Known problems with FLASH Release 3.1.1
Several problems have been identified in FLASH Release 3.1.1. In the hopes of saving much user frustration,
these bugs are identified below. FLASH Release 3.2, released July 2, 2009, addresses each
of these problems. We apologize profusely for any difficulties you may have encountered.
The top-most affected unit is shown for each problem. If you don't use that sub-branch of the code tree, then
the problem will not affect you.
Diagnostic messages may be lost in the event of an abort.
Grid_outsideBoundBox API function does not correctly handle the
situation when a particle is exactly on a block boundary.
It is possible for neighboring PARAMESH leaf blocks to be separated by
more than one (1) level of refinement. This bug breaks a fundamental assumption
made by the PARAMESH package.
Unnecessary print statements occur when using MHD.
PAUSE statements should be replaced by Driver_abortFlash.
Problem with Balsara's prologation and guardcell filling routines.
Possible danger: MPI error handler is set to ignore all errors.
PFFT does not generate a new FLASH to PFFT mapping when the grid
- Grid/GridParticles/GridParticlesMove, GridMain/paramesh/paramesh4, source/Grid/localAPI
Particles may be moved onto the wrong block when there are periodic
boundary conditions and the destination wrap-around neighbor is at a
different refinement level.
A race condition may occur between data messages generated from individual compute nodes and metadata
The Fortran code that appears when DEBUG_GRIDPARTICLES is defined does
- physics/Gravity/GravityMain/Poisson, physics/Hydro/HydroMain/split/PPM/PPMKernel
Pre-processor definitions GPOT_VAR and GPOL_VAR are inconsistent
when using constant gravity with active particles.
Does not work when combined with gravity.
Incorrect use of GPOL and GPOT for gravity extrapolation to n+1/2 step.
Possible danger: HDF5 Error handling switched off.
Preprocessor definition is interpreted incorrectly (This is a Lahey compiler-specific
Interface INTENT statement is inconsistent with the implementation. Unused memspace
causes HDF5 error messages. Lack of error code checking. Logic for
calling Particles_updateAttribute for writing a checkpoint file is
wrong if it was just called for writing a particles file.
Function argument mismatch.
Error with restarts when using the Cosmology unit: io_nextParticleZ
variable is not initialized correctly.
There are dangling hdf5 handles which may cause memory leakage.
Actual computation of Spitzer viscosity is wrong.
Logic for multiple particle types needs tweaking.
Internal energy not initialized properly. Inconsistent use of lb vs
gr_blkList(lb). LEAF should be ACTIVE_BLKS to be consistent with
logic. delta_max_par, delta_max not initialized correctly and isFirst
is an integer not a logical.