I've been trying to run simulations that consist of a supersonic
freestream flowing over a triangular obstacle. The problem I'm facing
most of the time is to startup the simulation, especially when the
flow turns around sharp corners. To provide an example, I uploaded a
script called "Wedge", which generates a NaN at the 15th phase. Does
the simulation crash because of a zero or negative pressure? Is there
another scheme that is better suited for this type of problem?

Jimmy,

You're using a linearized Riemann solver that cannot preserve
positive pressures in the strong expansion you get at sharp
corners. The script:

   amrcp bcg/axs.3

will generate 108 alternative solvers for the ReactiveEulerEquations.
Most of them will suffer from the same numerical weakness
but those that use the lax-friedrichs flux should prove
more roubust. For example, try using:

      plugin amr_sol
      ReactiveEulerEquations {
         space  = two-dimensional
      }
      BasicCodeGenerator {  
         scheme = contrib'axs                                            
         solver = jimmy                                              
         flux   = axs/lax-friedrichs                                      
         interp = axs/WENO-A
      }    

and if that doesn't work try using for interp axs/minmod .

James