v1.4.0

• Replaced the "Make 'Open with Code' open in WSL" installer option with a "Make launching VS Code from Explorer open in WSL" which includes double-clicking on files in addition to the context menu items.

  • Enabling this feature will disable the relevant options in VS Code's own installer to prevent updates from overwriting our changes. Disabling the feature or uninstalling wsl-tools reverts everything back.
  • Now launches VS Code the same way as the code shell script, using the (undocumented) --file-uri & --folder-uri options, as otherwise VS Code tries to determine if a path is a directory or not simply by whether it contains a dot...
  • If wslpath fails, it will now open in a local window instead of erroring (#3)

• recycle --rm is now viable as a shell alias for rm:

  • When run with --rm, files in WSL will be deleted Linux-side instead of going through the winapi. This greatly improves performance when deleting files in Linux and allows for use of sudo.
  • Added an -r/--recursive option, without which --rm will error if any path is a directory.
    • Also added a check to prevent deleting /, like rm.
  
  
  Benchmark versus rm and previous version:

  hyperfine \
    --export-markdown benchmark.md \
    --sort mean-time \
    --warmup 1 \
    --runs 5 \
    --shell none \
    -L file /mnt/c/Users/max/foo,/home/max/foo \
    -p 'touch {file}' \
    -C 'sleep 0.1s' \
    'rm {file}' \
    'v1.3.0/recycle --rm {file}' \
    'v1.4.0/recycle --rm {file}'
  

  Command	Mean [ms]	Min [ms]	Max [ms]	Relative
  🗑️ rm /home/max/foo	0.8 ± 0.1	0.7	0.9	1.00
  🗑️ rm /mnt/c/Users/max/foo	1.9 ± 0.1	1.8	2.0	2.42 ± 0.30
  🗑️ v1.4.0/recycle --rm /home/max/foo	6.8 ± 0.2	6.6	7.1	8.91 ± 1.00
  ♻️ v1.4.0/recycle --rm /mnt/c/Users/max/foo	71.2 ± 0.5	70.6	71.8	92.75 ± 10.08
  ♻️ v1.3.0/recycle --rm /mnt/c/Users/max/foo	71.8 ± 1.4	70.8	74.3	93.55 ± 10.32
  🗑️ v1.3.0/recycle --rm /home/max/foo	73.2 ± 1.7	71.6	75.5	95.37 ± 10.59

• When running the Windows exe, use an absolute path instead of relying on PATH. (This commit was cherry-picked onto v1.3.0 in order to run the above benchmark.)

• Enabled dark mode in installer