Bug#1118038: lmfit-py: tests fail: type int64 is not JSON serializable

[Santiago Vila]

Hi.

This is what debbisect told me:

bisection finished successfully
  last good timestamp: 20251019T083107Z
  first bad timestamp: 20251019T203802Z
the following packages differ between the last good and first bad timestamp:
  diffutils 1:3.10-4 -> 1:3.12-1
  python3-scipy 1.15.3-1.1 -> 1.16.2-2

So, yes, it was the new SciPy.

Drew Parsons wrote:
> It's possible the new version might help, though I haven't tested it.

Good news: I actually finished testing the new version a few minutes
ago, and it fixes the test issue.

(But debian/rules needs some minor adjustments).

> For this particular package I won't rush to make uploads myself since I've
> previously found upstream to be hostile,
> https://github.com/orgs/lmfit/discussions/891

Hmm, I've just read it. Maybe I would call that as "not a lot helpful"
more than "hostile".

If they do not want to support i386, they are in their right.
Then we, in turn, are free to support i386 or not.

I would be willing to try upgrading this package in my personal salsa
namespace and then ask you and the other uploaders (which I'm adding
to CC) for feedback before upload. Does this look like a good plan?


This is a summary of the current patches and what I would like
to do with them:

deactivate_test.patch
  this is very old and I'd like to drop it to see if it's still necessary

0004-jupyter_sphinx-is-not-yet-working-on-Debian.patch
  this one is still necessary (i've just checked)

exclude_test_for_32bit.patch
  I'm in doubt about this one. We could maybe try without it in Salsa CI
  (which has i386) and decide based on the outcome.

fix-test-failure-with-new-uncertainties.patch
  in theory, this one is already included upstream

drop-test-confidence-warnings.patch
  and also this one

Thanks.