fix failing CI tests

[Tieqiong]

closes #238

@sbillinge @bobleesj The failing CI was caused by problematic distribution of pdffit2 on pypi, namely the 1.4.4rc6 version. The problem is very likely caused by the wheel built by CI for windows not functional, even if the wheels had been tested locally. Let me research more

As for the test here, let's use 1.4.4rc4 for now, which build the wheel on the fly with gsl.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 95.11%. Comparing base (0e230e0) to head (1bc095c).
Report is 1 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #240   +/-   ##
=======================================
  Coverage   95.11%   95.11%           
=======================================
  Files          23       23           
  Lines        1126     1126           
=======================================
  Hits         1071     1071           
  Misses         55       55           

[bobleesj]

As for the test here, let's use 1.4.4rc4 for now, which build the wheel on the fly with gsl.

Yes, I had a Windows laptop and tried to install the latest pdffit2 but failed to install. I don't have a windows laptop now but glad that it works with 1.4.4rc4

[bobleesj]

Great progress! By the way, we’ve started adopting this practice for commits and issues: Billingegroup/scikit-package#212. It’s been quite useful now that things are rolling and the habit is now almost automatic.

[sbillinge]

@Tieqiong thanks, that's great that this is passing. I guess this was largely for testing so we don't want to merge this, right? Do we need to mark the rogue pypi release as broken (which means it won't be installed unless explicitly asked for), and then push out a new, unbroken release? @Tieqiong now the trouble-shooting has been done successfully, an issue with a set of Todos that we need to follow one after the other to clean things up would be great. Either that, or a release milestone (I will create it now) and issues we need to close to get there? This is HUGE! We have many pdfgui users and a number are bumping into numpy2 and other issues with installation.

[Tieqiong]

@sbillinge
It wouldn't hurt to merge this, and this can provide temporary CI coverage before a new pdffit2 gets released.

Yes let's mark pdffit2 1.4.4rc6 as broken. I guess this is more for us (when doing developmental works) as I don't think pypi will automatically install rc versions unless specified.

I'll gather the thoughts and make a list of issues. I think there are still potentially some small things needed for a perfect release.

[sbillinge]

@Tieqiong ok, I merged this pr and also yanked 1.4.4rc6 from pypi. Let me know if I need to do anything else