ax.twinx().plot() will reset the x_limits if only an axvspan was used on ax

[Redoubts]

Create a figure and axis, and add a vspan to that axis. If you twinx it, and apply a plot to the twin, the x_limits can adjust, potentially making the vspan not visible.

Minimal code sample:

fig, ax = plt.subplots(figsize=(12, 10))
ax.axvspan(1,2)
fig.canvas.draw()

bx = ax.twinx()
bx.plot([0,.5], [1,2])
fig.canvas.draw()

Note how the x-axis labels shifted away from the range (1,2).
I expected:

• Matplotlib version, Python version and Platform (Windows, OSX, Linux ...)
  Matplotlib 1.5.1 on OS X 10.11 with Python 2.7.10
• System python, pip install -U matplotlib (with SIP disabled)

[tacaswell]

the autoscaling is apparently not twin axis aware enough.

[patniharshit]

What is the issue here?

[ibnIrshad]

I think we've tracked down the problem. autoscale_view apparently ignores ALL axis limits when atleast one of the limits is infinite (in this case, the y-limits of a vspan are infinite). We have a PR which fixes this by taking into account the case when either y-limits or x-limits are infinite, but the other is finite. PR coming shortly.