Merge release 4.3 into development

* Add simple 2min autosave

* Remove test debug

* Le review feedback

* Don't load-save straight away

* Use Travis config validation

See https://blog.travis-ci.com/2019-10-24-build-config-validation for details

* Fix validation warnings

* Remove temporary aliases when calling resetProfile()

* Ensure aliases are disabled, cleaned up and then enabled when resetting profile

* Use latest VS image in Appveyor

* Use Qt 5.13.0 on appveyor builds

(cherry picked from commit 81ac771)

* Up to Qt 5.13.2

* Install right version in SDK as well

* New translations mudlet.ts (German)

* New translations mudlet.ts (Italian)

* New translations mudlet.ts (Italian)

* New translations mudlet.ts (Italian)

* New translations mudlet.ts (Russian)

* New translations mudlet.ts (Portuguese)

* New translations mudlet.ts (Portuguese, Brazilian)

* New translations mudlet.ts (Portuguese, Brazilian)

* New translations mudlet.ts (Portuguese, Brazilian)

* New translations mudlet.ts (Portuguese, Brazilian)

* New translations mudlet.ts (Portuguese, Brazilian)

* New translations mudlet.ts (Portuguese, Brazilian)

…ces (#3249)

* Refactor: remove multiple redundent semi-colons following closing braces

One was pointed out to me by CodeFactor in a recent commit to a PR and when
I looked I found a few more!

During the testing of this PR there were CI failures for the macOS case which
turned out to be because `pcre2` was being considered for `pcre` as a package -
the former has an updated formula currently and was tripping the code to
try and update the latter package - but that package does not need it. The fix was
to add the `-w` (whole-word) option to a couple of calls to `grep`.

Signed-off-by: Stephen Lyons <slysven@virginmedia.com>

This was coming from mishandling the XDG_DATA_DIRS environmental variable
- as environmental variables have to be converted to qmake variables before
they can be used easily in the qmake profile file.

Signed-off-by: Stephen Lyons <slysven@virginmedia.com>

This ought to help if any line number indexes are being used to access
the buffer underlying a console as then they can be revised downwards to
match the number of lines removed from the buffer/console concerned.

This commit also improves the UI for the setConsoleBufferSize(...) Lua API
function to bring it up to our current design style. It will now return
true on sucess or nil + an error message otherwise.

Signed-off-by: Stephen Lyons <slysven@virginmedia.com>

Redo_Enhance: overhaul stopwatches

This PR is a rework of the original PR #2516 squash-and-merged with the
bugfix that I proposed as PR #3162 but which became difficult to apply when
the original PR was eliminated from the development branch because it has
been reverted from the 4.2.0 release and then that had been merged into
the development branch!

This PR allows stopwatches to be stopped and read at any time afterwards,
for any number of times. It also:
* allows stopwatches to be destroyed - so ID numbers WILL get reused - with
  a Lua (bool) destroyStopWatch((int) id) function.
* allows them to be marked as persistent so that they are saved with the
  profile and reloaded again - if they were running then they will continue
  to increment when the profile is not loaded so they can be used to
  real time events outside of the profile/session; this uses a new Lua function
  (bool) setStopWatchPersistence((int) id, (bool) setPersistent).
* allows them to be adjusted (even so they become negative) so can be used
  to count down as well as count up time - whether the stopwatch concerned
  is running or not.
* have error messages that conform to our current style
* have more run-time handling - most actions that produce no effect will
  advise that this is the case -  e.g. stopping a stopwatch that was NOT
  running...
* can handle periods of time longer than a day - the previous
  implementation wrapped around after 24 hours.
* should not affected by DST changes - OS permitting.
* the Lua API also gains a getStopWatches() function that returns a table
  with the id numbers as keys and values as tables of:
  * (bool) isRunning
  * (bool) isPersistent
  * (string) name
  * (table) elapsedTime - containing broken down time:
    * (bool) negative
    * (int) days
    * (int) hours (0 to 23)
    * (int) minutes (0 to 59)
    * (int) seconds (0 to 59)
    * (int) milliSeconds (0 to 999)
    * (float) decimalSeconds floating point value of whole of the time in
    seconds (can be negative!)
* allows stop watches to be named - which is useful as then they can be
  identified in scripts; this means that all the stop watch functions
  can now take a name string as well as a numeric argument. Using an empty
  string will access the first (lowest id) stopwatch that does not have a
  name and the createStopWatch() function will accept an optional string
  argument as a name.  For simplicity each name must be unique and this is
  enforced for that function and the added setStopWatchName(id or name,
  newName) function. The latter will also accept an empty string as either
  the first or second argument; in the first case it will assign the name
  to the first unnamed stopwatch and the second will clear the name of the
  specified one.
* added getStopWatchBrokenDownTime(...) which returns the same broken
  down elements in a table for a single specified timer (day count; hours;
  minutes; seconds; milliseconds and whether the time is positive or
  negative {when preset with a negative adjustment and used as a count
  down})...

During debugging I found out that the process of loading an existing
profile that contained createStopWatch() calls was creating them during
the testing/loading phase so I had to add extra code to prevent that
function from taking effect whilst (bool) ~~`Host::mIsProfileLoadingSequence`~~
*revised to use a different, new, flag: `Host::mBlockStopWatchCreation`
which is cleared earlier in the loading sequence* is true. Then, when
testing the resetProfile() function I found that the same thing was
happening AND that the non-persistent stopwatches needed to be removed as
well - which is now solved by also preventing stopwatch creation whilst
(bool) Host::mResetProfile is set and by running a new method
(void) Host::removeAllNonPersistentStopWatches()! This thus allows
stopwatches to be created during the profile startup sequence when
lua scripts are run on loading (but after a prior compilation to test
for script validity has been done).

The code in (void) stopWatch::Host::adjustMilliSeconds(const qint64)
originally used (QDateTime) QDateTime::addMSecs(qint64) incorrectly in
that I had thought it adjusted the QDateTime it was called upon whereas it
returns a reference to the adjusted value - so needed to be invoked in a
different manner which was already being done in the original PR.

QString stopWatch::getElapsedDayTimeString() const made use of int for
some intermediate local variables but on Windows platforms the int type may
be 32-bit long and thus not big enough to contain 64-bit values - and some
other locals can be much shorter because of the limits that the code places
on their values but will give compiler warnings without static_cast <T>.

Use std::chrono_literals to specify some needed time intervals rather than
large long integer literal constants.

Refactor a block of code used to generate the broken down time as a Lua
table so that two instances are handled by a single helper function.

Additional code (the bug fix) has been included to provide
backwards-compatibility for Lua startStopWatch(id) function which recreates
prior start stop-watch behaviour when it is created:

The prior form of startStopWatch(...) would reset and restart the indicated
stopwatch each time that it was called.  This is not really compatible with
the revised functionality which allows the recorded time to be adjusted
even before the stopwatch is first used.  To allow existing scripts to
continue to experience the same API this commit adds an optional second
boolean argument to the startStopWatch(...) call ONLY WHEN the first
argument is an id number and NOT a string name.  If the second argument is
omitted (as it will be when using older scripts) or is true then each time
the function is used the stopwatch will be reset and restarted. Just in
case the same behaviour ***is*** wanted with stop watch created with a
**name** then a second boolean `true` will start the stopwatch from zero.

This PR replaces (and thus will) close #3162 .

A separate commit was appended and squashed in to reduce a code
duplication pointed out by CodeFactor:

Created a single function that is used to do the same thing in four
separate stop-watch functions.

Signed-off by: Stephen Lyons <slysven@virginmedia.com

This PR is a squash and merge of three commits with the following
edited summary of the commit messages:

This applies to the main console window for a profile and now reports all
mouse button presses and releases in the `sysWindowMousePressEvent` and
`sysWindowMouseReleaseEvent` instead of only the first three (left, right,
middle).

The code for the same things in Labels was checked and I found that we
were generally okay there already except that we were fabricating native
Qt mouse events for the mapper where it overlaid a label but we were using
an older (Qt4 compatible) event that only handled one mouse wheel at a time
and we were assuming it was for a vertical mouse wheel. As the owner of a
mouse with two wheels that assumption is misplaced so I have rejigged
`(bool) TLabel::forwardEventToMapper(QEvent*)` so that it can pass
horizontal and vertical mouse wheel events in the future.

The new constructor I mentioned in the previous paragraph was only
introduced in Qt 5.12 but not documented as such so it was found
necessary as a result of the CI process to reintroduce a different,
older and now deprecated in current code, `QWheelEvent` constructor
to support builds against Qt 5.11.x versions.

Signed-off-by: Stephen Lyons <slysven@virginmedia.com>

…2621)

This is a squash-and-merge of a number of commits with the following
edited summary:

This PR reorganises the 2D Painter code to draw the player room
indicator AFTER all the rooms have been drawn so that it does not get
overdrawn by rooms painted after the player's room - which it could be
if the rooms are at maximum size or the area is in "gridMode".

It also offers some alternative marker options in the form of an oversized
ring around the room which reduces the obstruction of the details of the
room when the marker is drawn on it. The alternatives are:
* a fixed red colour ring
* a fixed blue/yellow colour ring - two contrasting colours mean that it
can not be lost if drawn over a room that already has one of them as its
colour (environment) setting.
* a customizable two colour ring where the user can set the outer and inner
colours as they like - even to the same colour.

The ancient `(bool) Host::mMapStrongHighlight` option is still respected,
I note that it was introduced in 3cb1719
from 2011/01/20 (after Mudlet 1.0.5) and the code to control it was lost
in 8b0e8bb from 2012/12/31 "NEW Vadim
Peretokin: mapper remembers settings" just before the release of Mudlet
2.0...!

Refactor: ensure player room is drawn last

This is done by extracting the relevant room drawing code to a separate
inline function that is used whilst iterating through the area rooms to
paint them, skipping but noting if the player room is found and then that
function is used to paint the player's room afterwards.

Revise: made code acceptable to all compilers

The clang compiler does not like `auto` being used as a type in function
prototypes as it is not formally part of the C++14 standard but a GCC
extension - it got put in by the Qt Creator function extraction code
option on the refactor menu.

Revise: allow player room marking customisation to be saved and restored

They are saved with the profile's data rather than the map.

Revise: show transparency effects in custom player room marker colours

It is not possible to just colour the background of the `QPushButton`s that
set the custom colours to use for the *custom* player room marker as the
transparency effects do not show. Instead this commit arranges for an icon
with a black-and-white checker board pattern is over-painted with a brush
in the relevant colour so that when the opacity is less than 100% the
pattern is shown to an extent which is proportional to the degree of
transparency of the colour chosen.

Refactor: extract code to start 2D map speed-walk & simplify drawRoom code

Suggested in peer review comment:
#2621

Signed-off-by: Stephen Lyons <slysven@virginmedia.com>

/src/icons/black_white_transparent_check_1x3_ratio.png -- 0.43kb -> 0.38kb (11.19%)

Signed-off-by: ImgBotApp <ImgBotHelp@gmail.com>

This entails changing the name of the existing 32 Bits one so that another
one for 64 Bits can be placed alongside it in the same directory. The
Discord release zip file puts them (and the Linux and MacOs 64 bit ones) in
separate directories - which is a little inconvenient for us...

I have also put in some qDebug() messages - because I am having difficulty
getting the Discord RPC to respond to local builds - though it works fine
for Linux AppImages - which use the very same copy of the library for THAT
OS. They are enabled with a DEBUG_DISCORD #define visible to the
discord.cpp compilation unit...

Turns out the origin Windows libraries I had put in were both 32-bit ones -
the ones inserted now are described differently by the MSYS2 `file` command as:
* discord-rpc32.dll: "PE32 executable (DLL) (console) Intel 80386, for MS
                      Windows"
* discord-rpc64.dll: "PE32+ executable (DLL) (console) x86-64, for MS
                      Windows"
Further more they were certified and signed on "‎27 ‎November ‎2018 17:20" so
that is consistent with them being the (current) version 3.4.0 release
files from that same date.

Signed-off-by: Stephen Lyons <slysven@virginmedia.com>

* Update TTextEdit.cpp

Put in a couple checks for bounds in front of every place I can imagine an out of bounds check is needed.

* Update TTextEdit.cpp

Fixed the signs, and moved code around to return on what would have caused the crash rather than just break out of a loop and never return the selected text.

* Update TTextEdit.cpp

I think this undoes all my changes.

* Update TTextEdit.cpp

Undo final change.

* Update TTextEdit.cpp

try catch blocks and variable dumps.

* Update TTextEdit.cpp

Fix crash condition reported at #3239 (comment) .

* Update TTextEdit.cpp

Removed the debug code as per Vadi2's request in #3241 (comment) .

…alised

Forgot this when I was producting #2621 as I tend to use the other form of
the map in a floating/dockable window.

Signed-off-by: Stephen Lyons <slysven@virginmedia.com>

… is trimmed" (#3262)

This reverts commit 8f22484.

* New translations mudlet.ts (Italian)

* New translations mudlet.ts (Portuguese, Brazilian)

* New translations mudlet.ts (Portuguese, Brazilian)

* New translations mudlet.ts (German)

* New translations mudlet.ts (Russian)

* New translations mudlet.ts (Italian)

* New translations mudlet.ts (Italian)

* New translations mudlet.ts (Portuguese, Brazilian)

* New translations mudlet.ts (Italian)

* New translations mudlet.ts (German)

* New translations mudlet.ts (Russian)

* New translations mudlet.ts (Chinese Simplified)

* New translations mudlet.ts (Chinese Traditional)

* New translations mudlet.ts (Dutch)

* New translations mudlet.ts (English, United Kingdom)

* New translations mudlet.ts (French)

* New translations mudlet.ts (Greek)

* New translations mudlet.ts (Pirate English)

* New translations mudlet.ts (Polish)

* New translations mudlet.ts (Portuguese)

* New translations mudlet.ts (Spanish)

* New translations mudlet.ts (Portuguese, Brazilian)

* New translations mudlet.ts (Portuguese)

* New translations mudlet.ts (Russian)

* New translations mudlet.ts (Polish)

* New translations mudlet.ts (Polish)

* New translations mudlet.ts (Italian)

* New translations mudlet.ts (Portuguese, Brazilian)

* Add Mud Sound Protocol (MSP) Support

* Adjust URL to http, remove extra return

* Resolving code compliance checks

* Resolving code compliance checks

* Resolving code compliance checks

* Resolving code compliance checks

* Resolving code compliance checks

* Wildcards, CLAZY, slashes, break, nullptr and paths

* Resolved directory creation issue, downloads working

* Enhanced MSP with QAudio::CustomRole

* Backed out QAudio, Possible final revisions

* Updated Protocol Preferences in Settings

* Code compliance

* Code compliance

* Code compliance

* Trigger Windows & Linux builds

* Updated download methodology

* Code compliance

* Code compliance

* Resolving conflict

* Code compliance

* Introducing TMedia

* Code Compliance

* Code Compliance

* Sets .wav and .mid correctly

* Typo resolved

* Security checks

* GMCP Support

* Code compliance

* Added lua receiveMSP()

* Code compliance

* Resolving conflict

* Updated CMakeLists.txt

* Updated CMakeLists.txt

* Updated CMakeLists.txt

* Code compliance

* Registered receiveMSP

* Code compliance

* Undo missing GNU

* Refined code

* Enhance GMCP Support

* Enhance GMCP Support

* Enhance GMCP Support

* Code compliance

* Added Suppression

* Code compliance

* Enhanced suppression

* Implemented Priority

* Added Validation

* Code compliance

* Resolved bug

* Introducing GMCP.Media

* Code compliance

* Code compliance

* GMCP volume bug resolved

* mediaLocation by protocol

* List fix

* MediaKey constraints

* matchMediaKeyAndStopMediaVariants

* Code dedupe

* Code dedupe

* Resolving priority bug

* Bug fixes

* bool mediaContinue

* Trialing Video

* Untrialed video

* Untrialed video

* Removed target (no video)

* Minor update

* Documentation links

* Code compliance

* Wiki reference update

* Resolve bug in XMLImport.cpp

* Fix unintialised TMediaPlayer

* Removed pointers

* Fixed priority

* Removed folder making from tag

* Changed sub-directory generation

* Adjust label names in preferences

This code:

  local oldCL = createLabel
  function createLabel(name, posX, posY, width, height, fillBackground)
    oldCL(name, 0, 0, 0, 0, fillBackground)
    moveWindow(name, posX, posY)
    resizeWindow(name, width, height)
  end

Doesn't understand the new parent argument, and thus no label gets created.

* Add missing Geyser file to mudet.pro

* fix bloody typo

Co-authored-by: Slobodan Terzić <githzerai06@gmail.com>

This file was added by accident.

* Adding a bunch of tests for Other.lua
* Typo that bugged me

* Fix createStopWatch(nil) potentially crashing Mudlet

Also "name as string or autostart as boolean are optional, got nil!" sounds pretty confusing to me, so I've edited out the optional part, like with other messages. If it's optional, then don't complain that you got nil...

* Add note per review

Merge release-4.5 branch into development

Merge release-4.6 branch into development

* Update mudlet.pro

* Update CMakeLists.txt

* Fixing typo in error message for table.is_empty
* Adding tests and increasing coverage for TableUtils spec.

* Fix issue in _comp where table B containing keys table A does not have would still return true
* Move table.size test to after the existing pairs loop so it will potentially fail faster
* Further optimization to avoid relooping table A to get its size, instead track it while we're looping A anyway
* Finished all pending tests in TableUtils_spec as most of them used the _comp function internally, and testing they still work increases confidence in this change
* Removing duplicate but also strangely empty test

…release branches (#3412)

* Adjusting items in StringUtils to support the : function notation
* Adding StringUtils spec, to make sure that the changes I made didn't break anything
* Adjusting one test and adding one last one

* fix segfault with Composer window being left open

* Added closeEvent to Composer window which clears pointer to itself in cTelnet when it gets closed without pushing save/cancel (such as clicking X or right-click and close)
* Added to cTelnet destructor saying to close Composer window

* Use QPointer and set flag to delete on close (#8)

* Update dlgComposer.h

* Update dlgComposer.cpp

* Update ctelnet.cpp

* Update ctelnet.h

* Update dlgComposer.cpp

* missed part (#9)

* Update ctelnet.cpp

* Update ctelnet.cpp

* flyoutLabel fixes

Fixes some issues with flyoutLabels

* remove closeNest and Geyser.Label.currentLabel leaving menu closes whole menu

Made some changes that closeNest and Geyser.Label.currentLabel aren't now needed anymore
leaving an menu will close the whole menu now as expected