Re: [netcdfgroup] Superblock version 0

Dennis et al,

Now on a more positive note, my early testing finds that an older utility
program reads those new superblock 0 files just fine.  So far, I checked
only the three ref_hdf5_compat files from the new distro, and one test file
that I created locally.

This demonstrates the backward compatibility that we have been seeking.

--Dave


On Fri, Jun 3, 2016 at 5:26 PM, Dave Allured - NOAA Affiliate <
dave.allured@xxxxxxxx> wrote:

> Dennis,
>
> I use a simple hex dump for expediency.  This misses files with a nonzero
> user block, but I think those are pretty rare in the climate community.
>
> For example, here is one of the test files in the latest Unidata distro.
> But I get the same result with a locally created file.
>
> mac56:~/netcdf/netcdf-c-4.4.1-rc2/nc_test4 55> od -tx1 -N9
> ref_hdf5_compat2.nc
> 0000000    89  48  44  46  0d  0a  1a  0a  00
> 0000011
>
> The superblock version number is the 9th (last) byte.
>
> This number matches what is reported by ncdump -hs in the new release.  I
> have no reason to doubt the result from H5Pget_version.
>
> --Dave
>
>
> On Fri, Jun 3, 2016 at 4:59 PM, dmh@xxxxxxxx <dmh@xxxxxxxx> wrote:
>
>> HDF says the (superblock format) differences are:
>> Version 0 is the default format
>> Version 1 is the same as version 0 but with the âIndexed Storage
>>        Internal Node Kâ field for storing non-default B-tree âKâ value.
>> Version 2 has some fields eliminated and compressed from superblock
>>         format versions 0 and 1. It has added checksum support and
>>     superblock extension to store additional superblock metadata.
>> Version 3 is the same as version 2 except that the field âFile
>>      Consistency Flagsâ is used for file locking. This format version
>>       will enable support for the latest version.
>> It is possible that the H5Pget_version indicates not the true number,
>> but rather the lowest externally detectable version number or some such.
>> Do you have a pure hdf5 1.10 file to see that superblock number is
>> reported by ncdump -hs?
>> Or did you do a hex dump to see what is actually in the superblock?
>> =Dennis Heimbigner
>>  Unidata
>>
>>
>> On 6/3/2016 4:11 PM, Dave Allured - NOAA Affiliate wrote:
>>
>>> Netcdf group,
>>>
>>> My team is starting to test netcdf-4.4.1-rc2 with the latest HDF5
>>> library, currently 1.10.0-patch1.  I see that netcdf-4 files are now
>>> being generated with superblock version 0 for the first time in the
>>> history of netcdf-4.  Presumably this is a result of the recent format
>>> compatibility fix in 4.4.1.
>>>
>>> All the older netcdf-4 files that I sampled, back to the original
>>> netcdf-4.0 (2008), were created as superblock version 2.
>>>
>>> Are there any indications whether superblock 0 will be robust and
>>> interoperable, going into the future?
>>>
>>> --Dave
>>
>>