fix rounding errors in Hill functions

[murrayrm]

This is a small change in the way propensity functions are handled to avoid errors where X ** n fails if X is a small negative number (like -1e-30). This occurs fairly frequently in BioSCRAPE simulations when a variable is initialized to zero (and apparently numerical errors cause it to go negative).

[murrayrm]

The build that is failing seems to be unrelated to this PR. It looks like there is not an Arm-based wheel for libsbml under Python 3.8. I think this is due to GitHub switching to Arm-based MacOS workflows? Will look into it later.

[ananswam]

Yeah, was just looking into this as well. They just switched to apple silicon for macos it seems. Is it okay to just drop macos python 3.8 (perhaps drop python 3.8 entirely)?

Alternatively, it looks like macos-15-intel in the matrix works for Intel. So, if you wanted, you could run on ARM and Intel macOS and just drop 3.8 from the ARM one as below.

More info here

    1 --- a/.github/workflows/deploy_bioscrape.yml
    2 +++ b/.github/workflows/deploy_bioscrape.yml
    3 @@ -8,4 +8,7 @@
    4        matrix:
    5          os: [windows-latest, ubuntu-latest, macos-latest] # maybe add macos-15-intel here
    6          python-version: ["3.8", "3.11", "3.12"]
    7 +        exclude:
    8 +          - os: macos-latest
    9 +            python-version: "3.8"
   10
   11      steps:

[murrayrm]

Python 3.8 (and 3.9) is past end-of-life, so seems fine to drop it and just leave 3.11 and 3.12. We should probably eventually do 3.11 and 3.13 or 3.14 (most recent releases). I know there were some issues in BioCRNpyler with 3.13 and Numpy 2; not s ure about BioSCRAPE.