Since Resource::makeStream returns a MemoryReadStream which will
not attempt to free the resource memory, it is fine to always
dispose those streams and get rid of the separate resourceStream
property, which was a holdover from some past WIP resource design
which no longer exists.
makeSOLStream was leaking the SeekableSubReadStream object it
creates itself if it was not called with DisposeAfterUse::YES. That
substream is an implementation detail which should not rely on
the caller to be destroyed.
Since these later SCI32 games weren't really designed to work with
subtitles-only message mode, if this doesn't work consistently or
breaks the games a lot in other places, the subtitles-only message
type could possibly be implemented in some other way, like by
messing with the game's speech volume global instead.
Revise/Simplify old optional text rendering
Set things up a bit for selected text rendering
Add MacTextWindow::clearText and MacTextWindow::setSelection
Add MacTextWindow::appendText variant that accepts a MacFont
Currently the font specified in the constructor is not
being used, I explain why in gui.cpp, line 182.
Use the 2 defines in gui.h to specify whether:
1) The new text rendering in general is used
(USE_NEW_TEXT_RENDERER)
2) The MacTextWindow class is used
for rendering (USE_MACTEXTWINDOW)
When you change the Arboretum to Winter and exit, it changes to
a special FrozenArboretum room with the river and everything frozen.
Both the regular and frozen Arboretum rooms have a GondolierMixer..
when leavin one for the other, it's meant to do a two second fadeout
and then the new room's mixer takes over. This doesn't work for
ScummVM currently - it may be due to the simplicity of the QMixer
implementation. So to work around it for the forseeable future,
I've set the TurnOff times to 0 seconds so the old mixer's sounds
are immediately and properly stopped, letting the new mixer take over.
In b4c0be8b42d63cbf3c808be1a94839483f674ce9 keyboard events were
adjusted to send control characters to game scripts, which matches
how keyboard input works in SSCI. Unfortunately this broke games
using kMenuSelect because that kernel code was not expecting to
receive control characters.
Here is an amended list of known types of keyboard shortcuts, for
future reference:
* All games with text inputs (Ctrl+C clears text boxes)
* Most games using MenuBar, like QFG1EGA (Ctrl+P pauses the
game, Tab or Ctrl+I show inventory)
* QFG1VGA (Ctrl+S shows stats)
* Torin (Ctrl+N, Ctrl+O, Ctrl+S, etc. activate menu commands)
* LSL1VGA & LSL3 (Ctrl+Alt+X to bypass age check)
* Most in-game debuggers (Alt+T for teleport)
The shortcut handling code is still not 100% accurate since there
are some edge cases that are not implemented (e.g. in DOS/SSCI,
Shift+Ctrl+<key> usually sends the same key information as
Ctrl+<key>, but not if <key> is Tab), but it should now be working
in a consistent and rational manner for end-users.
Used by:
* All games with text inputs (Ctrl+C clears text boxes)
* QFG1VGA (Ctrl+S shows stats)
* Torin (Ctrl+N, Ctrl+O, Ctrl+S, etc. activate menu commands)
The branch that used to shift control keys for SCI versions <=
SCI_VERSION_1_MIDDLE was bogus; history indicates it was intended
to be used to fix backends that sent control characters when
Ctrl+Alt were used together, but that case is already handled by
the Alt-checking code just prior to that code. Games expect to
receive control characters only when Ctrl is the only active
modifier, and this seems to be consistent across all versions of
SCI engine from SCI0 all the way through at least SCI2.1.
Fixes Trac#6703, Trac#9837.
These objects should have been initialized only during the first
pass. Double-initialization does not cause any visible problem
problem during normal operation (mostly it just causes memory
waste by making Object::_baseVars/_baseMethod double up their
data), but could have silently allowed games to receive bogus data
for an out-of-bounds property or method index, instead of raising
an error.
1. In SCI0/1, selectors and offsets in the method block are
stored contiguously (all selectors, then all offsets), with a
null separator between the two runs. All the later versions of
SCI instead interleave selectors & offsets. Since these values
are already being copied into a new array anyway, code for
reading method selectors/offsets is now simplified by
interleaving this data when it is written into _baseMethod
for SCI0/1, so the same equation for retrieving method
selectors/offsets can be used across all SCI versions.
2. In SCI1.1-2.1 branch, the method count was being copied into
the first entry of the array, which meant that SCI1.1-2.1 had
extra code for dealing with the fact that the first entry was
not an entry. This has been fixed, and the extra code removed.
3. Data was being overread into _baseMethod in all games SCI0-2.1.
(SCI0/1 had an extra magic value of 2, and SCI1.1-2.1 had an
extra magic value of 3). Reviewing history, it's not clear why
this happened, other than that it appears to have been
introduced at 7b0760f1bc5c28abcede041a6e3930f84ff3d319. My best
guess is that this was a confusion between byte count and record
count, where the intent was to read an extra 2 bytes for the
null separator in SCI0/1, but it actually read 2 records
instead. (I do not have a guess on why SCI1.1 ended up with a
3.) This overreading has been removed.
