@Gerald-Meyers could you possibly trim that data down to the minimal needed to reproduce (or is this not data dependent)?

This bug doesn't seem data dependent. Here's a small reproduction

import matplotlib.pyplot as plt

fig, ax = plt.subplots(1, 1, figsize=(8, 8),
                        subplot_kw={'projection': 'polar'})
xvals, yvals = [1, 2, 3, 4, 5], [1, 2, 3, 4, 5]
ax.plot(xvals, yvals)
ax.set_yscale('log')
plt.show()

Which gives this error

/home/chahak/Documents/matplotlib/issue_24383.py:14: UserWarning: Attempt to set non-positive ylim on a log-scaled axis will be ignored.
  plt.show()
posx and posy should be finite values
posx and posy should be finite values

Appears to be related to RadialLocator.view_limits

In particular I'm suspicious that the vmin capping at 0 may not be desirable with log scale, but haven't fully worked out the details.

The 0 cap was added to preserve old behaviour from when it didn't support negative radii. However, I don't think log scales were supported then either.

I've idependently stumbled upon this investigating something else. An even smaller example:

import matplotlib.pyplot as plt

fig, ax = plt.subplots(subplot_kw={'projection': 'polar'})
ax.set_yscale('log')

plt.show()

posx and posy should be finite values
posx and posy should be finite values
posx and posy should be finite values
posx and posy should be finite values
/Users/dstansby/github/matplotlib/lib/matplotlib/scale.py:253: RuntimeWarning: overflow encountered in power
  return np.power(self.base, a)

At least one issue is the polar limits are not being set to finite values (ie. on a log scale, vmin > 0):

import matplotlib.pyplot as plt

for kw in [{}, dict(subplot_kw={'projection': 'polar'})]:
    fig, ax = plt.subplots(**kw)
    ax.set_yscale('log')
    print(ax.get_ylim(), kw)

(0.8912509381337456, 11.220184543019636) {}
(0.0, 1.0) {'subplot_kw': {'projection': 'polar'}}

@Gerald-Meyers could you possibly trim that data down to the minimal needed to reproduce (or is this not data dependent)?

I can assure you I tested this on 6 different data sets. The data is not the problem. It it matters, you are more than welcome to slice the data when you np.genfromtxt. The polar plot did function as expected without radial log-scale. However, the circular grid lines for the r values ranged from 0 to 1, not from 0 to 1e7.

For example, here is normal plot of one of my data sets without any log-scale over the domain [20,80]=Δ60 degrees.

Here is the same data set, but with the extra line of code

fig, ax = subplots( 1, 1,
                        figsize=(8, 8),
                        subplot_kw={'projection':'polar'} # <- extra line
                        )

It is worth noting that there are no non-positive(=negative,zero) values in this data set. And the same polar plot when I enable the radial log-scale with
plt.yscale('log')

Here is another data set over a 1 degree domain.

And the same data with polar as before.

And with log scale.

Clearly there are a few more issues than just 'log-scale' broken.

09192022_NaCl_0-2mmS_3mmD_0D_0-0.01inc_0.5dpm(high res).txt
09192022_NaCl_2mmS_3mmD_1D_0-0.05inc_0.5dpm(no shield).txt
09192022_NaCl_2mmS_3mmD_1D-0-0.05inc_0.5dpm.txt
09212022_LiF_0-2mmS_3mmD_0D_0-01inc_0-5dpm(high res).txt
09212022LiF_2mmS_3mmD_1D_0-05inc_10dpm(no shield).txt
09212022LiF_2mmS_3mmD_1D_0-05inc_10dpm(shield).txt

In any case, here is the rest of the data I was using.

I found a solution to this problem, just upgrade the package pip install --upgrade matplotlib