Horizontal colorbar position off in pdf when using subplots with constrained_layout

[maxnoe]

Bug report

Bug summary

When storing plots as pdf, the colorbar is at the wrong position, off by a small amount:

Code for reproduction

import matplotlib.pyplot as plt
import numpy as np
import matplotlib

print(matplotlib.__version__)

rng = np.random.default_rng(0)
image = rng.normal(size=(40, 40))

# only happens at this exact figure size....
fig, (ax1, ax2) = plt.subplots(1, 2, figsize=(2.976338729763387, 1.4881693648816936), constrained_layout=True)
img1 = ax1.imshow(image)
img2 = ax2.imshow(image)
fig.colorbar(img1, orientation='horizontal', ax=ax1)
fig.colorbar(img2, orientation='horizontal', ax=ax2)
plt.savefig('colorbar.pdf')

Actual outcome

png converted from pdf output using pdftoppm -r 600 -png -singlefile colorbar.pdf colorbar:

Matplotlib version

• Operating system: Manjaro Linux
• Matplotlib version (import matplotlib; print(matplotlib.__version__)): 3.4.2
• Matplotlib backend (print(matplotlib.get_backend())): pdf (since plot saved as pdf)
• Python version: 3.9.5

[jklymak]

It's not clear what you mean by "wrong". Can you clarify?

[LukasNickel]

More or less exact.
figsize=(2.97, 1.488) and some slightly different values will also do the trick

Edit:
Also happens without the constrained_layout

[LukasNickel]

It's not clear what you mean by "wrong". Can you clarify?

If you look closely at the colorbar, the colors and the border do not align perfectly
see here:

[jklymak]

I believe this is #6827. (Note it has nothing to do with constrained_layout.). Suggest we close as a duplicate unless you think this is different.

[LukasNickel]

Interesting!
It seems indeed this only happens with the dpi set to <=120

[jklymak]

I mean it always happens, you just can't see it at higher resolution 😉