If an early file in the game's signature list has a hash/size
mismatch, it is still necessary to continue to check the rest of
the candidate files for existence, since the non-existence of
candidate files is supposed to disqualify a game description as
matching a game to an unknown variant.
By quitting the file check early, the detector had been allowing
descriptions to randomly match if there happened to be an early
file in the detection list with the right name but wrong hash/size,
even if some of the other signature files did not exist at all.
Previous patches that removed shifts of constant negative values
to eliminate UB were valid, but did not correct all places where
this engine was potentially bit shifting negative values. There is
no reason to not just use multiplication and division and let the
compiler make the right choice for optimisation for an
architecture, so that is what this patch does.
This improves the output of static assertions in all compilers,
and prevents problems in MSVC 2015 where the mechanism for
triggering a compilation error in C++98 mode may cause errors when
that compiler builds in release mode.
Fixes Trac#10154.
Fixes out of bounds reads in the Myst ME intro videos.
OOB reads may happen because:
- The bitstream implementation reads 4 bytes at a time, and the buffer
size is not guaranteed to be a multiple of 4.
- The huffman parsing code reads a fixed amount of bits when it
sometimes needs not all of them.
Also fixed bits vs bytes mismatch for the size parameter of the calls to
the bitstream constructor, and removed a few unnecessary heap
allocations.
Fixes#10220.
While usage of these pointers was technically safe because they
were read through an alignment-aware API, taking the address of an
unaligned pointer was generating warnings in Clang, and is not
strictly necessary here. This change solves the warning and also
protects this code from any future change that might cause it to
start reading unsafely.
Warnings are as follows:
"In the GNU C Library, "minor" is defined by <sys/sysmacros.h>.
For historical compatibility, it is currently defined by
<sys/types.h> as well, but we plan toremove this soon.
To use "minor", include <sys/sysmacros.h> directly. If you did
not intend to use a system-defined macro "minor", you should
undefine it after including <sys/types.h>"
Passing overlapping buffers to C standard library memcpy, strcpy,
and strncpy is undefined behavior. In SSCI these operations would
perform a forward copy, and most stdlib implementations do the
same, but at least newer Linux glibc on x86 copies bytes in
reverse, so just using the standard library on this platform
results in broken output.
Because SSCI used a blind forward copy instead of memmove for
overlapping copy operations, this patch implements an explicit
forward copy to ensure that overlapping copies continue to operate
the same as in SSCI.
This fixes the Island of Dr. Brain v1.1 flamingo puzzle
(script 185, flamingos::init, localCall 4c3) on platforms that do
not perform forward copy in memcpy/strcpy/strncpy.
Thanks to @moralrecordings for research on this bug and an initial
patch using memmove.
Closes gh-1034.
Since version 2.3.12, freetype-config adds SYSROOT to all paths.
If we pass a --prefix that already includes SYSROOT, this will return
a duplicate SYSROOT. This patch detects that and adjusts prefix
accordingly.
Folks are confused about the new behaviour where the mouse is not
restricted to the game area in fullscreen, which is understandable.
This changes mouseIsGrabbed to use SDL directly in order to avoid
making changes to the user preference in the _inputGrabState.
Otherwise we'd either clobber the user's previous windowed mouse
grab preference, or require maintaining a second variable just to
track the original state, when we can have SDL do that for us.
This code was assuming that retval points to the start of the next
instruction, which is only true if the current parameter is the last
one. This fixes op_call printing.
I noticed that in Maemo the cursor was offset after the SDL refactoring
in de2bbe3b97
In Maemo when entering fullscreen, ScummVM receives a SDL_VIDEORESIZE
event with the native touchscreen resolution rather than the current
window size. This causes a call to notifyResize.
Before that refactoring, notifyResize did nothing (in SDL1).
Now it calls handleResize but doesn't actually set a new video mode.
This messes up the coordinate mapping, causing an overcorrection for
the cursor position.
