H5Part development and discussion

["Gsell Achim (PSI)" <achim.gsell AT psi.ch>]

Hi Gavin

> On 21 Jun 2017, at 18:50, Achtemeier, Gavin <gachtem AT sandia.gov> wrote:
> 
> 1.99.13 accessed from https://amas.psi.ch/H5hut/wiki/H5hutDownload
> 
> I came across the page following from LBNL's H5Part site.

We are migrating from Trac on 'amas.psich'  to our new Gitlab system 
'gitlab.psi.ch'. Unfortunately the information on the old Trac are outdated 
and on Gitlab not yet available. The version 1.99.13 is pretty old and the 
Fortran API is known to be broken.

You can download a current version with 

wget 
https://gitlab.psi.ch/H5hut/src/repository/archive.tar.bz2?ref=H5hut-2.0.0rc3 
-O H5hut-2.0.0rc3.tar.bz2

Please note:
- There are some API changes, mainly in the functions to open a file.
- Most examples have been rewritten. There are working examples for C and 
Fortran.

Best

Achim

> 
> Gavin
> 
> -----Original Message-----
> From: Gsell Achim (PSI) [mailto:achim.gsell AT psi.ch] 
> Sent: Wednesday, June 21, 2017 10:46 AM
> To: Achtemeier, Gavin <gachtem AT sandia.gov>
> Subject: Re: [EXTERNAL] Re: [H5part] Fortran API fixed?
> 
> 
>> On 21 Jun 2017, at 18:33, Achtemeier, Gavin <gachtem AT sandia.gov> wrote:
>> 
>> Achim,
>> 
>> Thanks for the response, the problem was all mine in that I wasn't linking 
>> to the HDF5 libraries.
>> 
>> Now that I'm using the library I am getting some unexpected behavior. It 
>> may be a bug, though it could also be my install and build.
>> 
>> In Fortran, when creating a new file the return is zero. Documentation 
>> contradicts itself with the open function claiming it returns an error 
>> code with success on 0 and the other functions stating that their first 
>> parameter is the file handle returned by the open function. I assume it 
>> should be returning a file handle.
>> 
>> When I pass this value to any other function it prints something similar 
>> to:
>> [proc 0] E: h5_writefileattrib_string_: Cannot create attribute 'another 
>> attrib 1' for '[error getting object name]'.
>> 
>> and if you write to the HDF5 file enough the program segfaults.
>> 
>> Files to (hopefully) give a minimal reproducible case are attached. System 
>> is a Red Hat Enterprise 6.8
>> 
>> Additionally the Fortran examples contain depreciated function names which 
>> will not allow it to compile as-is.
> 
> Which version of H5hu5 are you using? Current master from gitlab.psi.ch?
> 
> Best
> 
> Achim
> 
> 
>> 
>> Thanks,
>> Gavin
>> 
>> -----Original Message-----
>> From: Gsell Achim (PSI) [mailto:achim.gsell AT psi.ch] 
>> Sent: Wednesday, June 21, 2017 8:58 AM
>> To: Achtemeier, Gavin <gachtem AT sandia.gov>
>> Cc: h5part AT lists.psi.ch
>> Subject: [EXTERNAL] Re: [H5part] Fortran API fixed?
>> 
>> Hi Gavin
>> 
>>> On 19 Jun 2017, at 18:52, Achtemeier, Gavin <gachtem AT sandia.gov> wrote:
>>> 
>>> I have been having trouble linking to the H5hut/H5Part Fortran library, 
>>> it seems that the compiler does not recognize that the functions are 
>>> defined in the statically linked library. If I explicitly define the H5 
>>> functions my program uses as EXTERNAL  the compiler (both ifort and 
>>> Gfortran) gives “File not found:  'h5part.omesh.o'”.
>> 
>> Can you please send me a short example?
>> 
>>> Additionally I noticed there’s an open ticket(#4) regarding the Fortran 
>>> API having issues and would like to know if that’s still a known issue.
>> 
>> This issue has been fixed in commit #bdcc8f20
>> 
>> Best
>> 
>> Achim
>> 
>> 
>> Paul Scherrer Institut
>> WHGA/U136
>> CH-5232 Villigen PSI
>> +41 56 310 36 74
>> 
>> <Makefile_test><testmain.f><h5parttest.f><myoutput.txt>
> 
> Paul Scherrer Institut
> WHGA/U136
> CH-5232 Villigen PSI
> +41 56 310 36 74
> 

Paul Scherrer Institut
WHGA/U136
CH-5232 Villigen PSI
+41 56 310 36 74