Bug#1121356: opm-simulators FTBFS with scotch 7.0.10

[Markus Blatt]

Hi,

On 11/25/25 22:52, Drew Parsons wrote:
> On 2025-11-25 22:00, Adrian Bunk wrote:
>> [ Adding Drew to Cc. ]
>>
>> On Tue, Nov 25, 2025 at 08:57:29PM +0100, Markus Blatt wrote:
>>> Hi,
>>>
>>> Both DUNE and OPM have their own cmake scripts for finding 
>>> metis/parmetis or
>>> the appropriate wrappers from scotch.
>>>
>>> It seems like the libraries from scotch changed and they are now
>>> libscotchmetisv3 and libscotchmetisv5 instead of libscotchmetis.
>>>
>>> This broke assumptions made by upstream.
I took another look and it seems like DUNE and OPM are parsing headers 
to find out what version this is (v5 or v3).
That might be quite fragile.
>
>
> But it appears to be sorted out now, isn't it?
>
> opm-simulators are now building against the rebuilt libopm-grid-dev.
>
> Looks like this was just a transition bug caught in the gap of the 
> scotch transition. Like it hadn't waited before libopm-grid-dev was 
> rebuilt and ready.
>
You might be right that the error appeared due to this ( opm-grid still 
is linked to scotch, but the new build of opm-simulators does not find 
it) and is gone now (both don't find scotch.

Those builds are now without using scotch. But that is certainly another 
bug.

If this is true, should I wait for the transition to complete or is it  
OK to upload fixes now. It is kind of connected.

Thanks for the update.

Best,

Markus