On 10 Aug 2012, at 13:38, Ian Hinder wrote:
Hi,
Three months ago, I corrected a bug in the McLachlan boundary conditions. See https://trac.einsteintoolkit.org/ticket/848 for more details. All bug fixes tend to be discussed on trac.einsteintoolkit.org, but we should probably announce bugs like this which people might miss otherwise on the mailing list as well. Someone has just found that an unexplained numerical blowup was fixed by applying the above patch. The bug affects the Maxwell release, but is fixed in Lovelace. Anyone still using Maxwell and McLachlan radiative boundary conditions with conformalMethod = 1 should probably apply that fix or update to Lovelace. Similarly, ML_ADMConstraints is incorrect in Maxwell, and fixed in Lovelace.
I just had another report of someone who was bitten by this bug, so I want to reiterate:
If you are using ML_BSSN from the Maxwell release (ET_2011_10) or earlier with conformalMethod = 1 (which is quite likely), you should either upgrade to Lovelace (ET_2012_05) or apply the patches (http://git.carpetcode.org/McLachlan.git/commit/f5fda658d49efce252918ad5bfbc8... and http://git.carpetcode.org/McLachlan.git/commit/f25d614db972c4c938a4cc658bf26...) to the Lovelace version.
The effect of this bug is that the outer boundary condition for the 3-metric conformal factor (i.e. its determinant) is significantly wrong and can affect your simulation in unexpected ways.