice number would be 1e6 years = 3.1536e+13 seconds?
src/coreComponents/constitutive/fluid/multifluid/MultiFluidBase.hpp
Outdated
Show resolved
Hide resolved
…e.hpp Co-authored-by: Matteo Cusini <[email protected]>
|
@corbett5, @cssherman, and/or @rrsettgast please review, the change is minor and the issue appeared after FTE were enabled |
| real64 const mwInv = 1.0 / m_componentMolarWeight[ic]; | ||
| compMoleFrac[ic] = composition[ic] * mwInv; // this is molality (units of mole/mass) | ||
| dCompMoleFrac_dCompMassFrac[ic][ic] = mwInv; | ||
| totalMolality += compMoleFrac[ic]; | ||
| } | ||
|
|
||
| GEOS_ERROR_IF( totalMolality < LvArray::NumericLimits< real64 >::epsilon, "Zero total molality, all component concentrations are equal to zero." ); |
There was a problem hiding this comment.
This may go bad prior to totalMolality < LvArray::NumericLimits< real64 >::epsilon?? If so perhaps a higher number here? Also can we have a graceful recovery somehow? For example if totalMolality ~= 0 then that means that each compMoleFrac[ic] ~= 0, which means that composition[ic] ~=0? Should we put a limit on the minimum value of the primary variables? We are talking about void at this point?
There was a problem hiding this comment.
all elements composition can not be zero, the issue originates somewhere in the initialization (from user input) where the check is missing
| @@ -54,7 +54,7 @@ EventManager::EventManager( string const & name, | |||
| setDescription( "Start simulation time for the global event loop." ); | |||
|
|
|||
| registerWrapper( viewKeyStruct::maxTimeString(), &m_maxTime ). | |||
| setApplyDefaultValue( std::numeric_limits< real64 >::max() ). | |||
| setApplyDefaultValue( std::numeric_limits< integer >::max() * units::YearSeconds ). | |||
There was a problem hiding this comment.
Perhaps we shouldn't use a number based on std::numeric_limits< T >::max(). While the actual value of std::numeric_limits< real64 >::max() in seconds is around 5.7e300 years, this seems like a lot given the age of the universe is 13e9 years, and the age of the earth is 4e9 years. Perhaps a nice number would be 1e6 years = 3.1536e+13 seconds?
|
@rrsettgast the issue was that that maxTime get's converted into number of years as integer number and then overflows, that's why i put that fix |
one is when all component concentrations are zeros - division by zero
and another is when maxTime is not specified for events - there is overflow when huge double is caster into int