Some platforms with horribly broken MT handling (hai, PB) will manage to
screw something up so bad that we end up with slots never running
through initialState on a contact down, so they never get an initial_tev
recorded.
Ensure we do that if necessary when switching a buddy to voidState to
avoid crashes throughout the code, which assumes everything is sane and
doesn't guard accesses to initial_tev
Re: #11196, #10950 & #11111Closes: #11111
Namely, that swipes report the gesture's *starting* coordinates as `pos` (for hit-detection purposes), while pans (and basically everything else) use the *lift* coordinates (as well as provide a relative vector).
In each case, we now also provide a separate `start_pos`/`end_pos`, in case handlers need finer-grained data (e.g., anchoring UI elements to a swipe's lift coordinates).
Fix: #10539, and for context #6489, #6733, #6534
Reorganize and reword most of the settings to make it clear what actually ties into auto sync, and what doesn't. (Specifically, what happens when a pull attempts to sync forward or backward has nothing to do with auto sync, it applies in all cases; while the periodic sync *does* require auto sync).
The main point of contention, though, is that auto sync will now *always* attempt to setup network connectivity (i.e., on resume/suspend/close). Periodic sync will *not* though (the intent being that, if you use periodic sync, you're relying on the activity check to actually keep wifi on at all times)).
Since this may lead to a large amount of nagging about wifi toggles on devices w/ NetworkManager support, it is now *disabled* by default on those devices. (And given that it wouldn't have worked because of the lack of connectivity, that doesn't really make any practical difference ;p).
Additionally, given the fact that there's no way to make this behavior viable if the "before wifi" action is left at its default of "prompt", this feature now *requires* that to be set to "turn_on" (on devices where it can, of course); attempting to toggle it on will warn about that if necessary.
This change is retroactive (OTM).
Includes an assortment of fixes and cleanups, including migrating to the new LuaSettings API, which is why there's no longer a smattering of superfluous flushes.
* Get rid of the `canToggleGSensor` Device cap, it's now mandatory for `hasGSensor` devices. (This means Kindles can now toggle the gyro, fix#9136).
* This also means that `Device:toggleGSensor` is now implemented by `Generic`.
* Update the Screen & Gyro rotation constants to be clearer (c.f., https://github.com/koreader/koreader-base/pull/1568) (/!\ This might conceivably break some `rotation_map` user-patches).
* Input: Move the platform-specific gyro handling to Device implementations, and let Input only handle a single, custom protocol (`EV_MSC:MSC_GYRO`).
* Input: Refine the `rotation_map` disable method implemented in 43b021d37c. Instead of directly poking at the internal field, use a new method, `disableRotationMap` (/!\ Again, this might break some `rotation_map` user-patches).
* Input: Minor tweaks to event adjust hooks to make them more modular, allowing the Kobo implementation to build and use a single composite hook. API compatibility maintained with wrappers.
Basically:
* Use `extend` for class definitions
* Use `new` for object instantiations
That includes some minor code cleanups along the way:
* Updated `Widget`'s docs to make the semantics clearer.
* Removed `should_restrict_JIT` (it's been dead code since https://github.com/koreader/android-luajit-launcher/pull/283)
* Minor refactoring of LuaSettings/LuaData/LuaDefaults/DocSettings to behave (mostly, they are instantiated via `open` instead of `new`) like everything else and handle inheritance properly (i.e., DocSettings is now a proper LuaSettings subclass).
* Default to `WidgetContainer` instead of `InputContainer` for stuff that doesn't actually setup key/gesture events.
* Ditto for explicit `*Listener` only classes, make sure they're based on `EventListener` instead of something uselessly fancier.
* Unless absolutely necessary, do not store references in class objects, ever; only values. Instead, always store references in instances, to avoid both sneaky inheritance issues, and sneaky GC pinning of stale references.
* ReaderUI: Fix one such issue with its `active_widgets` array, with critical implications, as it essentially pinned *all* of ReaderUI's modules, including their reference to the `Document` instance (i.e., that was a big-ass leak).
* Terminal: Make sure the shell is killed on plugin teardown.
* InputText: Fix Home/End/Del physical keys to behave sensibly.
* InputContainer/WidgetContainer: If necessary, compute self.dimen at paintTo time (previously, only InputContainers did, which might have had something to do with random widgets unconcerned about input using it as a baseclass instead of WidgetContainer...).
* OverlapGroup: Compute self.dimen at *init* time, because for some reason it needs to do that, but do it directly in OverlapGroup instead of going through a weird WidgetContainer method that it was the sole user of.
* ReaderCropping: Under no circumstances should a Document instance member (here, self.bbox) risk being `nil`ed!
* Kobo: Minor code cleanups.
* ImageViewer: Minor code cleanups
* GestureDetector: Fix the `distance` field of `two_finger_pan` & `two_finger_swipe` gestures so that it's no longer the double of the actual distance traveled. Get rid of existing workarounds throughout the codebase that had to deal with this quirk.
* Kobo: Discriminate between the Touch A/B and the Touch C properly, and implement actual support for the A/B input quirks. This means the clunky touchscreen probe widget shown on fresh installs on those devices is now gone :}.
* Input: Fix an off-by-one in most adjustTouchMirrorX/Y callers (only rM was doing it right), and adjust their documentation to avoid similar mistakes in the future.
* GestureDetector: Unify logging to always display transformed coordinates for simple gestures.
* GestureDetector: Fix two-contact hold lifts to be computed at the midpoint between the two contacts, like their holds counterpart already did.
slots
This ensures we won't leave *any* slot in an undefined state because
we skipped parsing 'em because what we consumed first yielded a
gesture.
(In particular, this could leave a few slots dangling in the "hold"
state in corner cases involving spider-hand finger tapping ;p).
Cleans up the slot state clearing in GestureDetector to only clear the
necessary slots (e.g., two-finger gestures now only clear their own two
slots; and holds only clear their own slot).
The fact that every slot will be consumed ensures that every slot will
naturally get their contact up handled, which wasn't the case before,
hence those crappy workarounds.
As far as timerfd callbacks are concerned, this *does* introduce the
possibility of deadline collisions, so, do reimplement minimal safety
checks to ensure we run (and free) the right timerfd callback.
Some drivers can bundle the same slot multiple times in the same input
frame. We were only correctly coalescing *consecutive* slots, but some
drivers can do that in non-consecutive sequences (e.g., 1a -> 2a -> 1b
-> 2b), so, handle that, too.
Seen on neonode v2 grids.
For realz, this time.
This reverts the original attempt, because it was gratuitous
overcomplexification that turns out to be completely unnecessary.
This also fixes a few subtle MT handling snafus on some devices.
We'll only honor one (the last one) anyway, and we've already cleared
the timers, so the others are essentially lost and stuck in limbo.
(Repro: do a series of staggered holds on different words with different
fingers in a CRe document; the one that'll "take" is the last one,
but *every* slot will be stuck in a hold state, which can essentially
semi lock you into broken input until you manage to clear the right
slots by doing duplicate multi-slot holds again...).
* GestureDetector: Downgrade probeClockSource log messages to debug
We're seeing a bit more of 'em between autostandby and inhibitInput now,
and the whole thing was implemented long enough ago that it's no longer
critical to see this in random non-debug logs ;).
* Input: Reset gesture state in inhibitInput
Should prevent bogus gestures when re-enabling the handlers if we were
called in the middle of gesture detection.
* Decode EV_KEY:KEY_BATTERY
* Input: Only drop hovering *pen* events.
There are currently too many broken 0-pressure *finger* events being
reported on the Elipsa, making a dumb rejection highly annoying.
* Bump base
https://github.com/koreader/koreader-base/pull/1393
* Rely on actual events to detect loss of contact for the "snow"
protocol.
Allows simplifying the whole thing.
* Use `ipairs` over `pairs` for pure arrays.
* Input: Don't create a new TimeVal object for input frame timestamps, just promote our existing table by assigning it the `TimeVal` metatable.
* TimeVal: Export (const) `zero` & `huge` TimeVal objects, because they're common enough in our codebase. (NOTE: not actually const, that's a Lua 5.4 feature ;p).
* GestureDetector: Explain the behavior of the `last_tevs` & `first_tevs` tables, and why one needs a new object and not the other.
* Speaking of, simplify the copy method for `first_tevs`, because it doesn't need to create a new TimeVal object, we can just reference the original, it's unique and re-assigned for each frame.
* TimeVal: Log the results of the COARSE probes in debug logs
* GestureDetector: Print details of failed clock probes in debug logs
* GestureDetector: Skip the BOOTTIME probe when CLOCK_BOOTTIME is unsupported.
* Input: Decode ABS_DISTANCE events in debug logs
* Get rid of duplicated <linux/input.h> constants, use the FFI module everywhere (re #7536)
* Kobo: Get rid of the `touch_alyssum_protocol` quirk. Replace it by setting `main_finger_slot` to `1`, like on the H2O.
* Actually load librt properly on old Linux systems (fix#7472)
* Made sure SDL & Android honored timeouts properly in some edgy (or not so edgy, in SDL's case) cases (c.f., https://github.com/koreader/koreader-base/pull/1343).
* Try harder not to consume a timer callback when the timerfd backend is in use and `select` timed out *not* because of a timerfd expiring. This would take some particularly heinous timing to reproduce (e.g., a timerfd set, but a task deadline expiring *before* it). May or may not help with #7473.
* Decode `BTN_` KEY codes in verbose debug logs.
* Fix MultiTouch on the H2O (GestureDetector was hard-coding the two slots as 0 & 1, but the H2O kernel is a special snowflake that doesn't use slot 0, instead switching to slot 1 & 2).
* ReaderDictionary: Port delay computations to TimeVal
* ReaderHighlight: Port delay computations to TimeVal
* ReaderView: Port delay computations to TimeVal
* Android: Reset gesture detection state on APP_CMD_TERM_WINDOW.
This prevents potentially being stuck in bogus gesture states when switching apps.
* GestureDetector:
* Port delay computations to TimeVal
* Fixed delay computations to handle time warps (large and negative deltas).
* Simplified timed callback handling to invalidate timers much earlier, preventing accumulating useless timers that no longer have any chance of ever detecting a gesture.
* Fixed state clearing to handle the actual effective slots, instead of hard-coding slot 0 & slot 1.
* Simplified timed callback handling in general, and added support for a timerfd backend for better performance and accuracy.
* The improved timed callback handling allows us to detect and honor (as much as possible) the three possible clock sources usable by Linux evdev events.
The only case where synthetic timestamps are used (and that only to handle timed callbacks) is limited to non-timerfd platforms where input events use
a clock source that is *NOT* MONOTONIC.
AFAICT, that's pretty much... PocketBook, and that's it?
* Input:
* Use the <linux/input.h> FFI module instead of re-declaring every constant
* Fixed (verbose) debug logging of input events to actually translate said constants properly.
* Completely reset gesture detection state on suspend. This should prevent bogus gesture detection on resume.
* Refactored the waitEvent loop to make it easier to comprehend (hopefully) and much more efficient.
Of specific note, it no longer does a crazy select spam every 100µs, instead computing and relying on sane timeouts,
as afforded by switching the UI event/input loop to the MONOTONIC time base, and the refactored timed callbacks in GestureDetector.
* reMarkable: Stopped enforcing synthetic timestamps on input events, as it should no longer be necessary.
* TimeVal:
* Refactored and simplified, especially as far as metamethods are concerned (based on <bsd/sys/time.h>).
* Added a host of new methods to query the various POSIX clock sources, and made :now default to MONOTONIC.
* Removed the debug guard in __sub, as time going backwards can be a perfectly normal occurrence.
* New methods:
* Clock sources: :realtime, :monotonic, :monotonic_coarse, :realtime_coarse, :boottime
* Utility: :tonumber, :tousecs, :tomsecs, :fromnumber, :isPositive, :isZero
* UIManager:
* Ported event loop & scheduling to TimeVal, and switched to the MONOTONIC time base.
This ensures reliable and consistent scheduling, as time is ensured never to go backwards.
* Added a :getTime() method, that returns a cached TimeVal:now(), updated at the top of every UI frame.
It's used throughout the codebase to cadge a syscall in circumstances where we are guaranteed that a syscall would return a mostly identical value,
because very few time has passed.
The only code left that does live syscalls does it because it's actually necessary for accuracy,
and the only code left that does that in a REALTIME time base is code that *actually* deals with calendar time (e.g., Statistics).
* DictQuickLookup: Port delay computations to TimeVal
* FootNoteWidget: Port delay computations to TimeVal
* HTMLBoxWidget: Port delay computations to TimeVal
* Notification: Port delay computations to TimeVal
* TextBoxWidget: Port delay computations to TimeVal
* AutoSuspend: Port to TimeVal
* AutoTurn:
* Fix it so that settings are actually honored.
* Port to TimeVal
* BackgroundRunner: Port to TimeVal
* Calibre: Port benchmarking code to TimeVal
* BookInfoManager: Removed unnecessary yield in the metadata extraction subprocess now that subprocesses get scheduled properly.
* All in all, these changes reduced the CPU cost of a single tap by a factor of ten (!), and got rid of an insane amount of weird poll/wakeup cycles that must have been hell on CPU schedulers and batteries..
Follow up to b90f6db8: allow specifying an other
value for tap interval when the keyboard is shown
(a good value for tap interval on reader and UI
elements might be too long on the keyboard, and
prevent typing fast).
On devices where the event time is the uptime (time since
boot), we don't need to trash it and use TimeVal:now()
and lose precision. We can still use these times for
relative delays and durations computations, which is
mostly all we use them for.
We just need a real clock time in GestureDetector for
two timers (long-press and double tap), where we do
the event time type detection on the first event, on
all devices.
* Uses bunch of new plumbing in base to configure screen rotations in hardware (koreader normally does this via blit buffer rotations, except for android).
* Some PB specific kludges that used to pollute core/framebuffer_linux are brought into PBs frontend driver.
* Allow locking the gyro to the current screen mode (i.e., orientation).
* Tweak the "sticky rota" option to work both ways
* More rotation constant usage instead of magic numbers
Prevent the Tap>Hold setTimeout'ed function to trigger
on another later Tap.
Fix a Tap quickly following a Swipe (which triggers this
Tap>Hold timer) from becoming a Hold.
This commit standardizes the various todos around the code a bit in a manner recognized by LDoc.
Besides drawing more attention by being displayed in the developer docs, they're also extractable with LDoc on the command line:
```sh
ldoc --tags todo,fixme *.lua
```
However, whether that particular usage offers any advantage over other search tools is questionable at best.
* and some random beautification
Multiswipes consisting of mixed straight and diagonal strokes are not dependable and too easy to mess up, but making them mutually exclusive seems to work out quite well.
When multiswipes are enabled, this fixes the long-standing complaint that swiping to open the menu could unintentionally trigger some light panning. With the introduction of multiswipes, this problem has become more noticeable.
Reported by @poire-z, cf. https://github.com/koreader/koreader/pull/4640#issuecomment-466544922
Apparently it's natural for me to make the second swipe slightly longer than the first, so I never noticed a logic issue. I did notice that it seemed slightly harder to make 4-swipe multiswipes than I expected it to be, but those are not necessarily easy gestures to make.
The problem was that I needed to prevent obviously silly gestures like west west west east. In ignoring such duplication, what I accidentally did was to ignore any further movement west after the first multiswipe direction was detected, meaning that the following swipe east could still end up as a relatively western movement overall.
By simply updating the current multiswipe slot in case of the same direction, both problems are prevented. We'll never get the same direction twice, and X moves over to where it's supposed to be on the left.
The basic idea is that you gain an infinite amount of extra gestures,
although in practice you're probably mostly limited to your 16 basic
two swipe combinations and maybe a few three swipe ones.
Touch zone decouples screen size from gesture event registration.
The win here is each individual widget does not need to update
gesture range on screen rotate/resize anymore.
Another advantage is we now have a centralized ordered array to handle
all registered touch event listeners, makes it much easier to resolve
gesture range conflicts between multiple widgets.
This patch also includes the following changes:
* migrate readerpaging to use readerui's touch zone
* migrate readerfooter to use readerui's touch zone
* move inverse read direction setting to touch menu's setting tab
* moved kobolight widget from readerview into readerui
* various dead code cleanups and comments
