fix: route dead-key compose through IMContextSimple on Wayland#90
Merged
am-will merged 1 commit intoJul 6, 2026
Merged
Conversation
manuacl
added a commit
to manuacl/limux
that referenced
this pull request
May 25, 2026
manuacl
added a commit
to manuacl/limux
that referenced
this pull request
May 25, 2026
manuacl
force-pushed
the
fix/issue-89-dead-key-compose-fallback
branch
2 times, most recently
from
860d775 to
cb86a45
Compare
manuacl
added a commit
to manuacl/limux
that referenced
this pull request
May 25, 2026
manuacl
force-pushed
the
fix/issue-89-dead-key-compose-fallback
branch
from
cb86a45 to
2a2aca8
Compare
GTK4's IMMulticontext defaults to the "wayland" slave on Plasma 6
Wayland sessions that don't have ibus or fcitx5 running. That slave
claims dead-key events over text-input-v3 (filter_keypress returns
true and a preedit_changed signal fires with the bare dead-key glyph),
but KWin/Plasma 6 never delivers the follow-up commit. The dead-key
glyph flashes on screen, the follow-up key is consumed too, and the
composed character never appears. AZERTY users get raw `^e` instead
of `ê`.
Pair the existing IMMulticontext with a GtkIMContextSimple fallback
that drives libxkbcommon's compose tables in-process. When the
incoming keysym is a compose initiator (XK_dead_*, Multi_key) or the
fallback is already mid-compose, bypass the multicontext entirely so
KWin can't intercept the event; otherwise the multicontext keeps
priority so ibus / fcitx5 / CJK IMEs continue to work.
The IME plumbing lives in a new limux-host-linux::ime module split
into three submodules so pure logic is testable in isolation from
the GTK wiring:
ime/state.rs TerminalImeState machine + 6 unit tests
ime/routing.rs decide_routing, is_compose_initiator,
update_latch_after_fallback_first + 8 unit tests
ime/contexts.rs PaneIme, GTK signal wiring, ghostty FFI helpers
terminal.rs sheds ~270 lines. Regression tests pin the routing rules
(initiator and in-flight compose route to fallback first; plain keys
route to the multicontext first; the latch arms / disarms / stays
put under the expected fallback responses) and include a full
state-machine trace of the `^` + `e` → `ê` sequence.
A CLAUDE.md anchor entry points at the new module and a Pitfalls
entry warns against routing dead-key events through IMMulticontext
alone on Wayland.
Builds on the partial fix from PR am-will#20 (commit 536e6e9), which wired
IMMulticontext but did not cover the Wayland-default case.
Fixes am-will#89
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
manuacl
force-pushed
the
fix/issue-89-dead-key-compose-fallback
branch
from
2a2aca8 to
730ca0a
Compare
Owner
|
Thank you for the excellent write-up and implementation here. I landed this on The merged fix preserves your core approach: Could you please test the next Limux release on your Plasma 6 Wayland/AZERTY setup and confirm the dead-key cases are fixed? Thanks again for chasing this down. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
IMMulticontext'swaylandslave claims AZERTY dead-key events (^,¨, dead~) overtext-input-v3and emits apreedit_changedwith the bare glyph, but KWin never delivers the follow-upcommit— the dead-key glyph flashes on screen and the compose silently aborts (Dead-key compose (^/¨/dead ~) silently fails inside terminal panes on Plasma 6 Wayland #89).IMMulticontextwith a parallelGtkIMContextSimpleand bypass the multicontext entirely for compose initiators (XK_dead_*,Multi_key) and in-flight compose sequences, so libxkbcommon's compose tables run in-process and KWin can't intercept; otherwise the multicontext keeps priority so ibus / fcitx5 / CJK IMEs continue to work.limux-host-linux::imemodule (TerminalImeState, factories, pure-logic routing helpers, 14 unit tests including 8 regression tests pinning the routing rules);terminal.rssheds ~270 lines. Builds on PR fix: add IMContext for dead-key and compose-key support #20 / commit 536e6e9, which wiredIMMulticontextbut didn't cover the Wayland-default case.Fixes #89.
Test plan
cargo fmt --check— greencargo clippy --workspace --all-targets -- -D warnings— greencargo test -p limux-host-linux— 190/190 pass, including 14ime::tests::*(5 state-machine traces, 4 routing-decision regressions, 3 latch-transition regressions, 1 compose-initiator detection table, 1 full^+e→êend-to-end trace)^+e→êinside a terminal pane on Bazzite Plasma 6 Wayland (the reproducer from Dead-key compose (^/¨/dead ~) silently fails inside terminal panes on Plasma 6 Wayland #89)Shift+^(¨) +e→ëecho hello)Ctrl+C/Ctrl+Fshortcuts still reach their handlers; pressing^alone does not open the search bar (the regression PR fix: add IMContext for dead-key and compose-key support #20 fixed)🤖 Generated with Claude Code