There are a handful of tests which are failing because they fall slightly below the error threshold:
Ideal Gas:
- Conserved to Primitive Test - Screwball Schwarzschild, Strong Flow and Fieldless
- Conserved to Primitive Test - Screwball Schwarzschild, Strong Flow and Field
CompOSE Tabulated EOS:
- Conserved to Primitive Test - Fieldless, Strong Flow and Gravity
- Conserved to Primitive Test - Strong Flow, Field, and Gravity
These all fail at the $10^{-9}$ or $10^{-10}$ level when the error tolerance is $10^{-10}$. The tests themselves are ridiculous, with Lorentz factors in excess of 2000 for the ideal gas and 100 for the tabulated EOS, respectively. It might be worth doing some error analysis on the algorithm to find out if there are any terms which could be rewritten to be more numerically stable. Otherwise, it's perfectly reasonable to change the tests.
There are a few tests checking for unphysical situations which fail:
- Unphysical State Test - Negative Tau
- Unphysical State Test - Negative Tau, DEC Violation
- Unphysical State Test - Excess Momentum, DEC Violation
The root solver fails to bracket these solutions correctly. We need to think carefully about whether or not we want PrimitiveSolver to handle unphysical states or if that should be pushed back on the user with an error code instead.
I would like to add an automated CI pipeline to the repo for ease of use, but these tests need to pass first.
There are a handful of tests which are failing because they fall slightly below the error threshold:
Ideal Gas:
CompOSE Tabulated EOS:
These all fail at the$10^{-9}$ or $10^{-10}$ level when the error tolerance is $10^{-10}$ . The tests themselves are ridiculous, with Lorentz factors in excess of 2000 for the ideal gas and 100 for the tabulated EOS, respectively. It might be worth doing some error analysis on the algorithm to find out if there are any terms which could be rewritten to be more numerically stable. Otherwise, it's perfectly reasonable to change the tests.
There are a few tests checking for unphysical situations which fail:
The root solver fails to bracket these solutions correctly. We need to think carefully about whether or not we want PrimitiveSolver to handle unphysical states or if that should be pushed back on the user with an error code instead.
I would like to add an automated CI pipeline to the repo for ease of use, but these tests need to pass first.