Bug#1149312: pytorch-geometric: FTBFS against python 3.15 (several test errors)

[Margarita Manterola]

Package: src:pytorch-geometric
Version: 2.8.0-1
User: debian-pyt...@lists.debian.org
Usertags: python3.15
Tags: ftbfs, forky, sid
Severity: important
Hi!

While rebuilding the python related packages against the Python 3.15 
version we found that pytorch-geometric fails to build from source:
https://debusine.debian.net/debian/r-python-python3.15/artifact/4856933/

In the log there are several errors. One that is repeated a few times is 
this one:
                annotation = copy.copy(annotation)
              annotation.__args__ = args
                ^^^^^^^^^^^^^^^^^^^
E               AttributeError: readonly attribute

There's also this one:

      ForkingPickler(file, protocol).dump(obj)
E       TypeError: cannot pickle 'sqlite3.Connection' 
object
E       when serializing dict item '_connection'
E       when serializing 
torch_geometric.data.database.SQLiteDatabase state
E       when serializing 
torch_geometric.data.database.SQLiteDatabase object
E       when serializing dict item '_db'
E       when serializing 
torch_geometric.data.on_disk_dataset.OnDiskDataset state
E       when serializing 
torch_geometric.data.on_disk_dataset.OnDiskDataset object
E       when serializing tuple item 1
E       when serializing dict item '_args'
E       when serializing multiprocessing.context.Process 
state
E       when serializing multiprocessing.context.Process 
object

And this one:

      assert 
utils.supports_bipartite_graphs('SAGEConv')
E       AssertionError: assert False
E        +  where False = <function supports_bipartite_graphs 
at 0x7f81b7787480>('SAGEConv')
E        +    where <function supports_bipartite_graphs at 
0x7f81b7787480> = utils.supports_bipartite_graphs

Please note that while we ran into these errors during the rebuild 
against Python 3.15, that doesn't mean that the root cause for the 
failure is necessarily the new Python version. In this case, some of the 
test errors might be due to some other change in the dependencies.

Setting the severity to important for now. Once Python 3.15 is released 
and later added to python3-defaults, this bug will become release 
critical.

--
Regards,
Marga