[Debian-med-packaging] Bug#842815: Re-opening: Please support HDF5 1.10

[Bernd Rinn]

Hi Andreas,

the migration of JHDF5 to HDF5 1.10 is ongoing and mostly depend on me
having a block of time I can spend on it. Your analysis of the work that
needs to be done is right from what I can see. The plan is to switch to
using the JNI library from the HDF group wherever possible (it may still
be necessary to have a small JNI library though as some calls appear to
be missing).

I will keep you updated.

Kind regards,

Bernd

On 11/22/2016 03:31 PM, Andreas Tille wrote:
> Hi Bernd,
> 
> I'm hereby forwarding the main parts of the discussion of the bug report
> which might contain hints what to do to make jhdf5 compliant with hdf5
> version 1.10.
> 
> It would be nice if you could give us a hint if you were able to fix this.
> 
> Kind regards
> 
>       Andreas.
> 
> On Tue, Nov 15, 2016 at 12:37:20AM +0100, Gilles Filippini wrote:
>>>>> Thank you for this upload. Have you tested the build against hdf5-1.10
>>>>> in experimental?
>>>>
>>>> Hmmm, sorry, no - I just made sure that the *sid* pbuilder environment
>>>> is instaling hdf5-1.10.
>>>
>>> Ah. The fact that HDF5 1.8.16 builds libhdf5-10 might be confusing. HDF5
>>> 1.10 builds libhdf5-100 and is still in experimental.
>>>
>>>> If this is not sufficient please reopen ...
>>>> and please give some helping hint if the package might not build
>>>> properly.  Otherwise I have no idea what to do - patches are very
>>>> welcome.
>>>
>>> There are at least 3 majors changes brought by HDF5 1.10:
>>> 1- The widely used hid_t datatype is now uint64_t (instead of int)
>>> 2- The java wrapper library is now part of the HDF5 source tree
>>> 3- HDF5Exception is now a checked exception.
>>>
>>> (2) and (3) might be worked out with a new upload of hdf5, to ship the
>>> java wrapper library packages, patched to define HDF5Exception as an
>>> unchecked exception. I'd rather wait after the transition.
>>>
>>> (1) Involves patching libsis-jhdf5-java.
>>
>> I've spent some time trying to figure it out, and I'm afraid this is a
>> huge task I can't afford working on.
>>
>>> The best option would be a new upstream release. Do you know their
>>> schedule wrt HDF5 1.10?
>>
>> If there is a risk that libsis-jhdf5-java won't be ready for stretch,
>> then we could easily avoid the fastqc removal by disabling the fast5
>> file format support. It should be easy enough for anyone to convert
>> fast5 files to fastq using e.g. poretools.
>> What do you think?

-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 3937 bytes
Desc: S/MIME Cryptographic Signature
URL: <http://lists.alioth.debian.org/pipermail/debian-med-packaging/attachments/20161123/554f4b0c/attachment.bin>