When I use Spacemacs with the Treemacs feature inside it, like the picture, I opened it by Alt Shift m + p + t
Then use Ctrl x + u to open undo-tree
When I move the cursor on the undo-tree by Ctrl + p, the layout had been resized
Even I close the undo-tree window, the layout didn't back to the first picture.
Why it happened? How to avoid it?
The issue is caused by the function undo-tree-visualizer-update-diff, which contains a (balance-windows) command at the end. So the simple dirty trick is to re-define the function with that command commented out after loading undo-tree:
(with-eval-after-load 'undo-tree
;; [2023-10-20 Fri] modify this function from undo-tree.el to prevent
;; re-balancing windows ------------------------------------------------------
(defun undo-tree-visualizer-update-diff (&optional node)
;; update visualizer diff display to show diff between current state and
;; NODE (or previous state, if NODE is null)
(with-current-buffer undo-tree-visualizer-parent-buffer
(undo-tree-diff node))
(let ((win (get-buffer-window undo-tree-diff-buffer-name)))
(when win
;; (balance-windows); comment this line out
(shrink-window-if-larger-than-buffer win))))
)
~/.emacs.d/init.el.
Commented
Aug 28 at 18:06
undo-treeand am unaware of any code that resizes windows. It only usesswitch-to-buffer-other-window,switch-to-bufferanddisplay-buffer. The culprit may possibly lie elsewhere.treemacs, but did a quick git clone and grep forshrinkandenlargeand came up with the functiontreemacs--set-width. In your debugging quest, consider copyingtreemacs--set-widthto a*scratch*buffer and modify it by making it do nothing. e.g., comment out everything and then evaluate the modified function; e.g., the new function that could be evaluated looks like(defun treemacs--set-width (width) "Doc-string." nil)You could even just typeM-x eval-expression RET (defun treemacs--set-width (width) "Doc-string." nil) RET; and, then repeat your test.shrinkandenlargein their names are most likely responsible (with a few unlikely exceptions, e.g., where analistto adisplay-bufferfamily of functions is expressly set with a particular width). Once you track down the function responsible, then you can work your way backwards to where the function is called, and then devise a plan to deal with the situation. You could even just put in some messages in the function mentioned in my previous comment to see when it gets called ...