Prompt redraw (OSC133) leaves lines from old prompt

[thinety]

Issue Description

If the prompt wraps when resizing the terminal, extra lines from the old prompt are left in the scrollback history after redrawing.

I can confirm that this bug was introduced by dde3d4d. Before, the prompt was cleared before resizing pages, and thus before wrapping text. From reading the commit, I believe the new order of operations makes it easier to recover from a failed resize, and so I'm not sure how to proceed. I'm happy to contribute a fix after we settle on a solution.

Expected Behavior

No extra lines are left in the scrollback history. Example from commit 89b103d (parent of commit dde3d4d that introduced the bug):

Screencast.From.2026-07-26.03-04-10.mp4

Actual Behavior

Extra lines are being left in the scrollback history.

Screencast.From.2026-07-26.03-03-43.mp4

Reproduction Steps

1. Open a fish session in Ghostty
2. Setup a long prompt
3. Resize the terminal such that the prompt wraps
4. Notice the extra lines from the old prompt that are left in the scrollback history

Ghostty Logs

No response

Ghostty Version

Ghostty 1.3.2-main-+66fed652a

Version
  - version: 1.3.2-main-+66fed652a
  - channel: tip
Build Config
  - Zig version   : 0.16.0
  - build mode    : .Debug
  - app runtime   : .gtk
  - font engine   : .fontconfig_freetype
  - renderer      : renderer.generic.Renderer(renderer.OpenGL)
  - libxev        : io_uring
  - kernel version: 7.1.4-arch1-1
  - desktop env   : gnome
  - GTK version   :
    build         : 4.22.4
    runtime       : 4.22.4
  - libadwaita    : enabled
    build         : 1.9.2
    runtime       : 1.9.2
  - libX11        : enabled
  - libwayland    : enabled

OS Version Information

Arch Linux

(Linux only) Display Server

Wayland

(Linux only) Desktop Environment/Window Manager

GNOME 50.3

Minimal Ghostty Configuration

# default configuration

Additional Relevant Configuration

• fish, version 4.8.1

I acknowledge that:

• I have reviewed the FAQ and confirm that my issue is NOT among them.
• I have searched the Ghostty repository (both open and closed Discussions and Issues) and confirm this is not a duplicate of an existing issue or discussion.
• I have checked the "Preview" tab on all text fields to ensure that everything looks right, and have wrapped all configuration and code in code blocks with a group of three backticks (```) on separate lines.

[farfromexact]

The bisect to dde3d4d6b is very useful, and with the default configuration this looks like a Ghostty regression rather than a fish or OSC 133 configuration problem. The clean behavior in the parent 89b103dd5 gives you a good regression oracle.

I would approach the fix in two small steps:

1. Add a regression test for an OSC 133 prompt that wraps, followed by a resize, and assert that the old prompt cells are not left in scrollback.
2. Adjust the resize/redraw ordering so the invalidated prompt is cleared before the wrapped rows are committed, while retaining the newer failure-recovery behavior that motivated dde3d4d6b.

That test should use the same terminal-screen path as the real resize event rather than only testing a helper in isolation. It should also cover both a prompt that fits on one row and one that wraps, because the bug only appears in the latter case. I would not look for a user-facing config workaround first; the report already has a minimal reproduction and a source-level regression point. A focused PR with the test plus the smallest ordering fix would be the right next step.