resize-window doesn't apply given increment accurately

[xeyownt]

Describe the bug
I'm using remote-control and resize-window to build a tab layout with window of given size (using layout splits).
When applying the resize-window the new window sizes are not the one expected.

To Reproduce
Steps to reproduce the behavior:

1. Run the following script in a kitty --config NONE -o allow_remote_control=yes window:

#! /bin/bash

function do-resize() 
{
  INCREMENT=$1
  L1=$(kitty @ ls | jq '.[].tabs[].windows[] | select( .title == "left" ) | .columns')
  R1=$(kitty @ ls | jq '.[].tabs[].windows[] | select( .title == "right" ) | .columns')
  kitty @ resize-window --match title:left --increment $INCREMENT --axis horizontal
  L2=$(kitty @ ls | jq '.[].tabs[].windows[] | select( .title == "left" ) | .columns')
  R2=$(kitty @ ls | jq '.[].tabs[].windows[] | select( .title == "right" ) | .columns')
  echo "Left: $L1 -> $L2 [$((L2 - L1))] Right: $R1 -> $R2 [$((R2 - R1))]"
}
kitty @ launch --type tab --title "left" --tab-title "first" > /dev/null
kitty @ goto-layout --match title:first splits
kitty @ launch --match title:first --location vsplit --title "right" --keep-focus > /dev/null
do-resize 2
do-resize 2
do-resize 2
do-resize 2
do-resize 2

2. This gives the output below. We see that we sometimes get a wrong increment, and sometimes the increment on the left and right window is different.

Left: 106 -> 108 [2] Right: 106 -> 104 [-2]
Left: 108 -> 110 [2] Right: 104 -> 102 [-2]
Left: 110 -> 113 [3] Right: 102 -> 100 [-2]
Left: 113 -> 115 [2] Right: 100 -> 97 [-3]
Left: 115 -> 117 [2] Right: 97 -> 95 [-2]

3. Using an increment of 10, we get the following output, where the increment is always wrong. In fact the bigger the increment the bigger the difference.

Left: 106 -> 117 [11] Right: 106 -> 95 [-11]
Left: 117 -> 128 [11] Right: 95 -> 84 [-11]
Left: 128 -> 139 [11] Right: 84 -> 73 [-11]
Left: 139 -> 151 [12] Right: 73 -> 62 [-11]
Left: 151 -> 162 [11] Right: 62 -> 51 [-11]

4. The same applies for hsplit and --axis vertical.

Environment details

kitty 0.29.1 (a0341af3f8) created by Kovid Goyal
Linux zavcxl0007 6.1.0-6-amd64 #1 SMP PREEMPT_DYNAMIC Debian 6.1.15-1 (2023-03-05) x86_64
Debian GNU/Linux bookworm/sid zavcxl0007 /dev/tty

Running under: X11
Frozen: True
Paths:
  kitty: /home/peetersm/.local/kitty.app/bin/kitty
  base dir: /home/peetersm/.local/kitty.app/lib/kitty
  extensions dir: /home/peetersm/.local/kitty.app/lib/kitty-extensions
  system shell: /bin/bash
Loaded config overrides:
  allow_remote_control yes

Config options different from defaults:
allow_remote_control yes

Important environment variables seen by the kitty process:
	PATH                                /home/peetersm/bin/local:/home/peetersm/bin:/home/peetersm/.local/bin:/usr/local/bin:/usr/bin:/bin:/usr/local/games:/usr/games:/home/peetersm/.fzf/bin
	LANG                                en_US.UTF-8
	EDITOR                              vim
	SHELL                               /bin/bash
	DISPLAY                             :0
	USER                                peetersm
	XDG_MENU_PREFIX                     gnome-
	LC_MONETARY                         en_DK.UTF-8
	XDG_SESSION_DESKTOP                 gnome-xorg
	XDG_SESSION_TYPE                    x11
	LC_PAPER                            en_DK.UTF-8
	XDG_CURRENT_DESKTOP                 GNOME
	XDG_SESSION_CLASS                   user
	LC_MEASUREMENT                      en_DK.UTF-8
	XDG_RUNTIME_DIR                     /run/user/1000
	LC_TIME                             en_DK.UTF-8
	XDG_DATA_DIRS                       /usr/share/gnome:/usr/local/share/:/usr/share/
	LC_NUMERIC                          en_DK.UTF-8

[kovidgoyal]

windows sizes in kitty are based on biases not cell sizes so when you resize in cells you will get a bias that comes close to the asked for size but there is no guarantee it will be exact. That said I am sure the bias calculations could be improved for better consistency, patches are welcome, relevant code is in the layout folder.