grid: disarm a pending zero-width joiner on the C0 cursor controls #414

Merged

natsukium merged 2 commits from grid/zwj-c0-disarm into main 2026-09-25 05:41:27 +09:00

Conversation 0 Commits 2 Files changed 6 +115 -4

natsukium commented 2026-09-25 04:27:35 +09:00

Owner

Copy link

Closes #412

A pending zero-width joiner (Grid::zwj_pending, armed by the ZWJ fold and consumed by the next print) now disarms on the C0 controls that move the cursor: BS, HT, LF, VT, FF, CR. Before, when one of them put the cursor right behind a cluster ending in a joiner, the next emoji folded into that cluster:

• BS onto the cluster's spacer;
• HT pinned at the right edge right behind the cluster;
• CR parking at a DECLRMM left margin right behind the cluster;
• LF / VT / FF landing behind such a cluster on the next row.

Which controls

kitty's draw_control_char rebuilds its segmentation state at exactly BS, HT, LF/VT/FF and CR (from the cell left of the new cursor) and keeps it across BEL. felis keeps the GB11 join stream-scoped rather than re-deriving it from cells, so the equivalent is to disarm at those controls. BEL stays transparent: it leaves the cursor at the joiner (pinned by bell_keeps_a_pending_zwj_armed). CSI sequences are out of scope: SGR between a joiner and the next base keeps joining, as in kitty and ghostty.

Both print paths

The per-byte and batched paths share Sink::execute for controls, so the one reset covers both. cursor_controls_disarm_a_pending_zwj_on_both_paths runs the same streams through the #410 per-byte-vs-batched oracle at chunk sizes 1 and unbounded.

The #410 proptest was not extended: its oracle compares the two print paths, which share execute, so it cannot see this class. Each control has a deterministic test instead (crates/felis-grid/src/tests.rs), each verified to fail without the fix.

Docs

CHANGELOG "Fixed" entry; one sentence in docs/explanation/data-model/grid-and-cells.md "Building a cluster" on how far the continuation reaches.

Closes #412 A pending zero-width joiner (`Grid::zwj_pending`, armed by the ZWJ fold and consumed by the next print) now disarms on the C0 controls that move the cursor: BS, HT, LF, VT, FF, CR. Before, when one of them put the cursor right behind a cluster ending in a joiner, the next emoji folded into that cluster: - BS onto the cluster's spacer; - HT pinned at the right edge right behind the cluster; - CR parking at a DECLRMM left margin right behind the cluster; - LF / VT / FF landing behind such a cluster on the next row. ## Which controls kitty's `draw_control_char` rebuilds its segmentation state at exactly BS, HT, LF/VT/FF and CR (from the cell left of the new cursor) and keeps it across BEL. felis keeps the GB11 join stream-scoped rather than re-deriving it from cells, so the equivalent is to disarm at those controls. BEL stays transparent: it leaves the cursor at the joiner (pinned by `bell_keeps_a_pending_zwj_armed`). CSI sequences are out of scope: SGR between a joiner and the next base keeps joining, as in kitty and ghostty. ## Both print paths The per-byte and batched paths share `Sink::execute` for controls, so the one reset covers both. `cursor_controls_disarm_a_pending_zwj_on_both_paths` runs the same streams through the #410 per-byte-vs-batched oracle at chunk sizes 1 and unbounded. The #410 proptest was not extended: its oracle compares the two print paths, which share `execute`, so it cannot see this class. Each control has a deterministic test instead (`crates/felis-grid/src/tests.rs`), each verified to fail without the fix. ## Docs CHANGELOG "Fixed" entry; one sentence in `docs/explanation/data-model/grid-and-cells.md` "Building a cluster" on how far the continuation reaches.

natsukium added 1 commit 2026-09-25 04:27:35 +09:00

grid: disarm a pending zero-width joiner on the C0 cursor controls ... 560afbdb0c

The GB11 continuation is armed by the ZWJ fold and consumed by the
next print, but BS, HT, LF, VT, FF and CR left it armed. When the
control put the cursor right behind a cluster ending in a joiner (BS
onto the cluster's spacer, HT pinned at the right edge, CR parking at
a left margin, LF landing behind such a cluster on the next row), the
next emoji folded into that cluster instead of starting its own cell.
Whether it joined depended on the joiner having been the last print
before the control, not on anything the grid shows.

kitty rebuilds its segmentation state at exactly these controls from
the cell left of the new cursor. felis keeps the join stream-scoped
rather than re-deriving it from cells, so the equivalent here is to
disarm. BEL stays transparent, as in kitty: it leaves the cursor at
the joiner. Both print paths share `execute`, so one reset covers
them; the batched-versus-per-byte proptest cannot see this class, so
each control gets a deterministic test.

Assisted-by: Claude Code

natsukium added 1 commit 2026-09-25 04:49:47 +09:00

ci: rerun checks that failed under runner load ... 9433bb3222

The Criterion gate and one daemon serve test failed while the shared
runner host sat at load 12-19; neither touches the changed code path.

Assisted-by: Claude Code

natsukium merged commit a53cd63eb9 into main 2026-09-25 05:41:27 +09:00

natsukium deleted branch grid/zwj-c0-disarm 2026-09-25 05:41:27 +09:00

natsukium referenced this pull request from a commit 2026-09-25 05:41:29 +09:00

grid: disarm a pending zero-width joiner on the C0 cursor controls (#414)

natsukium referenced this pull request 2026-09-25 07:53:26 +09:00

bench: follow up the 9d1d842f cross-terminal run #405

Commenting is not possible because the repository is archived.

Reviewers

No reviewers

Labels

Clear labels   

priority/P0

Must be resolved before the v0.1.0 tag

  

priority/P1

Must be settled before the first public contract freezes

  

priority/P2

Explicitly deferred until after v0.1.0

  

release/v0.1.0

Work tracked for the first tagged release

  

status/blocked

Cannot start now: waits on an upstream release, another issue, or a recorded revisit trigger

  

status/planned

Triaged with a source-grounded plan and a passed pi review; ready to implement

  

type/bug

Shipped behavior that contradicts the docs or the code's own intent; needs a product code change

  

type/design

An open scope or trade-off question; needs a recorded decision in the owning explanation doc before code

  

type/test-gap

Documented or reviewed behavior that no test pins; usually split from a review finding, needs no product code change

  

type/tracker

Parent issue that tracks child work

No labels

priority/P0

priority/P1

priority/P2

release/v0.1.0

status/blocked

status/planned

type/bug

type/design

type/test-gap

type/tracker

Milestone

Clear milestone

No items

No milestone

Projects

Clear projects

No items

No project

Assignees

Clear assignees

No assignees

1 participant

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

natsukium/felis!414

No description provided.