search

amanzi / amanzi-tpls

TPL Collection for Amanzi
0 stars 2 forks source link

compatibility issue regarding Silo and HDF5 #1

Open ecoon opened 7 years ago

ecoon commented 7 years ago

There is an upstream issue in Silo in which a deprecated HDF5 call was made. With the recent update to newer HDF5, the current Silo release does not compile.

ecoon commented 7 years ago

Hi Ethan,

Apologies you are still running into this.

I've corrected Silo's trunk but am woefully behind in a new public release.

If you continue to encounter issues, pleas let me know and I can provide a pre-release for 4.10.3 (or maybe 4.10.4 at this point)

Mark

"Coon, Ethan T." wrote:

In version 1.8.13 released in 2014, HDF5 removed the function: H5Pset_fapl_mpiposix

It looks like Silo’s latest release 4.10.2 still uses this in silo_hdf5.c.

This:

https://support.hdfgroup.org/HDF5/doc/RM/RM_H5P.html#Property-SetFaplMpiposix

suggests that the call should be replaced by this:

https://support.hdfgroup.org/HDF5/doc/RM/RM_H5P.html#Property-SetFaplMpio

but I’m not sure I understand exactly what goes into the MPI_Info struct, or I’d have included a patch!

Thanks, and apologies if there is a better/preferred place for bug submissions.

Ethan

ecoon commented 7 years ago

So it sounds like we could upgrade to Silo's dev version, if it exists in the public -- will ask Mark this.

lipnikov commented 7 years ago

Ethan, Silo test output_silo passed for me after simple name change for H5Pget_fapl_mpiposix. Either this test is weak or the change really works.

ecoon commented 7 years ago

We don't actually use that branch of the code under typical usage (including that test). We'd have to use Silo with MPIIO. The issue is just getting something to compile, and I don't think the spack diff is much different than that.

I've done the patch and am testing now, will push a new patch for Silo, along with the needed SuperBuild changes, soon.

On Wed, May 31, 2017 at 12:14 AM, lipnikov notifications@github.com wrote:

Ethan, Silo test output_silo passed for me after simple name change for H5Pget_fapl_mpiposix. Either this test is weak or the change really works.

— You are receiving this because you were assigned. Reply to this email directly, view it on GitHub https://github.com/amanzi/amanzi-tpls/issues/1#issuecomment-305077917, or mute the thread https://github.com/notifications/unsubscribe-auth/AE43-wtvlYIQpQyVpW7-kSIZi-gChoZqks5r_OkxgaJpZM4NrBs3 .

--

Ethan Coon

http://www.lanl.gov/expertise/profiles/view/ethan-coon