This changes read receipt sending logic to allow it advance further into events
without tiles (such as edits or reactions) that may exist after the last
displayed event.
By allowing the read receipt to advance past such events, this also marks as
read any related notifications. For example, edits trigger notifications by
default since they are `m.room.message` events, and with this change, such edit
notifications can finally be marked read.
Part of https://github.com/vector-im/riot-web/issues/9745
If the debugging mode of showing hidden events in the timeline is enabled, we
should also show replacements using the same view source tile as we do for
reactions. This allows easy debugging of replacement event data and also makes
the edit event look visually distinct from regular messages.
Fixes https://github.com/vector-im/riot-web/issues/9937
This adds additional receipt storage to so that we can handle cases where the
receipts and events lists get out of sync. If we ever find a user who previously
had a receipt but momentarily no longer does, we recover their previous receipt
and go with that until we hear something new.
Part of https://github.com/vector-im/riot-web/issues/9745
This changes how we determine read receipts for the entire message panel. We now
calculate read receipts for all events up front, which makes it easier to handle
hidden events by moving their read receipts up to the last shown event for
display purposes.
Part of https://github.com/vector-im/riot-web/issues/9745
Smooth scrolling browsers (Firefox) use the relative area to determine how much scroll to apply. Because breadcrumbs are short vertically, the scroll amount is minimal (3 units) in the Y direction. On browsers which don't smooth scroll the units are usually much higher (100 in Chrome on Win 10). Users seem to expect the scrolling to be quicker due to the horizontal space on breadcrumbs, so we add a bit more power to their scroll when it looks small.
Fixes https://github.com/vector-im/riot-web/issues/9394
Re-use walkDOMDepthFirst we already had for turning the editor into
a string, which also allows us to detect when leaving a node better,
which is needed to decide how block elements create newline parts
is more centralized and robust manner, more in line with
how html works (consecutive block elements break the line)
instead of hacky special casing from before.
Also, to support every line of MD generated
within a BLOCKQUOTE, the parts an element creates are returned
and then centrally prepended with "> ", instead of passing in
the parts array and doing it in every parse... function.