Readback from "flattened" file

[guj]

Describe the bug
When running bpls, it did not return the correct dimension of variables.

To Reproduce
The file is on frontier: /lustre/orion/csc143/scratch/junmin/flattenTest/openpmd_000000.bp/

Expected behavior
bpls openpmd_000000.bp/ /data/0/particles/beam/id
shows
uint64_t /data/0/particles/beam/id {666}

but it should show
uint64_t /data/0/particles/beam/id {1000}

If you use -D option, you will see there are 1000 elements total.
uint64_t /data/0/particles/beam/id {666}
step 0:
block 0: [ 0:665]
block 5: [666:999]

Desktop (please complete the following information):

• Frontier
• Build (checked out Jan 28th)

[eisenhauer]

Can you permit access to that file?

[email protected] ~]$ ls -ld /lustre/orion/csc143/scratch/junmin/flattenTest/openpmd_000000.bp/
ls: cannot access '/lustre/orion/csc143/scratch/junmin/flattenTest/openpmd_000000.bp/': Permission denied
[[email protected] ~]$ ls -ld /lustre/orion/csc143/scratch/junmin/flattenTest/openpmd_000000.bp
ls: cannot access '/lustre/orion/csc143/scratch/junmin/flattenTest/openpmd_000000.bp': Permission denied
[[email protected] ~]$ ls -ld /lustre/orion/csc143/scratch/junmin/flattenTest/
ls: cannot access '/lustre/orion/csc143/scratch/junmin/flattenTest/': Permission denied
[[email protected] ~]$ ls -ld /lustre/orion/csc143/scratch/junmin/
drwx--S--- 8 junmin junmin 99328 Feb  5 22:23 /lustre/orion/csc143/scratch/junmin/
[[email protected] ~]$ 

[guj]

They all have drwxr-sr-x
ls -ld /lustre/orion/csc143/scratch/junmin/flattenTest/
drwxr-sr-x 3 junmin junmin 99328 Feb 5 22:25 /lustre/orion/csc143/scratch/junmin/flattenTest/

I also copied to /lustre/orion/csc143/world-shared/junmin/flattenTest
Hope you can access this one

[eisenhauer]

Huh. I see no change:

[[email protected] ~]$ ls -ld /lustre/orion/csc143/scratch/junmin/
drwx--S--- 8 junmin junmin 99328 Feb  5 22:23 /lustre/orion/csc143/scratch/junmin/
[[email protected] ~]$ 

But I can access the copied version, so I'll take a look.

[eisenhauer]

@pnorbert There are several calls that likely require special handling with flatten steps, but we may have to think about what semantics we want. For global variables, BP5 VarShape() currently returns the shape of the current timestep (selected step for random access mode, current for streaming), it starts at rank 0 looking for some rank that wrote the variable and returns the first size it finds (because all writing ranks should write the same shape). Elsewhere, we have taken pains to ensure that if you change the shape of a variable during a step, we take the last shape it had (I.E. we capture the shape during EndStep().). How should this work with flatten steps? Take the shape from the last timestep upon which a variable was written?

What about the shape of joined arrays? Min/max?