Keep a multi-row text-sizing character whole or erase it #33

Merged

natsukium merged 1 commit from fix/multirow-block-shift into main 2026-10-06 20:39:51 +09:00

Conversation 1 Commits 1 Files changed 7 +517 -4

natsukium commented 2026-10-06 19:32:51 +09:00

Owner

Copy link

Closes #24.

A text-sizing (OSC 66) character taller than one row is drawn from its primary over its whole footprint. Any cell move that carried some of its rows and not the others tore it: ICH or insert mode on one of its rows, DECIC/SR/SU in a region whose margin crossed it, IL/DL through it. The primary then drew over other text, or rows were left with no primary.

Before each move, a block that a row seam of the move splits is now erased whole, as kitty's nuke_multiline_char_intersecting_with does. A scroll into scrollback keeps the departing rows in history and clears only the rows left on screen.

A new property test over random moves and OSC 66 writes found three more ways to leave part of a block behind, fixed here too:

• an OSC 66 write over another block's lower rows erased only the cells it covered;
• SL/SR, DECIC/DECDC and DECBI/DECFI cut column seams row by row after earlier rows had moved;
• insert mode left a copy of the shifted cell under the cursor, which erased the moved sized character.

Tests: a table test whose every cut case fails on main; scrollback and alternate-screen cases; the property test (passed 4×3000 cases). Docs: reference and explanation of kitty-text-sizing.md, CHANGELOG.

Closes #24. A text-sizing (OSC 66) character taller than one row is drawn from its primary over its whole footprint. Any cell move that carried some of its rows and not the others tore it: ICH or insert mode on one of its rows, DECIC/SR/SU in a region whose margin crossed it, IL/DL through it. The primary then drew over other text, or rows were left with no primary. Before each move, a block that a row seam of the move splits is now erased whole, as kitty's `nuke_multiline_char_intersecting_with` does. A scroll into scrollback keeps the departing rows in history and clears only the rows left on screen. A new property test over random moves and OSC 66 writes found three more ways to leave part of a block behind, fixed here too: - an OSC 66 write over another block's lower rows erased only the cells it covered; - SL/SR, DECIC/DECDC and DECBI/DECFI cut column seams row by row after earlier rows had moved; - insert mode left a copy of the shifted cell under the cursor, which erased the moved sized character. Tests: a table test whose every cut case fails on main; scrollback and alternate-screen cases; the property test (passed 4×3000 cases). Docs: reference and explanation of `kitty-text-sizing.md`, CHANGELOG.

natsukium added 1 commit 2026-10-06 19:32:51 +09:00

grid: keep a multi-row OSC 66 block whole or erase it ... b290141b4f

A text-sizing block taller than one row is drawn from its primary over
its whole footprint. Any move that carried some of its rows and not the
others tore it: ICH on its top row shifted that row only, and DECIC, SR
or SU in a region whose margin crossed it, or IL and DL through it, did
the same between rows. The primary then drew over cells that held other
text, or rows were left with no primary at all. The earlier fix covered
a move's column boundaries only.

Before each move, a block that a row seam of the move splits is now
erased, as kitty's nuke_multiline_char_intersecting_with does. The seams
are the cursor row for a row-local shift, the region's top and bottom
for a column shift, and those plus the boundary between departing and
staying rows for a vertical scroll. A scroll into scrollback keeps the
departing rows of the block in history and clears only the rows left on
screen, so the character's text survives; when both the history seam and
the bottom seam cut one block, history wins.

A property test over random moves and OSC 66 writes found three more
ways to leave part of a block behind, fixed here too:

- An OSC 66 write cleared foreign sized cells only on its top row; its
  lower rows now erase a block they land on whole.
- SL/SR, DECIC/DECDC and DECBI/DECFI cut each row's column seams after
  the rows above had moved, so a moved row straddled the next row's seam
  and erased the block's remaining cells only there. All seams are now
  cut before any row moves.
- IRM left the old cell under the cursor after shifting the row; read as
  a sized primary it erased the moved original. The vacated span is now
  blanked, as ICH does.

The seam scan is gated on has_sized_cells and reads each seam row as one
slice: a scroll benchmark shows no change without sized text and a few
percent once an OSC 66 write has set the flag.

Closes #24

Assisted-by: Claude Code

natsukium commented 2026-10-06 19:45:29 +09:00

Author

Owner

Copy link

The Criterion gate flagged daemon_side/shell_prompt (+15.6%) and multi_cycle_end_to_end/plaintext_scroll_individual (+29.9%). A local A/B against main (separate target dirs, base and head alternated, three rounds) shows no regression:

bench	main (3 runs)	this branch (3 runs)
daemon_side/shell_prompt	36.1 / 35.4 / 35.4 µs	34.4 / 35.3 / 34.7 µs
multi_cycle_end_to_end/plaintext_scroll_individual	1.25 / 2.31 / 2.54 ms	1.15 / 1.21 / 1.44 ms
multi_cycle_end_to_end/plaintext_scroll_batched	1.44 / 2.44 / 1.12 ms	1.39 / 1.25 / 1.57 ms

The multi-cycle benches are bimodal on this host (1.1–2.5 ms on main alone); the seam scan returns before any work when no OSC 66 cell has been written.

The Criterion gate flagged `daemon_side/shell_prompt` (+15.6%) and `multi_cycle_end_to_end/plaintext_scroll_individual` (+29.9%). A local A/B against main (separate target dirs, base and head alternated, three rounds) shows no regression: | bench | main (3 runs) | this branch (3 runs) | | --- | --- | --- | | `daemon_side/shell_prompt` | 36.1 / 35.4 / 35.4 µs | 34.4 / 35.3 / 34.7 µs | | `multi_cycle_end_to_end/plaintext_scroll_individual` | 1.25 / 2.31 / 2.54 ms | 1.15 / 1.21 / 1.44 ms | | `multi_cycle_end_to_end/plaintext_scroll_batched` | 1.44 / 2.44 / 1.12 ms | 1.39 / 1.25 / 1.57 ms | The multi-cycle benches are bimodal on this host (1.1–2.5 ms on main alone); the seam scan returns before any work when no OSC 66 cell has been written.

natsukium force-pushed fix/multirow-block-shift from b290141b4f to 0dbe23dab7 2026-10-06 19:46:08 +09:00 Compare

natsukium force-pushed fix/multirow-block-shift from 0dbe23dab7 to d0584f663f 2026-10-06 19:53:09 +09:00 Compare

natsukium force-pushed fix/multirow-block-shift from d0584f663f to dfb85a51a9 2026-10-06 20:12:46 +09:00 Compare

natsukium merged commit b6257284f2 into main 2026-10-06 20:39:51 +09:00

natsukium deleted branch fix/multirow-block-shift 2026-10-06 20:39:51 +09:00

natsukium referenced this pull request from a commit 2026-10-06 20:39:52 +09:00

Merge pull request 'Keep a multi-row text-sizing character whole or erase it' (#33) from fix/multirow-block-shift into main

Sign in to join this conversation.

Reviewers

No reviewers

Labels

Clear labels   

Area/ci

CI workflows (.forgejo/workflows/)

  

Area/cli

felis-cli crate, the felis front-door binary

  

Area/client

felis-client crate, the winit GUI frontend

  

Area/client-core

felis-client-core crate, shadow screen, config, connect_or_spawn_daemon

  

Area/daemon

felis-daemon crate, session pool, IPC server

  

Area/deps

Dependency bumps

  

Area/docs

Documentation tree when no narrower area fits

  

Area/grid

felis-grid crate, cells, scrollback, image store

  

Area/manual

User-facing docs: tutorials, how-to, user-facing reference pages

  

Area/nix

Nix flake, home-manager module, package definitions

  

Area/protocol

felis-protocol crate, felis.proto, frame layout, version constants

  

Area/pty

felis-pty crate, PTY backends

  

Area/render

felis-render-wgpu crate, GPU pipeline, atlas

  

Area/shaping

felis-shaping crate, swash shaper, font fallback

  

Area/skills

Agent skills (skills/felis, .agents/skills/)

  

Area/transport

felis-transport crate, framing, socket/pipe, SSH carrier

  

Area/vt

felis-vt crate, VT parser, dispatch

  

Compat/Breaking

Breaking change that won't be backward compatible

  

Kind/Bug

Something is not working

  

Kind/Documentation

Documentation changes

  

Kind/Enhancement

Improve existing functionality

  

Kind/Feature

New functionality

  

Kind/Performance

Throughput, latency, or memory cost

  

Kind/Security

This is security issue

  

Kind/Testing

Issue or pull request related to testing

  

perf-regression-accepted

Passes the instruction-count bench gate for an intended cost; state why in a PR comment

  

Reviewed

Confirmed

Issue has been confirmed

  

Reviewed

Duplicate

This issue or pull request already exists

  

Reviewed

Invalid

Invalid issue

  

Reviewed

Won't Fix

This issue won't be fixed

  

Status

Blocked

Something is blocking this issue or pull request

  

Status

Need More Info

Feedback is required to reproduce issue or to continue work

  

Status

Upstream

Waiting on a change in an upstream project

No labels

Area/ci

Area/cli

Area/client

Area/client-core

Area/daemon

Area/deps

Area/docs

Area/grid

Area/manual

Area/nix

Area/protocol

Area/pty

Area/render

Area/shaping

Area/skills

Area/transport

Area/vt

Compat/Breaking

Kind/Bug

Kind/Documentation

Kind/Enhancement

Kind/Feature

Kind/Performance

Kind/Security

Kind/Testing

perf-regression-accepted

Reviewed

Confirmed

Reviewed

Duplicate

Reviewed

Invalid

Reviewed

Won't Fix

Status

Blocked

Status

Need More Info

Status

Upstream

Milestone

Clear milestone

No items

No milestone

Assignees

Clear assignees

No assignees

1 participant

Notifications

Subscribe

Due date

The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference

felis-terminal/felis!33

No description provided.