Signal CWD change to terminal via OSC-7

[dnkl]

The path still needs to be URL encoded, but this shows the gist of it. Tested in foot only so far. Will hopefully be able to try a few more tomorrow.

TODO

• URL encode path
• Test more terminals

Closes #1147

[jarun]

I have merged this PR. I will force push the appropriate changes as well.

Please let me know if you see any issues during your tests.

[jarun]

Please confirm commit 0556ac1 works fine in your tests.

[dnkl]

That commit works fine in both foot, and in gnome-terminal.

Technically, the path needs to be URL encoded. In reality, not doing will work in probably 99% of all cases. I tested switching to a directory named åöä&>, and it worked in gnome-terminal as well as in foot.

I also tested a directory that included a literal ESC, and that did not work. That's expected, since an ESC will typically reset the VT state machine.

It would be fairly simple to implement a URL encoder that allows plain ASCII but %NN encodes all other characters. Such an encoder would be overly eager, compared to what's actually necessary, but would produce "safe" output.

On the other hand, if we want to keep things simple, then let's skip URL encoding. The worst thing that happens is the new terminal doesn't spawn in the correct directory.

[jarun]

I tested switching to a directory named åöä&>, and it worked in gnome-terminal as well as in foot.

That's because by default nnn is linked to ncursesw (with wide char support) so wide chars work as they should under the hood.

[jarun]

BTW, given the restrictions in nnn, no one would agree to have an encoder for a single-line printf() that in fact, creates issues in certain workflows. Practically speaking, there's no ROI.

[dnkl]

Then it's all good from my side.

[jarun]

👍