Bug#1131087: trollimage: autopkgtest regression with NumPy 2.4

[Antonio Valentino]

On Tue, 17 Mar 2026 11:46:33 +0100 roehling at debian.org wrote:
> Source: trollimage
> Version: 1.28.0-1
> Severity: important
> User: debian-python at lists.debian.org
> Usertags: numpy2.4
> 
> Dear maintainer,
> 
> According to https://ci.debian.net data, your package trollimage has an
> autopkgtest regression with numpy.
> 
> The following architectures failed: amd64, arm64.
> 
> Hopefully relevant excerpt from
> https://ci.debian.net/packages/t/trollimage/unstable/amd64/69507135/
> follows:
> 
> 112s =================================== FAILURES ===================================
> 112s                 f"`dtype` inference failed in `{funcname}`.\n\n"
> 112s E           ValueError: `dtype` inference failed in `map_blocks`.
> 112s E           
> 112s E           Please specify the dtype explicitly using the `dtype` kwarg.
> 112s E           
> 112s E           Original error is below:
> 112s E           ------------------------
> 112s E           TypeError("nanquantile() got an unexpected keyword argument 'interpolation'")
> 112s E           
> 112s E           Traceback:
> 112s E           ---------
> 112s E             File "/usr/lib/python3/dist-packages/dask/array/core.py", line 469, in apply_infer_dtype
> 112s E               o = func(*args, **kwargs)
> 112s E             File "/usr/lib/python3/dist-packages/dask/array/reductions.py", line 1959, in _custom_quantile
> 112s E               return np.nanquantile(
> 112s E                      ~~~~~~~~~~~~~~^
> 112s E                   a,
> 112s E                   ^^
> 112s E               ...<5 lines>...
> 112s E                   **kwargs,
> 112s E                   ^^^^^^^^^
> 112s E               )
> 112s E               ^
> 112s msg        = '`dtype` inference failed in `map_blocks`.\n\nPlease specify the dtype explicitly using the `dtype` kwarg.\n\nOriginal...    ~~~~~~~~~~~~~~^\n        a,\n        ^^\n    ...<5 lines>...\n        **kwargs,\n        ^^^^^^^^^\n    )\n    ^\n'
> 112s                 f"`dtype` inference failed in `{funcname}`.\n\n"
> 112s E           ValueError: `dtype` inference failed in `map_blocks`.
> 112s E           
> 112s E           Please specify the dtype explicitly using the `dtype` kwarg.
> 112s E           
> 112s E           Original error is below:
> 112s E           ------------------------
> 112s E           TypeError("nanquantile() got an unexpected keyword argument 'interpolation'")
> 112s E           
> 112s E           Traceback:
> 112s E           ---------
> 112s E             File "/usr/lib/python3/dist-packages/dask/array/core.py", line 469, in apply_infer_dtype
> 112s E               o = func(*args, **kwargs)
> 112s E             File "/usr/lib/python3/dist-packages/dask/array/reductions.py", line 1959, in _custom_quantile
> 112s E               return np.nanquantile(
> 112s E                      ~~~~~~~~~~~~~~^
> 112s E                   a,
> 112s E                   ^^

The issue seems to be due to an incompatibility of dask 2024.12.1 with
recent versions of numpy.

In this specific case the "interpolation" keyword argument has been
removed from the "(nan)quantile" function in numpy v2.4 but it is still
used in dask.

Please note that this specific issue has been fixed upstream in
https://github.com/dask/dask/pull/12108.


kind regards
-- 
Antonio Valentino