Two more fixes based on Michael K. Darling's branch
(github.com/darlingm/tmux, pr5516-regression-fixes), taken as-is -
neither is caught by any test in regress/ yet, found by code review
rather than a failing test:
- screen_write_redraw_cb() (screen-write.c) reported damage for only a
single row, using ttyctx->ocy as if every fallback redraw were a
single-cell write. But it's also the callback for cases that can
legitimately span many rows - a large scroll-region fallback
(tty_redraw_region(), when tty_large_region() or the pane is
obscured), a full reset, and entering/leaving the alternate screen.
For those, only the top row of the affected area ever got marked as
damaged, leaving the rest stale until an unrelated redraw happened to
cover it. Changed the shared tty_ctx_redraw_cb typedef to carry
(py, ny) - the actual row range - and updated every call site to pass
the range it actually knows about, instead of hardcoding a single
row.
- window_pane_redraw_floating() (window.c) never refreshed the status
line after moving/resizing a floating pane, so a status format
depending on that pane's geometry (e.g. #{pane_width}) could go
stale until an unrelated status refresh happened. Added a
server_status_window(w) call.
Also confirmed the window_pane_scrollbar_intersects() parameter
naming cleanup (loop -> wp) discussed earlier was already done in an
earlier "Cleanup." commit - nothing left to do there.
All 9 tests in regress/ plus the two pre-existing floating-pane tests
plus a further 19-test sweep of redraw/tty/input/sync-adjacent
regress tests pass.
Co-Authored-By: Michael K. Darling <darlingm@gmail.com>
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Fixes the four bugs caught by the regression tests added in aed1209c
(popup-drag-status-line.sh, popup-drag-wide-character.sh,
popup-drag-pane-prompt.sh, switch-client-redraw.sh), based on fixes
from Michael K. Darling (github.com/darlingm/tmux, pr5516-regression-
fixes), reviewed and adapted:
- server-client.c: server_client_set_session()'s check for whether the
client's window actually changed compared old->curw to s->curw, but
when old == s these read the same, already-updated field, so a
same-session window switch was never detected. Compare against the
client's own cached redraw scene instead (redraw_client_has_window(),
new in screen-redraw.c/tmux.h). Taken from darlingm as-is.
- popup.c: popup_damage() only translated a popup's client-coordinate
rectangle into window coordinates, so a popup dragged across the
status line never triggered a status-line redraw once it moved away -
status_redraw()'s own "skip if content unchanged" optimization
suppressed it, since only the popup moved, not the status content.
Now detects overlap with the status line and forces a redraw via the
existing (previously unused) CLIENT_REDRAWSTATUSALWAYS flag, and
properly clips the reported rectangle to the pane area for
status-at-top/bottom/off. Taken from darlingm as-is.
- screen-redraw.c: redraw_draw_damage_rect() clipped a span to a damage
rectangle's raw geometric edges, which have no idea what's in the
grid, so a clip edge could land mid-character and tear a wide
character in half. Added redraw_damage_grow_span_clip(): widen the
clip by one cell on each edge that isn't already at the span's own
boundary. Reimplemented simpler than darlingm's version (which walked
grid cells per span type via a switch and direct grid lookups) -
since no grid cell is ever wider than two columns, an unconditional
one-cell margin is always enough to pull a split character back in,
with no need to inspect grid content at all.
- screen-redraw.c: redraw_draw_damage_rect() also never re-overlaid a
pane's active in-pane prompt after drawing its underlying content, so
damage crossing a prompt row erased it until an unrelated redraw
restored it. Factored the existing full-redraw prompt-building code
into a shared redraw_make_pane_prompt() helper and added
redraw_damage_draw_pane_prompt(), which recomposes the prompt over
the drawn range. Taken from darlingm as-is.
All 9 regression tests in regress/ now pass. redraw_damage_grow_span_clip
was verified independently by disabling it and confirming
popup-drag-wide-character.sh reproduces its original failure.
Co-Authored-By: Michael K. Darling <darlingm@gmail.com>
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
From github.com/darlingm/tmux, branch pr5516-regression-fixes. Adds 9
regression tests covering gaps found in the redraw-damage-rectangles
branch: screen-write full/region redraw fallback, same-session window
switches, wide-character clipping at damage edges, pane prompts and
status lines surviving damage, floating-pane status format refresh,
and multi-client damage delivery.
redraw-multiclient.sh is adapted here to use ASCII pane borders
(pane-border-lines simple) instead of darlingm's original UTF-8
borders: the original reliably "failed" under this test's nested
tmux-in-tmux harness (relaying through an outer tmux client) due to
that harness mis-rendering a cell that held a multi-byte UTF-8 border
character being overwritten by later plain content - confirmed to be
a nested-relay artifact, not a real bug, by replaying the identical
drag sequence against a real terminal (xterm), where it never
reproduces. ASCII borders avoid the artifact; the test still reliably
catches the real "damage consumed by only one client" bug it targets
(verified by reintroducing that bug and confirming the test fails).
The other 8 tests are added verbatim from darlingm's branch. Four of
them (popup-drag-status-line.sh, popup-drag-wide-character.sh,
popup-drag-pane-prompt.sh, switch-client-redraw.sh) currently FAIL on
this branch, since the source fixes they test for have not been
merged yet - only the tests are being added here.
Co-Authored-By: Michael K. Darling <darlingm@gmail.com>
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
* refs/remotes/tmux-openbsd/master:
Correct a comparison ('\0' should be 0).
If a pane has no scrollbar, don't bother to redraw it, and if redraw is deferred change scrollbar redraws into redraw-all-scrollbars not redraw entire window. GitHub issue 5529.
* refs/remotes/tmux-openbsd/master:
Always use synchronized updates in copy mode (like when not in the active pane), since copy mode will almost always moved the cursor. From Artem Goldenberg in GitHub issue 5525.
* refs/remotes/tmux-openbsd/master:
Do not leak cached status line, from Jeong, Heon.
Free pane fallback range when freeing pane, from Jeong, Heon.
Do not leak path when destroying client, from Jeong, Heon.
Do not leak format when drawing preview, from Jeong, Heon.
Do not leak buffer name on failure, from Jeong, Heon.
popup_handle_drag()'s MOVE and SIZE branches each called
server_redraw_client(c) unconditionally, redrawing the client's entire
window on every drag step. Report damage for just the popup's old and
new rectangle instead, via a new popup_damage() (translating from raw
client/tty coordinates into window coordinates), and set
CLIENT_REDRAWOVERLAY so the popup itself still redraws.
redraw instead of full client redraw
The three interactive mouse-drag paths that move or resize a floating
pane (move-pane -M's Alt-drag, split-window/new-pane's interactive
resize, and resize-pane's own border drag) each unconditionally called
server_redraw_window(w), redrawing every pane in the window for a change
that only ever disturbs the floating pane's own old and new rectangle.
Switch all three to window_pane_redraw_floating().
window_set_active_pane() unconditionally called server_redraw_window(w)
on every active-pane change, redrawing every pane's content even though
only the previous and new active pane's border/status appearance
actually changed. Unzooming (which does change every pane's geometry)
still gets the full redraw; otherwise this now only redraws borders and
status.
redraw_client_damage() (added previously, unused until now) is called
from server_client_check_redraw()'s normal redraw pass, and
server_client_any_pane_redraw() now also checks for pending window
damage so a client with only damage (no PANE_REDRAW/PANE_REDRAWSCROLLBAR
flags) still gets its redraw pass run.
server_client_check_redraw() now returns whether the redraw was deferred
(waiting for outstanding tty output to drain) rather than performed. A
deferred redraw no longer escalates to a full CLIENT_REDRAWWINDOW to
avoid losing what was pending - server_client_loop() now only clears
PANE_REDRAW, PANE_REDRAWSCROLLBAR and window damage once every client
viewing a window actually drew this pass (tracked via a new per-window
redraw_deferred flag), otherwise they're left in place and retried in
their normal, narrowly-scoped form.
server_client_set_session() now redraws only if the client's session or
current window actually changed, not on every call (e.g. switch-client
-t= from clicking a pane name in the status line resolves here even
when nothing besides the active pane changed).
A drag callback's mouse_drag_update() now opens a sync region itself
(tty_sync_start()) before its first write, so a fast-path write it makes
directly and a later correction arriving via redraw_client_damage() end
up in the same atomic terminal update instead of two visible frames.
window_pane_redraw_floating() reports damage for only a floating pane's
old and new rectangle (via redraw_damage_window(), grown by one cell to
cover its border frame - see the "floating" case in screen-redraw.c),
instead of the caller falling back to a full window redraw. Any other
pane whose *scrollbar strip* - not its whole body - intersects either
rectangle still gets PANE_REDRAWSCROLLBAR directly, since scrollbars
aren't covered by the damage system.
Introduce a per-window list of damaged rectangles (struct redraw_damage)
and redraw_damage_window() to record them, redraw_client_damage() to
consume them by composing exactly the damaged cells (via a new
redraw_draw_damage_rect(), which also force-refreshes any pane-status
span it touches, since window_make_pane_status()'s content-diff check
has no way to know the physical cells were disturbed by something else).
redraw_draw_span() now takes an explicit [clip_x, clip_x + clip_n) range
instead of always drawing a span's full width, so a damage rectangle can
redraw just the portion of a span it actually covers.
screen_write_redraw_cb() - the fallback when a write can't be applied
directly to the terminal - now reports damage for just the affected row
via this mechanism, instead of unconditionally flagging the whole pane
for a full redraw.
wp->flags |= PANE_REDRAWSCROLLBAR was set unconditionally whenever a pane
reserved a scrollbar, even if layout_fix_panes() left its geometry
completely unchanged - forcing a needless scrollbar redraw on every layout
pass. Move it inside the existing "did this pane's geometry actually
change" check.
* refs/remotes/tmux-openbsd/master:
When a pane is respawned, reset the control mode offsets. GitHub issue 5498 from Alex Rattray.
Correctly clear clients referring to a session with destroy-unattached, GitHub issue 5497 from Jeong, Heon.
Restore umask if bind fails, from nachalsa at naver dot com.
