Github actions: DEVEL_BUILD=OFF build fails because of numpy version

[KrisThielemans]

One of the native jobs fails with

https://github.com/SyneRBI/SIRF-SuperBuild/actions/runs/20039734624/job/57470452101#step:14:1121

This is due to SyneRBI/SIRF#1363.

Fixing this in SIRF might be virtually impossible, @evgueni-ovtchinnikov is that correct?

The GHA native jobs use pip and

SIRF-SuperBuild/docker/user_python-ubuntu.sh

Line 61 in 06c7d30

${PYTHON} -m pip install -U -r requirements.txt

SIRF-SuperBuild/docker/requirements.txt

Line 4 in 06c7d30

numpy # CIL

I think it would make sense to ask for numpy>=2 in our requirements.txt. (not sure why it gives us numpy 1.*). It would help others as well

@casperdcl what do you think?

[casperdcl]

numpy is downgraded by https://github.com/SyneRBI/SIRF-SuperBuild/actions/runs/19765369296/job/56636734269?pr=973#step:11:3724

Collecting numpy<2.0.0,>=1.23 (from cil==25.0.1.dev13+g1cf84720a.d20251128)

which is the same error as in https://github.com/SyneRBI/SIRF-SuperBuild/actions/runs/20039734641/job/57470452118.

Note that TomographicImaging/CIL#2218 is required for numpy>=2 support

[KrisThielemans]

of course it wasn't going to be easy...

[evgueni-ovtchinnikov]

Fixing this in SIRF might be virtually impossible, @evgueni-ovtchinnikov is that correct?

I am afraid it is. AcquisitionDataView objects cannot be possibly used, and asarray() falling back to as_array() in PET may lead to confusion. Requiring numpy.2* is the only way to fix this I believe.