pmb->loc broken?

[pgrete]

@lroberts36 I'm trying to calculate the effective offsets for each block for writing outputs.

Within a block like (on a mesh with 64x96 and blocks of 32x32)

    for (auto &pmb : pm->block_list) {
...
              std::cout << "Block " << pmb->gid << " writes chunk of [" << chunk_extent[0]
                        << " " << chunk_extent[1] << " "
                        << "] with offset [" << chunk_offset[0] << " " << chunk_offset[1]
                        << "] and logical locs [" << pmb->loc.lx2() << " "
                        << pmb->loc.lx1() << "]\n";
...

I get the following output:

Block 0 writes chunk of [32 32 ] with offset [0 0] and logical locs [0 0]
Block 1 writes chunk of [32 32 ] with offset [0 0] and logical locs [0 0]
Block 2 writes chunk of [32 32 ] with offset [0 0] and logical locs [0 0]
Block 3 writes chunk of [32 32 ] with offset [0 0] and logical locs [0 0]
Block 4 writes chunk of [32 32 ] with offset [0 0] and logical locs [0 0]
Block 5 writes chunk of [32 32 ] with offset [0 0] and logical locs [0 0]

Am I missing sth? IIRC previously `loc.lx#` returned the effective logical location in a global mesh at a given level.
This is with all changes from current `develop`.

[lroberts36]

No, this is the expected behavior as the grid you are using results in a forest of six trees. Each block has a logical location relative to the tree it is in, but 'LogicalLocation' now also contains the index of the tree in which it is contained accessible via 'LogicalLocation::tree()'. Getting on a plane now (and writing this from my phone), I will look at how to extract the offsets you want tomorrow morning. They are probably already contained in the neighbor block information.

[lroberts36]

For now, you could use Forest::GetLegacyTreeLocation using the locations stored in the MeshBlocks to get the old global logical location, but this will not work with general forests.

[pgrete]

Perfect, that did the job! Thanks for the quick reply.
Now I also understand why this showed up in downstream testing (as I previously used a mesh that was part of a single tree in Parthenon).
One more reminder to add downstream testing of more complex mesh structures.