h5dump fails to parse HDF5 path names containing certain ASCII characters

[gheber]

h5dump --enable-error-stack --dataset="/ic_antennas/antenna[]&name" ic_antennas.h5

fails with the following error stack:

HDF5-DIAG: Error detected in HDF5 (1.15.0) thread 0:
  #000: H5D.c line 402 in H5Dopen2(): unable to synchronously open dataset
    major: Dataset
    minor: Can't open object
  #001: H5D.c line 363 in H5D__open_api_common(): unable to open dataset
    major: Dataset
    minor: Can't open object
  #002: H5VLcallback.c line 1980 in H5VL_dataset_open(): dataset open failed
    major: Virtual Object Layer
    minor: Can't open object
  #003: H5VLcallback.c line 1947 in H5VL__dataset_open(): dataset open failed
    major: Virtual Object Layer
    minor: Can't open object
  #004: H5VLnative_dataset.c line 319 in H5VL__native_dataset_open(): unable to open dataset
    major: Dataset
    minor: Can't open object
  #005: H5Dint.c line 1423 in H5D__open_name(): not found
    major: Dataset
    minor: Object not found
  #006: H5Gloc.c line 438 in H5G_loc_find(): can't find object
    major: Symbol table
    minor: Object not found
  #007: H5Gtraverse.c line 836 in H5G_traverse(): internal path traversal failed
    major: Symbol table
    minor: Object not found
  #008: H5Gtraverse.c line 613 in H5G__traverse_real(): traversal operator failed
    major: Symbol table
    minor: Callback failed
  #009: H5Gloc.c line 395 in H5G__loc_find_cb(): object 'antenna' doesn't exist
    major: Symbol table
    minor: Object not found
HDF5-DIAG: Error detected in HDF5 (1.15.0) thread 0:
  #000: H5L.c line 1337 in H5Lget_info2(): unable to get link info
    major: Links
    minor: Can't get value
  #001: H5VLcallback.c line 5414 in H5VL_link_get(): link get failed
    major: Virtual Object Layer
    minor: Can't get value
  #002: H5VLcallback.c line 5382 in H5VL__link_get(): link get failed
    major: Virtual Object Layer
    minor: Can't get value
  #003: H5VLnative_link.c line 263 in H5VL__native_link_get(): unable to get link info
    major: Links
    minor: Object not found
  #004: H5Lint.c line 1905 in H5L_get_info(): name doesn't exist
    major: Links
    minor: Object already exists
  #005: H5Gtraverse.c line 836 in H5G_traverse(): internal path traversal failed
    major: Symbol table
    minor: Object not found
  #006: H5Gtraverse.c line 613 in H5G__traverse_real(): traversal operator failed
    major: Symbol table
    minor: Callback failed
  #007: H5Lint.c line 1867 in H5L__get_info_cb(): name doesn't exist
    major: Links
    minor: Object not found
HDF5 "ic_antennas.h5" {h5dump error: unable to get link info from "/ic_antennas/antenna"

}

Notice the []& in the link name, which is legitimate. This is not a shell issue because the correct string is passed to the C program:

(gdb) p *argv@argc
$1 = {0x7fffffffe608 "/home/gerd/.local/bin/h5dump", 0x7fffffffe625 "-d", 
  0x7fffffffe628 "/ic_antennas/antenna[]&name", 0x7fffffffe644 "ic_antennas.h5"}

That means that the HDF5 path name is not correctly parsed.

Interestingly, other tools don't have this issue, e.g.,

gerd@deborah:~/work$ h5ls ic_antennas.h5/"ic_antennas/antenna[]&name"
antenna[]&name           Dataset {3/Inf}

Don't tools share the same pathname parser?

Ludovic Fleury of CEA reported this issue.

[gheber]

The error appears to be there in many (all?) releases. I've confirmed it for 1.8.12, 1.12.0, and develop.

[byrnHDF]

However "[]" is part of the h5dump argument string for subsetting.

[byrnHDF]

Did it work if the dataset name was in quotes like for the h5ls example?

[gheber]

Did it work if the dataset name was in quotes like for the h5ls example?

No. (The quotes got dropped when I copied it.)

[gheber]

However "[]" is part of the h5dump argument string for subsetting.

What's the toggle to suppress that behavior? It's obviously wrong to "trespass" on the user namespace in that way.

[byrnHDF]

Unfortunarely, someone decided that using "[]" as a shortcut for subsetting did not consider all the ramifications.
h5dump is the only tool that uses that shortcut. easiest would be to deprecate that notation or someone will need to fix the H5_get_option function in the library to honor the quotes around the dataset name.

[gheber]

👍 (deprecation)

What would the point of the -[s,S,c,k] options be otherwise?

[byrnHDF]

Actually the problem is in the parse_subset_params functions used by h5dump and h5diff. (Looks like h5diff might have that problem?) So it is even simpler. (Plus that function should probably in the the tools library proper)

[byrnHDF]

Just need to add some checking for quotes - maybe.

[byrnHDF]

h5dump has this option: --no-compact-subset but h5diff does not.