diff options
| author | Igor Soarez <igor@soarez.org> | 2026-10-25 06:41:34 +0200 |
|---|---|---|
| committer | Igor Soarez <igor@soarez.org> | 2026-10-25 06:41:34 +0200 |
| commit | b77427190408cbe3ebea5677cd67bce53e73f313 (patch) | |
| tree | acffe9de5554747d0ff505e4428a748e2c40ae7d /tests | |
| parent | dd3462c3cd0a7ff8dbb31c5b4e6d133a04506419 (diff) | |
terminals: ask Ghostty only once focus has settled
across a Space switch (a full screen window) tmux sees the new client
focused about 0.3 s before Ghostty's front window changes, so learn paired
the client with the terminal that had focus. live, raising a full screen
window and back crossed the two pairings, and from then on every jump
raised the wrong window (old binary: 0 of 10 jumps right).
learn now waits a second before asking Ghostty, then still requires the
client to be focused. it also drops pairings of clients no longer attached,
which piled up with every closed window.
e2e step 20: Ghostty lagging behind tmux's focus, crossed pairings healing,
pruning; the old learn fails all five. live, two Ghostty windows, one full
screen: 16 of 16 jumps right, starting from crossed pairings.
Diffstat (limited to 'tests')
| -rwxr-xr-x | tests/e2e.sh | 20 |
1 files changed, 20 insertions, 0 deletions
diff --git a/tests/e2e.sh b/tests/e2e.sh index f874c15..13a698f 100755 --- a/tests/e2e.sh +++ b/tests/e2e.sh @@ -737,6 +737,26 @@ r switch-client -c "$TTY1" -t '=elsewhere' renv "$DINGGER" focus --pane "$AG" expect "a window already on the agent's session wins over a more recent one" "tail -1 $DIR/ghostty-raised" BBBB-2222 50 expect "and the other window is not moved" "client_session $TTY1" elsewhere +# Across a Space switch Ghostty still names the terminal that had focus +# for a moment: the other window's. +focus_out viewer; focus_in viewer2 AAAA-1111 +sleep 0.5; echo BBBB-2222 >"$DIR/ghostty-focused" +sleep 1 +expect "Ghostty lagging behind tmux's focus: the right terminal" "paired $TTY2" BBBB-2222 50 +expect "and the other window keeps its own" "paired $TTY1" AAAA-1111 +# Pairings crossed (as the lag used to leave them): Ghostty's lagging +# answer is then the client's own wrong terminal. +r set -g "@dingger_terminal_$(basename "$TTY1")" "1 BBBB-2222 $FAKE_APP" +r set -g "@dingger_terminal_$(basename "$TTY2")" "1 AAAA-1111 $FAKE_APP" +focus_out viewer2; focus_in viewer2 AAAA-1111 +sleep 0.5; echo BBBB-2222 >"$DIR/ghostty-focused" +expect "crossed pairings heal: the focused client's" "paired $TTY2" BBBB-2222 50 +expect "and the other's wrong one is dropped" "paired $TTY1" "" +focus_out viewer2; focus_in viewer AAAA-1111 +expect "until its next focus" "paired $TTY1" AAAA-1111 50 +r set -g @dingger_terminal_ttysgone "1 DDDD-4444 $FAKE_APP" +focus_out viewer; focus_in viewer2 BBBB-2222 +expect "a client no longer attached loses its pairing" "r show -gv @dingger_terminal_ttysgone 2>/dev/null" "" 50 echo viewer >"$DIR/ghostty-race"; : >"$DIR/ghostty-queries" focus_in viewer CCCC-3333 expect "Ghostty asked" "wc -l <$DIR/ghostty-queries | tr -d ' '" 1 50 |
