[Paraview] ParaView 3.6.2 Release and hdf5-1.8

[Jerry Clarke]

This has been fixed in Xdmf a while ago ...
There are :

#if (H5_VERS_MAJOR>1)||((H5_VERS_MAJOR==1)&&(H5_VERS_MINOR>=8))

directives in XdmfH5Driver.cxx in several places.

Jerry


François Bissey wrote:
> On Fri, 08 Jan 2010 10:21:22 Favre Jean wrote:
>  > I have tried compiling pv3.6.2 with my system's hdf5 and have run 
> into the
>  >  following issue:
>  >
>  > regardless of what api-version I use to configure & compile hdf5 (v16 or
>  >  v18), I cannot compile pv because of the following error:
>  >
>  > /apps-vis/ParaView3-6-2/Utilities/Xdmf2/libsrc/XdmfH5Driver.cxx:174: 
> error:
>  >  braces around scalar initializer for type ‘herr_t (*)(H5FD_t*, unsigned
>  >  char*, unsigned int, hbool_t)’ Badly placed (.
>  > /apps-vis/ParaView3-6-2/Utilities/Xdmf2/libsrc/XdmfH5Driver.cxx:174: 
> error:
>  >  invalid conversion from ‘haddr_t (*)(H5FD_t*)’ to ‘herr_t (*)(H5FD_t*,
>  >  H5FD_mem_t, hid_t, haddr_t, hsize_t)’ Badly placed (.
>  >
>  > etc.
>  >
>  > here is my last config for hdf5:
>  >
>  > ./configure --enable-parallel --enable-linux-lfs --enable-shared
>  >  --with-default-api-version=v18
>  >
>  > The good news:
>  >
>  > I was able to make the compilation finish after copying XdmfH5Driver.cxx
>  >  from the CVS version of ParaView. And so far in execution trials, 
> nothing
>  >  has broken.
>  >
>  > Has anyone else seen this? I had initially configured hdf5 with the 
> default
>  >  (i.e without specifying the last option  
> --with-default-api-version=v??).
>  >  I then tried with v16, and v18, and in both cases, I could not compile,
>  >  until I overwrote the source with the cvs copy.
>  >
> In Gentoo we have a patch for XdmfH5Driver.cxx - It was introduced for
> paraview 3.6.1 but it may still be current.
> I am attaching for reference. One thing to note conditional compilation is
> introduced for HDF5 v1.6 and 1.8. From the look of the patch the code
> that is left alone is geared to HDF5 v1.8 but paraview ship with version 1.6
> so paraview version may have been patched to include some v1.8 bits.
> Most interesting are these 2 bits that are not conditional:
>  @@ -159,6 +165,7 @@
>      H5FD_dsm_close,             /*close         */
>      H5FD_dsm_cmp,               /*cmp           */
>      NULL,                       /*query         */
> +    NULL,                       /*type map      */
>      NULL,                       /*alloc         */
>      NULL,                       /*free          */
>      H5FD_dsm_get_eoa,           /*get_eoa       */
> @@ -168,6 +175,7 @@
>      H5FD_dsm_read,              /*read          */
>      H5FD_dsm_write,             /*write         */
>      NULL,                       /*flush         */
> +    NULL,                       /*truncate      */
>      NULL,                       /*lock          */
>      NULL,                       /*unlock        */
>      H5FD_FLMAP_SINGLE           /*fl_map        */
> 
> this is in the definition of H5FD_class_t H5FD_dsm_g, that would fit 
> with your
> error message ie line 174 the definition supplied is too short and 
> misses some
> fields.
>