Neotree Update Breaks Workflow

[jupl]

Description

Something changed where Neotree is becoming disruptive to my workflow:

• When I have Neotree open (SPC p t) in a project and try to open a terminal using SPC a s t the terminal no longer pops up from underneath with the message split-window: Cannot split side window or parent of side window. SPC w p m to show the message buffer does work though.
• When I have Neotree open (SPC p t) in a project and try to open Magit using SPC g s with git-magit-status-fullscreen enabled the status window appears with Neotree still visible when it wasn't visible before. This is problematic because now when I want to commit (c c) it only show one pane instead of two and it doesn't show the COMMIT_EDITMSG buffer anymore and have to look for it.

In both cases things work when I hide the Neotree pane, which is an inconvenience to hide every time I want to do something like the above.

EDIT: There were some commits to Neotree within a week that may be contributing to this. I did a rollback which had an older Neotree and everything is working.

Reproduction guide 🪲

• Start Emacs with packages up to date
• Follow above steps

Observed behaviour: 👀 💔

• Mentioned above

Expected behaviour: ❤️ 😄

• Not mentioned above

System Info 💻

• OS: darwin
• Emacs: 25.1.1
• Spacemacs: 0.200.2
• Spacemacs branch: master (rev. 7bdde17)
• Graphic display: t
• Distribution: spacemacs
• Editing style: vim
• Completion: helm
• Layers:

(typescript colors docker editorconfig emacs-lisp html javascript markdown org-plus osx react restclient shell shell-scripts syntax-checking vimscript yaml
            (auto-completion :variables auto-completion-return-key-behavior nil auto-completion-tab-key-behavior 'complete auto-completion-enable-help-tooltip t)
            (clojure :variables clojure-enable-fancify-symbols t)
            (git :variables git-magit-status-fullscreen t)
            (version-control :variables version-control-diff-tool 'diff-hl))

[bmag]

Regarding shell-pop (SPC a s t), this seems to be a problem in shell-pop when shell-pop-full-span is non-nil and neotree is open. @jupl Can you try and reproduce it with stock emacs + neotree + shell-pop, and report upstream to shell-pop?
You can use SPC q D to restart emacs with only neotree and shell-pop (without spacemacs), or run emacs -Q and evaluate this:

(package-initialize)
(require 'neotree)
(require 'shell-pop)

Regarding Magit, there's an open a merged PR (which I can't find right now #7085) for using a function provided by Magit, instead of our own custom one. You can try it for yourself by setting magit-display-buffer-function to magit-display-buffer-fullframe-status-v1, and if that doesn't work then it should be another report upstream (to Magit).

The crucial change in Neotree is that now it displays its buffer in a proper side window (with display-buffer-in-side-window). When there is a side window, its parent (the frame-root-window) can't be split, unless ignore-window-parameters is nil (usually it is let-bound to temporarily ignore parameters). Any feature that uses (split-window (frame-root-window)) is potentially affected.

Edit: found the PR, it has been merged 6 hours ago

[jupl]

@bmag Thank you for the thorough explanation!

Regarding shell-pop and neotree yes I can reproduce the issue on stock Emacs. When I go report it do I just mention neotree and shell-pop-full-span?

[bmag]

You can use the "description", "reproduction guide", "observed behaviour" and "expected behaviour" sections from our bug-report template (example). You should definitely mention neotree and shell-pop-full-span, and maybe also link to my earlier comment.

[jupl]

kyagi/shell-pop-el#50

[jupl]

Did a commit fix this? I just tried develop latest and it appears to be working now in Spacemacs. If so I'll take down the upstream issue. Upstream issue still apparent.

[lotuc]

Encountering the same problem, if I want to show my shell by SPC-' I need to close the neotree first. So pity they cannot show the same time, really annoying.

[d12frosted]

Shouldn't it be reported to neotree instead of shell-pop? Especially after this comment?