scrollbar in notebook when not needed?

[jseabold]

Also at 55129be. I don't see an issue related to this, though I recall there being some discussion about new scrollbar features. Consider the below. Is this normal/expected? Is there something I can set in my config. I generated a new profile and didn't see any new variables that might affect this.

[minrk]

Not normal, and not what I experience. Will look into it. Just to confirm, does this still happen up to date as of today?

[jseabold]

Hmm, well, I just tried to pull down master and check and now I get the error messages mentioned in #3265.

[~/src/ipython-skipper] (master)
|23 $ git show -n 1 --format=%h
73770ff

[~/src/ipython-skipper] (master)
|24 $ git clean -xdf
Removing build/
Removing setupbase.pyc
Removing setupext/__init__.pyc
Removing setupext/install_data_ext.pyc
Removing setupext/setupext.pyc

then removed .../dist-packages/IPython/ and .../dist-package/ipython*.egg-info. Did a build then system-wide install.

[minrk]

run python setup.py develop again, or setup.py install or pip install -e ., however you do you dev installs.
Or just

python setup.py submodule

[jseabold]

I almost always install system-wide unless I'm installing from pypi. Removed everything again except profiles and bin scripts. Made sure easy-install.pth was clean. Tried pip install -e and setup.py develop with the same results. Notebook server output

WARNING:root:405 HEAD / (127.0.0.1) 0.44ms
WARNING:root:404 GET /static/components/codemirror/lib/codemirror.css (127.0.0.1) 0.64ms
...

[minrk]

try an explicit python setup.py submodule

[jseabold]

Ok, that bit of magic fixed the problem. What did I just do?

Will see if I can replicate scrollbar issue. Cursory glance says I can't.

[minrk]

it's just git submodule init && git submodule update. The other commands should imply this, but apparently it's not firing, probably because the components directory exists. My guess is this will only happen for people updating IPython dev across the submodule introduction, it probably shouldn't happen for fresh checkouts.

[jseabold]

Ok that makes sense. Thanks.

Come to think of it, this was a notebook that I wrote using IPython.nbformat.current.write. Is it possible that there's something in the json that makes the scrollbar appear? Updating to current master from where I was broke my fragile notebook generating code ('BlockingKernelManager' object has no attribute 'iopub_channel'), so I need to see what happened to try and replicate. Will look into it (and then slowly back away from the bleeding edge).

[jseabold]

Whew ok. Saw my problem with the new kernel code. Nope, I can't replicate the scrollbar issue. I'll close this. Sorry for all the noise.

Another data point for you for the submodules. I'm not used to using git submodules, so I don't know if it's expected. I can't just checkout an old revision. I have to manually remove the components folder.

[minrk]

yeah, that's because we made the somewhat unfortunate choice to turn an existing folder into a submodule rather than putting the submodules into a new folder. Thanks for the data points.

[minrk]

In fact, both git issues here are related to that same decision - the false positive in the initial check for whether the submodule was up to date, and the difficulty rolling back and forth across the transition are both due to the fact that we turned a folder into a submodule rather than adding a submodule in a new location.

[jseabold]

Just in case anyone comes across this and reads the comments this far. The appearance of the scrollbars was due to the zoom in chromium. I see the scrollbar whether I need it or not at some zoom levels (noticed at 110% and 150% but not at other levels).

[minrk]

And for even further reference, this commit: codemirror/codemirror5@0bfab07 should fix the issue, so if there's a CM release before 1.0, we should be set.