Conditions are evaluated left to right. If you want short-
circuiting to protect from invalid read, you have to do the
range checks before the pointer dereferencing.
The EGA version of Gobliiins, similar to Little Red Riding Hood,
claims a few resources are larger than they actually are. The original
happily reads past the resource structure, but we'll instead fix
the size of the resource after loading.
This fixes bug #7162.
GCC 6's -Wmisleading-indentation warns about the debugC() statement
being wrongly indented, as if belonging into the else's branch, which
it clearly doesn't.
It shouldn't be necessary anymore, since initOPL() cleans everything.
If, however, suddenly a piece of music sounds weird in a Gob game,
this is the place to look for.
This fixes a race condition between OPL::reset() and the callback
timer.
This is what the original game does too, and the scripts query for
all-uppercase CD track names.
Interestingly, this being wrong has been mostly inconsequential. It
does, however, manifest in one bug in the underwater wreck level of
Gobliins 2. After calling the Moray Eel by lighting the lamp with
Winkle, speaking to said eel has the scripts wait for the currently
playing background track to end before playing the voice lines,
instead of ending the audio track forcefully. The track is only about
a minute long, so it's "only" annoying, not a game-stopper.
The scripts also try to compare the CD track name with some different,
all-uppercase names, so this is possibly relevant in some other places
as well. No such bug report exists at the moment, though.
See also the forum post with the bug report:
http://forums.scummvm.org/viewtopic.php?p=81733#81733
A somewhat related bug report is #2999 "GOB2 : Moray Eel Game Freeze
(Wreck Scene)" (<http://sourceforge.net/p/scummvm/bugs/2999/>). At
that time, the script expression parser still did a stricmp for the
NEQ expression (which is wrong), and that made the bug not trigger.
Commit 5c58b9a3a4, which fixed that
incorrect NEQ behaviour, then made this bug here appear.
This is a left-over vestige of when the gob engine recorded the size
of each variable currently in use, because it byte-swapped all
variables on save/load depending on the machine ScummVM ran on.
The gob engine doesn't do that all anymore (instead, the variables are
always stored in the endianness of the original game), so
getAddressOff8() doesn't store "this is an 8-bit variable" anymore,
making the call a NOP.
This fixes Coverity Scan issue #1267142.
And clip to [0, width), [0, height) before drawing.
This fixes bug #6864, which is a regression I introduced in
51fd528fe5 when I changed all
the drawing code to use the Surface class.
I thought that having unsigned coordinates makes sense, but
for some reason, Fascination sets _destSpriteX (which maps
to left in fillRect()) to -1, expecting the drawing code to
clip.
Quoth Luc Schrijvers (Begasus) on scummvm-devel:
"Haiku is stepping up and there is a recipe (script to build
scummvm) in the haikuporters git repo [...] I'll still do a
full build and make it available for download that could be
upped on your download page when ready"
Draw::_vgaSmallPalette was supposed to be a 16-color VGA
palette / EGA palette, but we've been using the first 16
entries of Draw::_vgaPalette for that for ages now.
The only parts where we didn't were codepaths never
actually taken in the gob games, so for the sake of
clarity (and fixing Coverity issues), this commit removes
those as well.
Fixes Coverity issues 1192648 and 1192649.
This fixes a Protracker module in the OS/2 version of
Hopkins FBI (bug #3612101). In row 0x30 of the first
pattern, the set channel offset effect in the fourth track
pushes the offset past the sample (repeat) length.
This is not error; the mixing function already handles this
case flawlessly. No assert() is necessary there.
Sven Hesse