[SyneRBI/SIRF-SuperBuild] Github actions: DEVEL_BUILD=OFF build fails because of numpy version (Issue #979)

Kris Thielemans

unread,

Dec 8, 2025, 3:32:53 PM12/8/25

to SyneRBI/SIRF-SuperBuild, Subscribed

KrisThielemans created an issue (SyneRBI/SIRF-SuperBuild#979)

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
https://github.com/SyneRBI/SIRF-SuperBuild/blob/06c7d308a3592549ed55798af91dc91377f05d16/docker/user_python-ubuntu.sh#L61
https://github.com/SyneRBI/SIRF-SuperBuild/blob/06c7d308a3592549ed55798af91dc91377f05d16/docker/requirements.txt#L4

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?

—
Reply to this email directly, view it on GitHub, or unsubscribe.
You are receiving this because you are subscribed to this thread.

Casper da Costa-Luis

unread,

Dec 8, 2025, 3:43:43 PM12/8/25

to SyneRBI/SIRF-SuperBuild, Subscribed

casperdcl left a comment (SyneRBI/SIRF-SuperBuild#979)

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

—
Reply to this email directly, view it on GitHub, or unsubscribe.

You are receiving this because you are subscribed to this thread.

Kris Thielemans

unread,

Dec 8, 2025, 3:59:07 PM12/8/25

to SyneRBI/SIRF-SuperBuild, Subscribed

KrisThielemans left a comment (SyneRBI/SIRF-SuperBuild#979)

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

—
Reply to this email directly, view it on GitHub, or unsubscribe.

You are receiving this because you are subscribed to this thread.

Evgueni Ovtchinnikov

unread,

Dec 9, 2025, 5:21:42 AM12/9/25

to SyneRBI/SIRF-SuperBuild, Subscribed

evgueni-ovtchinnikov left a comment (SyneRBI/SIRF-SuperBuild#979)

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.