Follow up to the merge of PR #10560 which skips input_delay when less
than 1 KiB of program output is pending.
Fix a data race: the I/O thread called
vt_parser_pending_input_is_interactive() which read read.pos under the
parser lock, but the main thread modifies read.pos without the lock
while parsing, so taking the lock on one side only did not protect it.
Instead, vt_parser_commit_write() now returns whether the pending input
is small, computed from read.sz and write.pending, which are only
modified with the lock held. This may overestimate while a parse is in
progress or when an incomplete escape code is retained, in which case
we simply fall back to the normal input_delay behavior. It also saves a
lock/unlock cycle per read in the I/O thread.
Bound the number of early main loop wakeups: since render() bypasses
repaint_delay whenever input was read, input_delay was the effective
cap on parse/render frequency for streaming output. Waking the main
loop on every small read removed that cap entirely for programs that
emit a continuous stream of small writes, costing CPU and increasing
the chance of drawing partial screen updates. Now an early wakeup is
only allowed if the previous wakeup was not itself early, so typed
character echo is still delivered immediately while streams of small
writes get at most two wakeups per input_delay.
Replace the magic 1024 with SMALL_PENDING_INPUT_THRESHOLD, rename
pending_input_is_interactive() to pending_input_is_small() since the
check has nothing to do with interactivity, clarify the input_delay
option documentation and add a changelog entry.
189d78283c introduced the application_has_finished_launching variable,
but it's never read. This causes a compile error on macOS 26.6.2 with
Clang 23.1.2.
Remove the unused variable and assignment so the source can be compiled.
Signed-off-by: Jiamu Sun <39@barroit.sh>
Use Unicode word boundaries with the stdlib regexp engine
Go's regexp only treats ASCII characters as word characters for \b, while
regexp2 uses Unicode ones, so switching the builtin path pattern to the
stdlib engine changed which paths were marked next to non-ASCII text, for
example config.yaml文件 marked config.yaml. Match against a proxy of the
text in which every non-ASCII rune is replaced by a single ASCII byte that
is a word character iff the rune is one, then map the offsets back.
find_matches_in_lines was passing the exclusive end offset of a match to
find_pos, so a match ending at the end of the joined screen lines mapped
to no line and was dropped. It was not highlighted, not counted and n/N
skipped it. Since screen lines are not padded, this hit any match on the
last character of a line, for example searching for foo in "x := foo".
Look up the line holding the last matched byte instead.
regexp2 backtracks on the path pattern, retrying it at every position
of a long line that has no spaces, so the path and linenum types took
seconds on a screen full of such text (about 36s for a 380x100 window
of hex). The builtin patterns dont need lookarounds, so use the linear
time stdlib engine for them. User supplied --regex patterns stay on
regexp2 so lookarounds keep working (#6265).
Window drop previews are cleared before tab drop previews on each tab drag
update. With the shared hover state, this cancels the tab hover timer on every
update, preventing the 0.6 second tab switch.
Preserve a tab hover timer when clearing a window drop preview. Keep the
existing cancellation behavior for window drags and cover the regression.
- Dropping a tab on the tab bar always re-orders. Only windows are merged
into a tab when dropped on its middle. Single window tabs still switch
tabs on hover so they can be dropped into the content area.
- Draw the vertical insertion marker in the blank line between tabs so it
does not hide a tab title, using a single cell only when crowded.
- Make the dropped tab active after a drop from another OS Window, instead
of whichever tab ended up at the stale active index after re-ordering.
- Dropping the window of a single window tab at an insertion point moves
its tab, preserving title and layout, and is a no-op next to itself.
- Do not leave an empty tab behind if moving a dropped window fails.
- Highlight the new tab button only when appending.
- Unify the window and tab drag hover state and share the tab bar drop
hit testing code.