xcreate_cross_section failed on certain cases

[wjlei1990]

Hi,

I am using xcreate_cross_section(source code in src/tomography/postprocess_sensitivity_kernels/create_cross_section.F90) to generate some cross section files for my model(which has the same mesh as s362ani). However, this function failed on certain cases.

For example, I am try to make a cross section from (1S, 135E) to (1S, 170E) and the program failed.

 vertical cross-section:

 mesh:  
   processors =          384
   nproc_eta / nproc_xi =            8           8
   nex        =          256
   nspec      =        10496
   nglob      =       699193

 model parameter: vs

   mesh directory: DATABASES_MPI/
  model directory: DATABASES_MPI/
 output directory: vcs-CoralSea-vs

 array size:
   ibool   =    320.31250000000000      MB
   x,y,z   =    170.70141601562500      MB

   model   =    320.31250000000000      MB

 total MPI processes:          384

 location search by : kd-tree search
   uses internal GLL points

 no topography
   depth given with respect to sea-level

 no ellipticity
   radius and latitude of points with respect to a spherical Earth

 New rotating points switched longitude orders:   342.50275141032267        17.497248589677330     
 New rotating points switched longitude orders:   342.50275141032267        17.497248589677330     

 creating target cross-section points... 

   vertical cross-section

   starting depth         :    0.00000000     km
   reaching maximum depth :    2800.00000     km
   depth increment        :    10.000000000000000      km

   start location lat/lon =   -1.00000000       135.000000    
     end location lat/lon =   -1.00000000       170.000000    

   epicentral distance    :    34.9944954     degrees

   great-circle increments:   0.100000001     degrees
   total number of great-circle increments =         3600
   total path length of great-circle =    359.899994     degrees

   total number of points =      1011600
   total size for point arrays =    19.294738769531250      MB

 colat/lon:    91.000000000000000        135.00000000000000      /   91.000000000000000        170.00000000000000     
 normal:    1.0485185417645855        152.50000000000000     
 rotated colat,lon =    90.000000000000000        342.50275141032267      /   90.000000000000000        17.497248589677330

 New rotating points switched longitude orders:   342.50275141032267        17.497248589677330     
 New rotating points switched longitude orders:   342.50275141032267        17.497248589677330     
 New rotating points switched longitude orders:   342.50275141032267        17.497248589677330     
 New rotating points switched longitude orders:   342.50275141032267        17.497248589677330     
 New rotating points switched longitude orders:   342.50275141032267        17.497248589677330     
 New rotating points switched longitude orders:   342.50275141032267        17.497248589677330     
 New rotating points switched longitude orders:   342.50275141032267        17.497248589677330     
 New rotating points switched longitude orders:   342.50275141032267        17.497248589677330     
...

Here is some debug print I got from the output file:

 colat/lon:    91.000000000000000        135.00000000000000      /   91.000000000000000        170.00000000000000     
 normal:    1.0485185417645855        152.50000000000000     
 rotated colat,lon =    90.000000000000000        342.50275141032267      /   90.000000000000000        17.497248589677330

(Error begins from here)
New rotating points switched longitude orders:   342.50275141032267        17.497248589677330  

I think in the code, when it rotate the normal plane and change the (theta/phi), there are some errors popped at the line here.

---

PS1: Actually, I tried using different latitudes, from 45S to 10N(longtitudes are the same, from 135E to 170E), all the north latitudes works but all the south do NOT work.

I also tried to flip the start and end points since the program is complaing about the longitude. But there are still errors...

PS2: I am using a relatively old version of SPECFEM3D GLOBE. However, I compared my version with the most recent version, it seems nothing has been changed to the create_cross_section.F90 file.