possible 1.2.1 regression in ax.axhline

[ghost]

The 1 pixel workaround seems to work fine for single plots, but I may have
uncovered another regression. Not sure if related or not:

moved here from #1882.

In [1]: fig,ax=plt.subplots(2)
   ...: ax[0].bar(0,5,0.8,log=True, bottom=1,label='a')
   ...: ax[0].bar(1,5000,0.8,log=True, bottom=1,label='a')
   ...: ax[0].axhline(0, color='k', linestyle='--')
Out[1]: <matplotlib.lines.Line2D at 0x38b21d0>

1.2.0

1.2.1

[neggert]

Can I ask again what the expected behavior is? Obviously the plot produced by 1.2.1 isn't correct, but it doesn't look like your line is visible in 1.2.0 either. And it shouldn't be. 0 is outside the bounds of the plot (and always will be for log plots).

[ghost]

up till 1.2.1, IIUC, it was interpreted as a no-op,I understand it's incorrect technically,
but I'm not sure what breaking it achieved. Since it's already in 1.2.1, the horse has left
the stable, and I'll just fix it, I guess it will break other people's code as well though, and
you might get less friction by restoring the no-op behavior with the next point release. or not.

Let me see if fixing this shakes out anything else.

[ghost]

Ok, everything seems to be working fine. Up to you whether to roll back the behaviour
for future release or close as wontfix.

Thank you.

[neggert]

I guess that's up to the repo collaborators. To fix this without undoing the improvements in f6dfcea, I think it would require some work on the set_yscale('log', nonposy=True) code, which I don't know much about.

[mdboom]

I think I'm going to declare this "won't fix". Thanks for the report, and thanks @neggert for the improvements that came out of the discussion.