05.04.2016 4:27, Elena Pourmal пишет:
3) Symbol H5F_LIBVER_18 definition is missing in 1.10.0. I used to pass
the value to H5Pset_libver_bounds to ensure files are readable by HDF5
1.8+. How do I proceed with HDF 1.10 to achieve the same effect (i.e.
readability by HDF5 1.8+)?
Unfortunately, HDF5 1.10.0 doesn’t have a flag to get 1.8 behavior. We
ran out of time and the feature didn’t make into the release.
Focus of HDF5 1.10.0 was on the new SWMR and VDS features that are not
compatible with 1.8. In 1.10.1 we will finish integration of the promised
HPC improvements and will add control for library versioning.
In 1.10.0, if one specifies the |H5F_LIBVER_LATEST flag to the
H5Pset_libver_bounds function, the file will be created according to |
>the latest version of the HDF5 file format and may not be readable by
HDF5 1.8.*. We will create a document that describes |
>which objects are compatible with 1.8 when the |H5F_LIBVER_LATEST flag
It would suffice to introduce H5F_LIBVER_18_110 equals to the previous value of H5F_LIBVER_18. It is understandable that using new 1.10 features (e.g. SWMR) may introduce backward incompatibility for 1.8 readers.
I'm a bit cautious to use H5F_LIBVER_LATEST everywhere because my understanding is its value may change in the future. The header file says
H5F_LIBVER_LATEST /* Use the latest possible format available for storing objects*/
but usually I'd like to ensure 1.8 compatibility even if new format is introduced in 1.12 and H5F_LIBVER_LATEST value is increased.
I can wait for the next update of 1.10
4) The symbols H5F_ACC_* are now #defined using H5OPEN treat, meaning
that H5open call is needed before access to constant values. However,
the values are defined as plain constants -- why use H5OPEN?
Thank you for bringing this to our attention. We discussed it today and
agreed that using H5CHECK and H5OPEN with the constants is overkill and
We will be looking into the issue.
Thank you for your support,
This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.