[MDAL-Developer] Compiling with HDF5 1.10

[Paul Harwood]

OK - I will create a ticket.

I have the recipe for a Windows build using conda - pace the HDF5 and GRIB
problems - and can publish that if you want - create a ticket (suggesting
where to put the info) and I will PR to it at some point over the next few
weeks.

Of course - we could also add MDAL to conda-forge. But I am not sure I am
up to working out how to do that at the moment. I have created a Unity
package that loads MDAL into a Unity project -
https://github.com/ViRGIS-Team/mdal-upm - that also shows how to use MDAL
in c#. That is what how our project consumes MDAL now. The Unity package
will now auto update for each new release of MDAL.

I am also scheduled to try the same on OSX and Linux over the next few
weeks. Will let you know.

I have no problems with the GDAL dependency. I have problems with the
OSGEO4W dependency because that is basically a pit of despair - and GDAL
have moved to conda as far as I can see.

It would seem to me that managing the dependency to GDAL is no worse than
the job we would have managing the dependency to next level down and there
are probably more interesting things people could be spending their time on
- but there may be some functional restrictions brought by GDAL that I am
not aware of.

On Wed, 21 Oct 2020 at 11:22, Peter Petrik <
peter.petrik at lutraconsulting.co.uk> wrote:

> Yeah, I haven't  tested  with  that version. Best  to create a ticket so
> we can look at it later. I think that we  may even add a windows build
> based on conda to the MDAL repo, if you like to do so.
> For GDAL dependency, we want to remove it, but first we need to implement
> GRIB and NetCDF drivers without need of GDAL.
>
> Peter
>
> On Wed, Oct 21, 2020 at 11:58 AM Paul Harwood <runette at gmail.com> wrote:
>
>> Heho
>>
>> I am compiling through a GH Action - which you can see for yourself here :
>>
>>
>> https://github.com/ViRGIS-Team/mdal-upm/blob/main/.github/workflows/winbuild2.yml
>>
>> IMPORTANT NOTE - before you say. This action CURRENTLY has the
>> -DWITH_HDF5=OFF. That is so it works despite this issue. If it were not to
>> have this flag then one would see the error messages I mentioned.
>>
>> Also :
>>
>> - I know this is always a clean environment because - GH Action.
>> - This is VERY DEFINITELY compiling against a different version of HDF5
>> since I am building against the conda repo of GDAL 3.1.3 and not the
>> OSGEO4W repo. Sorry, I thought I made that clear in the title.
>>
>> As far as I am aware the conda repo uses HDF5 1.10.6 and the OSGEO4W repo
>> seems to use 1.8.14 (or 1.8.11 in the choco repo).
>>
>> I know this is NOT the MDAL supported version - but it is the current
>> "preferred" build for GDAL and - since the latest HDF5 version is 1.12,
>> using 1.8 does seem very backward. QGIS and MDAL probably need to grapple
>> with this at some point.
>>
>> However - it seems that there are some compile-time changes needed after
>> 1.8.14. I just wondered if anyone had already solved this?
>>
>> In the short term - for us compiling without HDF is not a huge issue. And
>> - I guess we will have to get around to solving it sometime.
>>
>> Br
>> P
>>
>> On Wed, 21 Oct 2020 at 08:38, Peter Petrik <
>> peter.petrik at lutraconsulting.co.uk> wrote:
>>
>>> Hi Paul,
>>>
>>> can you please post here your CMAKE command you use to generate the
>>> build system and the output of the command.
>>> Also make sure you start from the empty build directory when running it.
>>> This looks like you are linking against wrong version of hdf5 library.
>>>
>>> Peter
>>>
>>>
>>> On Tue, Oct 20, 2020 at 9:18 PM Paul Harwood <runette at gmail.com> wrote:
>>>
>>>> I am not sure this is a bug as such - so I thought I would try out the
>>>> new mailing list before raising as an issue.
>>>>
>>>> For various reasons - I am trying to compile MDAL against the current
>>>> conda version of GDAL - 3.1.3 - and the conda versions of the deps.
>>>>
>>>> It is mostly working - pace the testing issues I have raised. However I
>>>> get a fatal error (shown below) when compiling against HDF5 - which in this
>>>> case is v1.10.6.
>>>>
>>>> This appears to be a known issue - for instance here
>>>> https://github.com/conda-forge/gdal-feedstock/issues/25 and other
>>>> mentions. There seems to be a need for a compiler definition or flag and I
>>>> have tried playing with this - but nothing works.
>>>>
>>>> Does anyone have any idea of what I need to do? Or where to put the
>>>> compiler definition?
>>>>
>>>>
>>>> mdal_hdf5.obj : error LNK2001: unresolved external symbol H5T_C_S1_g
>>>> [C:\Users\runes\Documents\GitHub\MDAL\build\mdal\mdal.vcxproj]
>>>> mdal_hec2d.obj : error LNK2001: unresolved external symbol H5T_C_S1_g
>>>> [C:\Users\runes\Documents\GitHub\MDAL\build\mdal\mdal.vcxproj]
>>>> mdal_hdf5.obj : error LNK2001: unresolved external symbol
>>>> H5T_NATIVE_INT_g
>>>> [C:\Users\runes\Documents\GitHub\MDAL\build\mdal\mdal.vcxproj]
>>>> mdal_flo2d.obj : error LNK2001: unresolved external symbol
>>>> H5T_NATIVE_INT_g
>>>> [C:\Users\runes\Documents\GitHub\MDAL\build\mdal\mdal.vcxproj]
>>>> mdal_hec2d.obj : error LNK2001: unresolved external symbol
>>>> H5T_NATIVE_INT_g
>>>> [C:\Users\runes\Documents\GitHub\MDAL\build\mdal\mdal.vcxproj]
>>>> mdal_hdf5.obj : error LNK2001: unresolved external symbol
>>>> H5T_NATIVE_FLOAT_g
>>>> [C:\Users\runes\Documents\GitHub\MDAL\build\mdal\mdal.vcxproj]
>>>> mdal_flo2d.obj : error LNK2001: unresolved external symbol
>>>> H5T_NATIVE_FLOAT_g
>>>> [C:\Users\runes\Documents\GitHub\MDAL\build\mdal\mdal.vcxproj]
>>>> mdal_hec2d.obj : error LNK2001: unresolved external symbol
>>>> H5T_NATIVE_FLOAT_g
>>>> [C:\Users\runes\Documents\GitHub\MDAL\build\mdal\mdal.vcxproj]
>>>> mdal_hdf5.obj : error LNK2001: unresolved external symbol
>>>> H5T_NATIVE_DOUBLE_g
>>>> [C:\Users\runes\Documents\GitHub\MDAL\build\mdal\mdal.vcxproj]
>>>> mdal_flo2d.obj : error LNK2001: unresolved external symbol
>>>> H5T_NATIVE_DOUBLE_g
>>>> [C:\Users\runes\Documents\GitHub\MDAL\build\mdal\mdal.vcxproj]
>>>> mdal_hdf5.obj : error LNK2001: unresolved external symbol
>>>> H5T_NATIVE_UINT8_g
>>>> [C:\Users\runes\Documents\GitHub\MDAL\build\mdal\mdal.vcxproj]
>>>> C:\Users\runes\Documents\GitHub\MDAL\build\mdal\Debug\mdal.dll : fatal
>>>> error LNK1120: 5 unresolved externals
>>>> [C:\Users\runes\Documents\GitHub\MDAL\build\mdal\mdal.vcxproj]
>>>>
>>>> _______________________________________________
>>>> MDAL-Developer mailing list
>>>> MDAL-Developer at lists.osgeo.org
>>>> https://lists.osgeo.org/mailman/listinfo/mdal-developer
>>>>
>>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/mdal-developer/attachments/20201021/14b664e9/attachment-0001.html>