On application close we used to access invalid memory associated with
the audio subsystem. This was because on destruction of static variables
we tried to access state associated with an already destructed item.
These variables seem to have no good reason to be static. They should be
tied to a single CoreAV instance and are only accessed through an
existing CoreAV instance.
The previous implementation of hiding the date line would cause 100% cpu
usage. When the date line was shown it would hide the top line, causing
the date line to be hidden again due to a state change in which dates
was visible.
This is a minimal patch to work around the issue by pretending the line
covered by the date line is the first visible line when the dateline is
shown
Fixes#5620
Before a bug in qTox would make it possible for a user to try to send an empty
action type message. This would fail to send at toxcore, but still be persisted
in history, causing it to fail every time FauxOfflineEngine tried to resend
it. Moving these stuck messages into the broke_messages table will stop qTox
from attempting to deliver them on each connect, and display in the GUI to
users that the messages aren't really pending anymore.
Otherwise sending just "/me " calls down to core with 0-length action, which
causes a toxcore send error.
There may be messages that are stuck in history as pending with "/me " content,
causing them to fail to send on every start. Those should be moved to the new
broken_messages table, similar to stuck pending messages from #5776. There's no
reason this fix to stop adding more broken messages needs to wait on the
history fix, though.
"Sent" to "Delivered", since that's what we're tracking also rename "isOffline"
to "isPending", which is boolean the same, but "isOffline" was opposite of its
correct value before, so boolean reverse the assignment.
Fix#5776
Due to a long standing bug, faux offline message have been able to become stuck
going back years. Because of recent fixes to history loading, faux offline
messages will correctly all be sent on connection, but this causes an issue of
long stuck messages suddenly being delivered to a friend, out of context,
creating a confusing interaction. To work around this, this upgrade moves any
faux offline messages in a chat that are older than the last successfully
delivered message, indicating they were stuck, to a new table,
`broken_messages`, preventing them from ever being sent in the future.
This allows upgrade steps to query the db at the last version and run C++ code
on the results, then do a single transaction to make the upgrade, instead of
all actions of each upgrade step being required to be part of the overall
upgrade transaction.
* Fix callback hookup order in ChatHistory
* Add correct call to SessionChatLog to insert an unfinished message
* Fix incorrect logic in parsing History database response
This commit makes LoginScreen return QDialog::Rejected (0) and
QDialog::Accepted (1) instead of C standard return values. This should be more
robust with regards to special cases in Qt implementation.
Fixes#5781
qTox added specialized tray icon backends over time to work around bugs with
the default implementation, since then our GTK+2 backend has stopped working,
our appindicator backend was never selected by default by cmake, and
statusnotifier was never selected at SystemTrayIcon construction, leaving us
only ever using the broken GTK+2 backend, or theoretically the Unity backend,
which I didn't see selected on Ubuntu 16.04 Unity. In all other cases we would
fall back to the Qt backend.
Qt icon has improved over time, and our platform specific icons have become
stale, with GTK+2 becoming deprecated and GTK+3 not having a similar feature,
QSystemTrayIcon has been tested on a variety of DEs and works as well or better
in all cases, as shown in the table at
https://github.com/qTox/qTox/issues/5881#issuecomment-541892457Fix#5881Fix#5859
seccomp is used by the docker host, and the default version shipped with
Ubuntu 16.04 does not support syscalls used by Qt5's configure inside the
docker container. Upgrading libseccomp2 resolves this issue.
Fix#5874
Errors are parsed and printed, but which specific node is being connected to
isn't very relevant to any errors that would occur, and the nodes list is
already updated to prune offline nodes based on nodes.tox.chat periodically.
This provides some extra privacy about which connections are being made,
even though the bootstrap nodes are already public.
ffmpeg dependency unbound attempts to symlink to /usr/local/sbin, failing if
the directory does not exist. /usr/local/sbin does not exist by default on
10.13 High Sierra, so prompt to create it if it doesn't exit, and create it for
travis CI automatically.
Fix#5515
brew install might take a long time to build Qt or ffmpeg. Travis kills us if
we don't output for 10 minutes. Travis also kills us if we output too much, so
verbose isn't an option. So just output some dots...
