Due to the need to calculate the audio duration with millisecond
precision from the file size, it is possible to overflow a 32-bit
integer in games with long background audio loops, like RAMA,
during the calculation of the duration.
It is also not necessary to give some framerate here, so eliminate
the unnecessary explicit Timestamp construction with the second
argument.
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.
DPCM decompression algorithms in SSCI operate directly on 8- and
16-bit registers, so any sample that ends up being out-of-range
during decompression gets wrapped by the CPU, not clipped.
This does not fix any known problem with AUD files, but there are
some VMDs (e.g. GK2 5280.VMD) which are known to contain OOR
samples. Making this code more accurate should prevent trouble
with any other similar files.
This provides a complete implementation of kDoAudio through
SCI2.1mid, plus partial implementation of SCI3 features.
Digital audio calls shunted through kDoSound have also been
updated to go through the SCI32 audio mixer, though these shunts
are a bit hacky because the ScummVM implementation of kDoSound
does not currently match how SSCI kDoSound is designed.
It is probably possible in the future to just replace the SCI1.1
audio code (audio.cpp) with the new SCI32 code, since the major
differences seem to be that (1) SCI1.1 only supported one digital
audio playback channel (this is configurable already), (2) it
had extra commands for CD audio playback and queued sample
playback.
Colin Snover
Thierry Crozat
Filippos Karapetis