Re: big issue with latest hdf5 upgrade to 1.10 in stretch
Christophe TROPHIME
Research Engineer
CNRS - LNCMI
25, rue des Martyrs
BP 166
38042 GRENOBLE Cedex 9
FRANCE
Tel : +33 (0)4 76 88 90 02
Fax : +33 (0) 4 76 88 10 01
Office U 19
M@il : christophe.trophime@lncmi.cnrs.fr
----- Original Message -----
> From: "Anton Gladky" <gladk@debian.org>
> To: "Christophe Trophime" <christophe.trophime@lncmi.cnrs.fr>
> Cc: "Gilles Filippini" <pini@debian.org>, dparsons@debian.org, "Debian Science List"
> <debian-science@lists.debian.org>
> Sent: Monday, February 13, 2017 8:10:29 PM
> Subject: Re: big issue with latest hdf5 upgrade to 1.10 in stretch
>
> Christoph is right. Med-ficher fails on autopkgtests since the last
> NMU. Unit tests are also failing, but they are not causing FTBFSs
> due to explicit ignoring of its failures.
>
> @Christoph, is there any chance to fix it in med-ficher?
I don't know how to fix the pb.
I had some contact with upstream. As far as I've understood they wont fix this because
hdf5 1.10 introduce some problem of compatibility...
A specific release (4.xx) will adress this issue and fix the pbs of compatibility
but in the meantime we are stuck!!
>
> Regards
>
> Anton
>
>
> 2017-02-13 15:12 GMT+01:00 Christophe Trophime
> <christophe.trophime@lncmi.cnrs.fr>:
> >
> >
> >
> >
> > Christophe TROPHIME
> > Research Engineer
> >
> > CNRS - LNCMI
> > 25, rue des Martyrs
> > BP 166
> > 38042 GRENOBLE Cedex 9
> > FRANCE
> >
> > Tel : +33 (0)4 76 88 90 02
> > Fax : +33 (0) 4 76 88 10 01
> > Office U 19
> > M@il : christophe.trophime@lncmi.cnrs.fr
> >
> > ----- Original Message -----
> >> From: "Gilles Filippini" <pini@debian.org>
> >> To: "Christophe Trophime" <christophe.trophime@lncmi.cnrs.fr>
> >> Cc: dparsons@debian.org, debian-science@lists.debian.org
> >> Sent: Monday, February 13, 2017 10:56:40 AM
> >> Subject: Re: big issue with latest hdf5 upgrade to 1.10 in stretch
> >>
> >> Hi,
> >>
> >> On 2017-02-13 10:49, Christophe Trophime wrote:
> >> > ----- Original Message -----
> >> >> From: "Drew Parsons" <dparsons@debian.org>
> >> >> To: "Christophe Trophime" <christophe.trophime@lncmi.cnrs.fr>,
> >> >> debian-science@lists.debian.org
> >> >> Sent: Monday, February 13, 2017 10:34:23 AM
> >> >> Subject: Re: big issue with latest hdf5 upgrade to 1.10 in stretch
> >> >>
> >> >> On Mon, 2017-02-13 at 08:46 +0100, Christophe Trophime wrote:
> >> >> > Hi,
> >> >> > I would like to draw your attention to a problem that makes several
> >> >> > packages unusable.
> >> >> >
> >> >> > Since the upgrade of hdf5 to version 1.10 there is an issue with the
> >> >> > med-fichier library
> >> >> > which is used for several packages including gmsh, code_saturne,
> >> >> > code_aster.
> >> >> >
> >> >> > Actually med-fichier 3.xx do not support hdf5 format for version
> >> >> > greater than 1.8.xx.
> >> >> > So the upgrade of hdf5 breaks all the packages that rely on med-
> >> >> > fichier library.
> >> >> > This is a big issue especially as hdf5 1.10 will be in the next
> >> >> > stable debian version.
> >> >>
> >> >>
> >> >> What is the failure mode? med-fichier 3.0.6-10.1 has already been
> >> >> patched to build against hdf5 1.10.
> >> >>
> >> >
> >> > Typical failure is the following from Salome Forum:
> >> > Hello all,
> >> >
> >> > I'm trying to set up a CFD case with Salome, export the mesh in a MED
> >> > file and then process it with Code Saturne. CS doesn't read the MED
> >> > file and gives the error in the preprocessing stage:
> >> >
> >> > _MEDdatasetRd.c [139] : Erreur à la lecture du dataset
> >> > _MEDdatasetRd.c [139] : COO
> >> > HDF5-DIAG: Error detected in HDF5 (1.10.0-patch1) thread 0:
> >> > #000: ../../../src/H5Dio.c line 170 in H5Dread(): can't read data
> >> > major: Dataset
> >> > minor: Read failed
> >> > #001: ../../../src/H5Dio.c line 418 in H5D__read(): unable to set up
> >> > type info
> >> > major: Dataset
> >> > minor: Unable to initialize object
> >> > #002: ../../../src/H5Dio.c line 953 in H5D__typeinfo_init(): not a
> >> > datatype
> >> > major: Invalid arguments to routine
> >> > minor: Inappropriate type
> >> > _MEDdatasetRd.c [142] : _dataset = 2
> >> > _MEDdatasetRd.c [143] : _i = 0
> >> > _MEDdatasetRd.c [144] : (*filter).memspace[_i] = 2
> >> > _MEDdatasetRd.c [145] : (*filter).diskspace[_i] = 3
> >> > _MEDmeshAdvancedRd30.c [462] : Erreur à la lecture du dataset
> >> > _MEDmeshAdvancedRd30.c [462] :
> >> > _MEDmeshAdvancedRd30.c [463] : meshname = "Mesh_water"
> >> > _MEDmeshAdvancedRd30.c [463] : numit = -1
> >> > _MEDmeshAdvancedRd30.c [463] : numdt = -1
> >> > _MEDmeshAdvancedRd30.c [463] : _datagroupname2 = "NOE"
> >> > _MEDmeshAdvancedRd30.c [464] : _datagroupname3 = ""
> >> > _MEDmeshAdvancedRd30.c [464] : _profilename = "MED_NO_PROFILE_INTERNAL"
> >> >
> >> > The problem can be because I installed Salome from the
> >> > SALOME-7.6.0-DB07.tgz archive whilst Code Saturne is a Debian package
> >> > and is installed via the Debian package system which is much easier
> >> > than installing the original package. So the med and hdf5 libraries of
> >> > CS come from the Debian repositories. I can't use newer versions of
> >> > Salome (7.7.1 or 7.8.0) as they produce MED3.1 files but my CS reads
> >> > only MED3.0 files.
> >> >
> >> > The details of my machine:
> >> >
> >> > Debian stretch (testing) on AMD64
> >> >
> >> > Salome 7.6.0 from SALOME-7.6.0-DB07.tgz
> >> >
> >> > Code Saturne 4.3 Debian package
> >> >
> >> > libmed 3.0.6-10.1 Debian package
> >> >
> >> > hdf5 1.10.0 Debian package
> >> >
> >> > Not knowing if this problem comes from how I create the mesh I tried
> >> > to run one of the test cases from the Code Saturne tutorials but ran
> >> > into other problems.
> >> >
> >> > If you need any other information pleas let me know. Any hint on how
> >> > to go further is very welcome.
> >> >
> >> > Best regards
> >> >
> >> > Elektrix
> >> > """
> >> >
> >> > Basically you can generate file but can no longer read it...
> >>
> >> Something like a test case as simple as possible would help.
> >> Thanks in advance,
> >
> > Take any med file from saturne, aster and salome and you will run into the
> > problem...
> > It's maybe more a break of compatibility from med-fichier with hdf5 1.8.xx
> >
> >
> >>
> >> _g.
> >>
> >>
> >
>
Reply to: