felis-terminal/felis

Watch

Tell an ST-terminated OSC from an interrupted one #110

Merged

natsukium merged 1 commit from vt/osc-terminator into main

2026-10-12 15:11:42 +09:00

natsukium commented

2026-10-12 14:19:11 +09:00

Owner

Second step of #67. The parser dispatched an OSC body as soon as an ESC arrived in the string, so a sink could not tell ESC ] … ESC \ from ESC ] … ESC [0m. OSC 7501 may apply a report only once BEL or ST actually ended it, so the grid wiring needs that distinction.

What changes

Sink::osc_dispatch(body, bell_terminated: bool) becomes osc_dispatch(body, end: OscEnd) with OscEnd::{Bel, St, Interrupted}.
An ESC that leaves the OSC string holds the body until the next byte arrives. \ dispatches the body as St, and any other byte (including ESC, CAN, SUB, or a C0 control) dispatches it as Interrupted. That byte is then processed exactly as before, so sinks see events in the same order.
CAN/SUB inside the body still dispatch an empty body, now marked Interrupted.
The grid ignores end, just as it ignored bell_terminated, so every OSC applies as before. The only difference is timing: an OSC ended by a lone ESC dispatches when the following byte arrives, not at the ESC itself.
The new osc_pending flag is carried in the parser's upgrade dump. check_restored rejects a pending OSC outside Escape. The dump has not shipped in a release, so DUMP_VERSION stays the same.

This is an API break for out-of-tree Sink implementors.

Tests

Unit tests:
- an OSC interrupted by a CSI or an ESC dispatch, or by a second OSC
- a doubled ESC
- CAN/SUB or a C0 control after the ESC
- an ESC at a chunk end, completed by the next chunk as ST or as an interruption
- consecutive ST-terminated OSCs
- an ESC outside any OSC, which dispatches no OSC
New property osc_dispatch_reports_how_each_osc_ended: a random stream of OSCs, each with a random ending, fed at random chunk boundaries, must yield exactly the expected (body, end) list.
The state-dump test that resumes a parser stopped mid-sequence now includes an interrupted OSC, split at every byte.
cargo-mutants on the diff: every mutant is caught except transition_to's !matches!(next, State::OscString) guard, which is the equivalent mutant already documented above that match.
just check passes.

Second step of #67. The parser dispatched an OSC body as soon as an ESC arrived in the string, so a sink could not tell `ESC ] … ESC \` from `ESC ] … ESC [0m`. OSC 7501 may apply a report only once BEL or ST actually ended it, so the grid wiring needs that distinction. ## What changes - `Sink::osc_dispatch(body, bell_terminated: bool)` becomes `osc_dispatch(body, end: OscEnd)` with `OscEnd::{Bel, St, Interrupted}`. - An ESC that leaves the OSC string holds the body until the next byte arrives. `\` dispatches the body as `St`, and any other byte (including ESC, CAN, SUB, or a C0 control) dispatches it as `Interrupted`. That byte is then processed exactly as before, so sinks see events in the same order. - CAN/SUB inside the body still dispatch an empty body, now marked `Interrupted`. - The grid ignores `end`, just as it ignored `bell_terminated`, so every OSC applies as before. The only difference is timing: an OSC ended by a lone ESC dispatches when the following byte arrives, not at the ESC itself. - The new `osc_pending` flag is carried in the parser's upgrade dump. `check_restored` rejects a pending OSC outside `Escape`. The dump has not shipped in a release, so `DUMP_VERSION` stays the same. This is an API break for out-of-tree `Sink` implementors. ## Tests - Unit tests: - an OSC interrupted by a CSI or an ESC dispatch, or by a second OSC - a doubled ESC - CAN/SUB or a C0 control after the ESC - an ESC at a chunk end, completed by the next chunk as ST or as an interruption - consecutive ST-terminated OSCs - an ESC outside any OSC, which dispatches no OSC - New property `osc_dispatch_reports_how_each_osc_ended`: a random stream of OSCs, each with a random ending, fed at random chunk boundaries, must yield exactly the expected `(body, end)` list. - The state-dump test that resumes a parser stopped mid-sequence now includes an interrupted OSC, split at every byte. - cargo-mutants on the diff: every mutant is caught except `transition_to`'s `!matches!(next, State::OscString)` guard, which is the equivalent mutant already documented above that match. - `just check` passes.

natsukium added 1 commit

2026-10-12 14:19:11 +09:00

vt: tell an ST-terminated OSC from one an ESC interrupted

fuzz / cargo fuzz nightly long-run (pull_request) Has been skipped

Details

darwin / detect relevant changes (pull_request) Successful in 8s

Details

darwin / darwin docs-only no-op (pull_request) Has been skipped

Details

darwin / build felis (aarch64-darwin) (pull_request) Successful in 1m9s

Details

windows / detect relevant changes (pull_request) Successful in 18s

Details

windows / windows docs-only no-op (pull_request) Has been skipped

Details

windows / cargo nextest (Windows) (pull_request) Successful in 7m28s

Details

bench / instruction-count regression gate (pull_request) Successful in 1m4s

Details

fuzz / cargo fuzz smoke (per target) (pull_request) Successful in 2m55s

Details

pr / nix flake check (pull_request) Successful in 51s

Details

pr / cargo build / clippy / test / deny (pull_request) Successful in 3m19s

Details

pr / wire schema is compatible with the base (pull_request) Successful in 3s

Details

windows / frontend smoke (Windows) (pull_request) Successful in 1m9s

Details

windows / package felis (x86_64-pc-windows-msvc) (pull_request) Has been skipped

Details

pr / frontend smoke (x86_64-linux) (pull_request) Successful in 50s

Details

windows / cargo clippy (Windows cross) (pull_request) Successful in 26s

Details

f8583ebd60

The parser dispatched an OSC body the moment an ESC arrived in the
string, so a sink saw `ESC ] … ESC [0m` exactly as it saw
`ESC ] … ESC \`. OSC 7501 program status may apply a report only once a
real terminator arrived; the sink needs to know which one it got.

`Sink::osc_dispatch` now takes an `OscEnd` (Bel, St, Interrupted) in
place of `bell_terminated`. An ESC leaving the OSC string holds the
body until the next byte: `\` dispatches it as St, anything else as
Interrupted, and that byte is then processed as before, so the event
order a sink sees is unchanged. CAN/SUB inside the body still dispatch
an empty body, now marked Interrupted.

The grid ignores the ending as it ignored `bell_terminated`, so every
OSC is applied as before; only the timing moves: an OSC ended by a lone
ESC dispatches when the byte after that ESC arrives, not at the ESC.
The pending flag rides the parser's upgrade dump, which has not shipped
in a release yet, so the dump version stays.

Assisted-by: Claude Code

natsukium merged commit 3ce6b6719b into main

2026-10-12 15:11:42 +09:00

natsukium deleted branch vt/osc-terminator

2026-10-12 15:11:42 +09:00

natsukium referenced this pull request from a commit

2026-10-12 15:11:43 +09:00

Merge pull request 'Tell an ST-terminated OSC from an interrupted one' (#110) from vt/osc-terminator into main

natsukium referenced this pull request

2026-10-12 15:24:18 +09:00

Queue OSC 7501 reports and prompt starts as ordered grid effects #115