numpy build broken with SAGE_FAT_BINARY=yes

[mkoeppe]

Follow up from #29537.

Depends on #29537

CC: @embray @kliem @dimpase @vbraun

Component: packages: standard

Author: Jonathan Kliem

Branch/Commit: 128e8ef

Reviewer: Matthias Koeppe

Issue created by migration from https://trac.sagemath.org/ticket/31521

[mkoeppe]

comment:1

On macOS, with SAGE_FAT_BINARY=yes, I just got:

    [numpy-1.19.5]   Running from numpy source directory.
    [numpy-1.19.5]   Cythonizing sources
    [numpy-1.19.5]   usage: setup.py [global_opts] cmd1 [cmd1_opts] [cmd2 [cmd2_opts] ...]
    [numpy-1.19.5]      or: setup.py --help [cmd1 cmd2 ...]
    [numpy-1.19.5]      or: setup.py --help-commands
    [numpy-1.19.5]      or: setup.py cmd --help
    [numpy-1.19.5]
    [numpy-1.19.5]   error: option --cpu-dispatch not recognized
    [numpy-1.19.5]   ********************************************************************************
    [numpy-1.19.5]   Error building a wheel for numpy-1.19.5

[mkoeppe]

comment:3

Indeed, --cpu-dispatch has only been introduced in the 1.20.x series. We are still at 1.19.5.

Wondering how this has possibly passed in previous tests?!

[mkoeppe]

comment:4

And I guess the prefix SAGE_ can be removed in the variable names in sage_conf...

[kliem]

comment:5

Reverting for now until numpy is updated.

---

New commits:

128e8ef

Revert "do not allow numpy intrinsics when building fat binary"

[kliem]

Author: Jonathan Kliem

[kliem]

Commit: 128e8ef

[kliem]

Branch: public/31521

[mkoeppe]

Reviewer: Matthias Koeppe

[mkoeppe]

comment:6

Thanks!

But we'll need another ticket then to actually fix the nonportability of the build...

[mkoeppe]

comment:7

Replying to @mkoeppe:

But we'll need another ticket then to actually fix the nonportability of the build...

That's now #31565.

[vbraun]

Changed branch from public/31521 to 128e8ef