Fix bug when stacking histograms with non-integer weights - #1372

Can we have a test so that this doesn't crop back up.

Also, I have to wonder why the original code explicitly cast the zeros as integers? Maybe there was a reason for it?

I'll work on a test for this.

I noticed that when running the existing tests, I get this warning

/usr/local/lib/python2.7/dist-packages/matplotlib/axes.py:8080: DeprecationWarning: Implicitly casting between incompatible kinds. 
In a future numpy release, this will raise an error. Use casting="unsafe" if this is intentional.     
  m += mlast 

It seems like np.histogram can return an array of either floats or ints depending on what it's given as input. We probably want to get rid of this warning, so I'll change this to grab the type from the np.histogram bins.

+1.

I milestoned this for 1.2. Regarding the discussion on the mailing list, 1.2 should be reserved only for critical bugs. Does this qualify as critical?

It looks like something it would be very nice to include. I would prefer seeing it rebased against v1.2.x and merged there directly rather than cherry-picked.

Alright, rebased onto 1.2.x. That was a learning experience in git.

Was that the proper way to rebase it? I notice it seems to have grabbed a bunch of 1.2.x changes and included them in the PR.

Nevermind, figured it out. See #1406.

@neggert, Thanks very much for the rebase, which I committed, as well as the original PR.