/*

Simple DirectMedia Layer

Copyright (C) 1997-2024 Sam Lantinga <[email protected]>

This software is provided 'as-is', without any express or implied

warranty. In no event will the authors be held liable for any damages

arising from the use of this software.

Permission is granted to anyone to use this software for any purpose,

including commercial applications, and to alter it and redistribute it

freely, subject to the following restrictions:

1. The origin of this software must not be misrepresented; you must not

claim that you wrote the original software. If you use this software

in a product, an acknowledgment in the product documentation would be

appreciated but is not required.

2. Altered source versions must be plainly marked as such, and must not be

misrepresented as being the original software.

3. This notice may not be removed or altered from any source distribution.

*/

#include "../SDL_internal.h"

#ifdef HAVE_LIMITS_H

#include <limits.h>

#endif

#ifndef INT_MAX

/* Make a lucky guess. */

#define INT_MAX SDL_MAX_SINT32

#endif

#ifndef SIZE_MAX

#define SIZE_MAX ((size_t)-1)

#endif

/* Microsoft WAVE file loading routines */

#include "SDL_hints.h"

#include "SDL_audio.h"

#include "SDL_wave.h"

#include "SDL_audio_c.h"

/* Reads the value stored at the location of the f1 pointer, multiplies it

* with the second argument and then stores the result to f1.

* Returns 0 on success, or -1 if the multiplication overflows, in which case f1

* does not get modified.

*/

static int SafeMult(size_t *f1, size_t f2)

{

if (*f1 > 0 && SIZE_MAX / *f1 <= f2) {

return -1;

}

*f1 *= f2;

return 0;

}

typedef struct ADPCM_DecoderState

{

Uint32 channels; /* Number of channels. */

size_t blocksize; /* Size of an ADPCM block in bytes. */

size_t blockheadersize; /* Size of an ADPCM block header in bytes. */

size_t samplesperblock; /* Number of samples per channel in an ADPCM block. */

size_t framesize; /* Size of a sample frame (16-bit PCM) in bytes. */

Sint64 framestotal; /* Total number of sample frames. */

Sint64 framesleft; /* Number of sample frames still to be decoded. */

void *ddata; /* Decoder data from initialization. */

void *cstate; /* Decoding state for each channel. */

/* ADPCM data. */

struct

{

Uint8 *data;

size_t size;

size_t pos;

} input;

/* Current ADPCM block in the ADPCM data above. */

struct

{

Uint8 *data;

size_t size;

size_t pos;

} block;

/* Decoded 16-bit PCM data. */

struct

{

Sint16 *data;

size_t size;

size_t pos;

} output;

} ADPCM_DecoderState;

typedef struct MS_ADPCM_CoeffData

{

Uint16 coeffcount;

Sint16 *coeff;

Sint16 aligndummy; /* Has to be last member. */

} MS_ADPCM_CoeffData;

typedef struct MS_ADPCM_ChannelState

{

Uint16 delta;

Sint16 coeff1;

Sint16 coeff2;

} MS_ADPCM_ChannelState;

#ifdef SDL_WAVE_DEBUG_LOG_FORMAT

static void WaveDebugLogFormat(WaveFile *file)

{

WaveFormat *format = &file->format;

const char *fmtstr = "WAVE file: %s, %u Hz, %s, %u bits, %u %s/s";

const char *waveformat, *wavechannel, *wavebpsunit = "B";

Uint32 wavebps = format->byterate;

char channelstr[64];

SDL_zeroa(channelstr);

switch (format->encoding) {

case PCM_CODE:

waveformat = "PCM";

break;

case IEEE_FLOAT_CODE:

waveformat = "IEEE Float";

break;

case ALAW_CODE:

waveformat = "A-law";

break;

case MULAW_CODE:

waveformat = "\xc2\xb5-law";

break;

case MS_ADPCM_CODE:

waveformat = "MS ADPCM";

break;

case IMA_ADPCM_CODE:

waveformat = "IMA ADPCM";

break;

default:

waveformat = "Unknown";

break;

}

#define SDL_WAVE_DEBUG_CHANNELCFG(STR, CODE) \

case CODE: \

wavechannel = STR; \

break;

#define SDL_WAVE_DEBUG_CHANNELSTR(STR, CODE) \

if (format->channelmask & CODE) { \

SDL_strlcat(channelstr, channelstr[0] ? "-" STR : STR, sizeof(channelstr)); \

}

if (format->formattag == EXTENSIBLE_CODE && format->channelmask > 0) {

switch (format->channelmask) {

SDL_WAVE_DEBUG_CHANNELCFG("1.0 Mono", 0x4)

SDL_WAVE_DEBUG_CHANNELCFG("1.1 Mono", 0xc)

SDL_WAVE_DEBUG_CHANNELCFG("2.0 Stereo", 0x3)

SDL_WAVE_DEBUG_CHANNELCFG("2.1 Stereo", 0xb)

SDL_WAVE_DEBUG_CHANNELCFG("3.0 Stereo", 0x7)

SDL_WAVE_DEBUG_CHANNELCFG("3.1 Stereo", 0xf)

SDL_WAVE_DEBUG_CHANNELCFG("3.0 Surround", 0x103)

SDL_WAVE_DEBUG_CHANNELCFG("3.1 Surround", 0x10b)

SDL_WAVE_DEBUG_CHANNELCFG("4.0 Quad", 0x33)

SDL_WAVE_DEBUG_CHANNELCFG("4.1 Quad", 0x3b)

SDL_WAVE_DEBUG_CHANNELCFG("4.0 Surround", 0x107)

SDL_WAVE_DEBUG_CHANNELCFG("4.1 Surround", 0x10f)

SDL_WAVE_DEBUG_CHANNELCFG("5.0", 0x37)

SDL_WAVE_DEBUG_CHANNELCFG("5.1", 0x3f)

SDL_WAVE_DEBUG_CHANNELCFG("5.0 Side", 0x607)

SDL_WAVE_DEBUG_CHANNELCFG("5.1 Side", 0x60f)

SDL_WAVE_DEBUG_CHANNELCFG("6.0", 0x137)

SDL_WAVE_DEBUG_CHANNELCFG("6.1", 0x13f)

SDL_WAVE_DEBUG_CHANNELCFG("6.0 Side", 0x707)

SDL_WAVE_DEBUG_CHANNELCFG("6.1 Side", 0x70f)

SDL_WAVE_DEBUG_CHANNELCFG("7.0", 0xf7)

SDL_WAVE_DEBUG_CHANNELCFG("7.1", 0xff)

SDL_WAVE_DEBUG_CHANNELCFG("7.0 Side", 0x6c7)

SDL_WAVE_DEBUG_CHANNELCFG("7.1 Side", 0x6cf)

SDL_WAVE_DEBUG_CHANNELCFG("7.0 Surround", 0x637)

SDL_WAVE_DEBUG_CHANNELCFG("7.1 Surround", 0x63f)

SDL_WAVE_DEBUG_CHANNELCFG("9.0 Surround", 0x5637)

SDL_WAVE_DEBUG_CHANNELCFG("9.1 Surround", 0x563f)

SDL_WAVE_DEBUG_CHANNELCFG("11.0 Surround", 0x56f7)

SDL_WAVE_DEBUG_CHANNELCFG("11.1 Surround", 0x56ff)

default:

SDL_WAVE_DEBUG_CHANNELSTR("FL", 0x1)

SDL_WAVE_DEBUG_CHANNELSTR("FR", 0x2)

SDL_WAVE_DEBUG_CHANNELSTR("FC", 0x4)

SDL_WAVE_DEBUG_CHANNELSTR("LF", 0x8)

SDL_WAVE_DEBUG_CHANNELSTR("BL", 0x10)

SDL_WAVE_DEBUG_CHANNELSTR("BR", 0x20)

SDL_WAVE_DEBUG_CHANNELSTR("FLC", 0x40)

SDL_WAVE_DEBUG_CHANNELSTR("FRC", 0x80)

SDL_WAVE_DEBUG_CHANNELSTR("BC", 0x100)

SDL_WAVE_DEBUG_CHANNELSTR("SL", 0x200)

SDL_WAVE_DEBUG_CHANNELSTR("SR", 0x400)

SDL_WAVE_DEBUG_CHANNELSTR("TC", 0x800)

SDL_WAVE_DEBUG_CHANNELSTR("TFL", 0x1000)

SDL_WAVE_DEBUG_CHANNELSTR("TFC", 0x2000)

SDL_WAVE_DEBUG_CHANNELSTR("TFR", 0x4000)

SDL_WAVE_DEBUG_CHANNELSTR("TBL", 0x8000)

SDL_WAVE_DEBUG_CHANNELSTR("TBC", 0x10000)

SDL_WAVE_DEBUG_CHANNELSTR("TBR", 0x20000)

break;

}

} else {

switch (format->channels) {

default:

if (SDL_snprintf(channelstr, sizeof(channelstr), "%u channels", format->channels) >= 0) {

wavechannel = channelstr;

break;

}

case 0:

wavechannel = "Unknown";

break;

case 1:

wavechannel = "Mono";

break;

case 2:

wavechannel = "Setero";

break;

}

}

#undef SDL_WAVE_DEBUG_CHANNELCFG

#undef SDL_WAVE_DEBUG_CHANNELSTR

if (wavebps >= 1024) {

wavebpsunit = "KiB";

wavebps = wavebps / 1024 + (wavebps & 0x3ff ? 1 : 0);

}

SDL_LogDebug(SDL_LOG_CATEGORY_AUDIO, fmtstr, waveformat, format->frequency, wavechannel, format->bitspersample, wavebps, wavebpsunit);

}

#endif

#ifdef SDL_WAVE_DEBUG_DUMP_FORMAT

static void WaveDebugDumpFormat(WaveFile *file, Uint32 rifflen, Uint32 fmtlen, Uint32 datalen)

{

WaveFormat *format = &file->format;

const char *fmtstr1 = "WAVE chunk dump:\n"

"-------------------------------------------\n"

"RIFF %11u\n"

"-------------------------------------------\n"

" fmt %11u\n"

" wFormatTag 0x%04x\n"

" nChannels %11u\n"

" nSamplesPerSec %11u\n"

" nAvgBytesPerSec %11u\n"

" nBlockAlign %11u\n";

const char *fmtstr2 = " wBitsPerSample %11u\n";

const char *fmtstr3 = " cbSize %11u\n";

const char *fmtstr4a = " wValidBitsPerSample %11u\n";

const char *fmtstr4b = " wSamplesPerBlock %11u\n";

const char *fmtstr5 = " dwChannelMask 0x%08x\n"

" SubFormat\n"

" %08x-%04x-%04x-%02x%02x%02x%02x%02x%02x%02x%02x\n";

const char *fmtstr6 = "-------------------------------------------\n"

" fact\n"

" dwSampleLength %11u\n";

const char *fmtstr7 = "-------------------------------------------\n"

" data %11u\n"

"-------------------------------------------\n";

char *dumpstr;

size_t dumppos = 0;

const size_t bufsize = 1024;

int res;

dumpstr = SDL_malloc(bufsize);

if (!dumpstr) {

return;

}

dumpstr[0] = 0;

res = SDL_snprintf(dumpstr, bufsize, fmtstr1, rifflen, fmtlen, format->formattag, format->channels, format->frequency, format->byterate, format->blockalign);

dumppos += res > 0 ? res : 0;

if (fmtlen >= 16) {

res = SDL_snprintf(dumpstr + dumppos, bufsize - dumppos, fmtstr2, format->bitspersample);

dumppos += res > 0 ? res : 0;

}

if (fmtlen >= 18) {

res = SDL_snprintf(dumpstr + dumppos, bufsize - dumppos, fmtstr3, format->extsize);

dumppos += res > 0 ? res : 0;

}

if (format->formattag == EXTENSIBLE_CODE && fmtlen >= 40 && format->extsize >= 22) {

const Uint8 *g = format->subformat;

const Uint32 g1 = g[0] | ((Uint32)g[1] << 8) | ((Uint32)g[2] << 16) | ((Uint32)g[3] << 24);

const Uint32 g2 = g[4] | ((Uint32)g[5] << 8);

const Uint32 g3 = g[6] | ((Uint32)g[7] << 8);

switch (format->encoding) {

default:

res = SDL_snprintf(dumpstr + dumppos, bufsize - dumppos, fmtstr4a, format->validsamplebits);

dumppos += res > 0 ? res : 0;

break;

case MS_ADPCM_CODE:

case IMA_ADPCM_CODE:

res = SDL_snprintf(dumpstr + dumppos, bufsize - dumppos, fmtstr4b, format->samplesperblock);

dumppos += res > 0 ? res : 0;

break;

}

res = SDL_snprintf(dumpstr + dumppos, bufsize - dumppos, fmtstr5, format->channelmask, g1, g2, g3, g[8], g[9], g[10], g[11], g[12], g[13], g[14], g[15]);

dumppos += res > 0 ? res : 0;

} else {

switch (format->encoding) {

case MS_ADPCM_CODE:

case IMA_ADPCM_CODE:

if (fmtlen >= 20 && format->extsize >= 2) {

res = SDL_snprintf(dumpstr + dumppos, bufsize - dumppos, fmtstr4b, format->samplesperblock);

dumppos += res > 0 ? res : 0;

}

break;

}

}

if (file->fact.status >= 1) {

res = SDL_snprintf(dumpstr + dumppos, bufsize - dumppos, fmtstr6, file->fact.samplelength);

dumppos += res > 0 ? res : 0;

}

res = SDL_snprintf(dumpstr + dumppos, bufsize - dumppos, fmtstr7, datalen);

dumppos += res > 0 ? res : 0;

SDL_LogDebug(SDL_LOG_CATEGORY_AUDIO, "%s", dumpstr);

SDL_free(dumpstr);

}

#endif

static Sint64 WaveAdjustToFactValue(WaveFile *file, Sint64 sampleframes)

{

if (file->fact.status == 2) {

if (file->facthint == FactStrict && sampleframes < file->fact.samplelength) {

return SDL_SetError("Invalid number of sample frames in WAVE fact chunk (too many)");

} else if (sampleframes > file->fact.samplelength) {

return file->fact.samplelength;

}

}

return sampleframes;

}

static int MS_ADPCM_CalculateSampleFrames(WaveFile *file, size_t datalength)

{

WaveFormat *format = &file->format;

const size_t blockheadersize = (size_t)file->format.channels * 7;

const size_t availableblocks = datalength / file->format.blockalign;

const size_t blockframebitsize = (size_t)file->format.bitspersample * file->format.channels;

const size_t trailingdata = datalength % file->format.blockalign;

if (file->trunchint == TruncVeryStrict || file->trunchint == TruncStrict) {

/* The size of the data chunk must be a multiple of the block size. */

if (datalength < blockheadersize || trailingdata > 0) {

return SDL_SetError("Truncated MS ADPCM block");

}

}

/* Calculate number of sample frames that will be decoded. */

file->sampleframes = (Sint64)availableblocks * format->samplesperblock;

if (trailingdata > 0) {

/* The last block is truncated. Check if we can get any samples out of it. */

if (file->trunchint == TruncDropFrame) {

/* Drop incomplete sample frame. */

if (trailingdata >= blockheadersize) {

size_t trailingsamples = 2 + (trailingdata - blockheadersize) * 8 / blockframebitsize;

if (trailingsamples > format->samplesperblock) {

trailingsamples = format->samplesperblock;

}

file->sampleframes += trailingsamples;

}

}

}

file->sampleframes = WaveAdjustToFactValue(file, file->sampleframes);

if (file->sampleframes < 0) {

return -1;

}

return 0;

}

static int MS_ADPCM_Init(WaveFile *file, size_t datalength)

{

WaveFormat *format = &file->format;

WaveChunk *chunk = &file->chunk;

const size_t blockheadersize = (size_t)format->channels * 7;

const size_t blockdatasize = (size_t)format->blockalign - blockheadersize;

const size_t blockframebitsize = (size_t)format->bitspersample * format->channels;

const size_t blockdatasamples = (blockdatasize * 8) / blockframebitsize;

const Sint16 presetcoeffs[14] = { 256, 0, 512, -256, 0, 0, 192, 64, 240, 0, 460, -208, 392, -232 };

size_t i, coeffcount;

MS_ADPCM_CoeffData *coeffdata;

/* Sanity checks. */

/* While it's clear how IMA ADPCM handles more than two channels, the nibble

* order of MS ADPCM makes it awkward. The Standards Update does not talk

* about supporting more than stereo anyway.

*/

if (format->channels > 2) {

return SDL_SetError("Invalid number of channels");

}

if (format->bitspersample != 4) {

return SDL_SetError("Invalid MS ADPCM bits per sample of %u", (unsigned int)format->bitspersample);

}

/* The block size must be big enough to contain the block header. */

if (format->blockalign < blockheadersize) {

return SDL_SetError("Invalid MS ADPCM block size (nBlockAlign)");

}

if (format->formattag == EXTENSIBLE_CODE) {

/* Does have a GUID (like all format tags), but there's no specification

* for how the data is packed into the extensible header. Making

* assumptions here could lead to new formats nobody wants to support.

*/

return SDL_SetError("MS ADPCM with the extensible header is not supported");

}

/* There are wSamplesPerBlock, wNumCoef, and at least 7 coefficient pairs in

* the extended part of the header.

*/

if (chunk->size < 22) {

return SDL_SetError("Could not read MS ADPCM format header");

}

format->samplesperblock = chunk->data[18] | ((Uint16)chunk->data[19] << 8);

/* Number of coefficient pairs. A pair has two 16-bit integers. */

coeffcount = chunk->data[20] | ((size_t)chunk->data[21] << 8);

/* bPredictor, the integer offset into the coefficients array, is only

* 8 bits. It can only address the first 256 coefficients. Let's limit

* the count number here.

*/

if (coeffcount > 256) {

coeffcount = 256;

}

if (chunk->size < 22 + coeffcount * 4) {

return SDL_SetError("Could not read custom coefficients in MS ADPCM format header");

} else if (format->extsize < 4 + coeffcount * 4) {

return SDL_SetError("Invalid MS ADPCM format header (too small)");

} else if (coeffcount < 7) {

return SDL_SetError("Missing required coefficients in MS ADPCM format header");

}

coeffdata = (MS_ADPCM_CoeffData *)SDL_malloc(sizeof(MS_ADPCM_CoeffData) + coeffcount * 4);

file->decoderdata = coeffdata; /* Freed in cleanup. */

if (!coeffdata) {

return SDL_OutOfMemory();

}

coeffdata->coeff = &coeffdata->aligndummy;

coeffdata->coeffcount = (Uint16)coeffcount;

/* Copy the 16-bit pairs. */

for (i = 0; i < coeffcount * 2; i++) {

Sint32 c = chunk->data[22 + i * 2] | ((Sint32)chunk->data[23 + i * 2] << 8);

if (c >= 0x8000) {

c -= 0x10000;

}

if (i < 14 && c != presetcoeffs[i]) {

return SDL_SetError("Wrong preset coefficients in MS ADPCM format header");

}

coeffdata->coeff[i] = (Sint16)c;

}

/* Technically, wSamplesPerBlock is required, but we have all the

* information in the other fields to calculate it, if it's zero.

*/

if (format->samplesperblock == 0) {

/* Let's be nice to the encoders that didn't know how to fill this.

* The Standards Update calculates it this way:

*

* x = Block size (in bits) minus header size (in bits)

* y = Bit depth multiplied by channel count

* z = Number of samples per channel in block header

* wSamplesPerBlock = x / y + z

*/

format->samplesperblock = (Uint32)blockdatasamples + 2;

}

/* nBlockAlign can be in conflict with wSamplesPerBlock. For example, if

* the number of samples doesn't fit into the block. The Standards Update

* also describes wSamplesPerBlock with a formula that makes it necessary to

* always fill the block with the maximum amount of samples, but this is not

* enforced here as there are no compatibility issues.

* A truncated block header with just one sample is not supported.

*/

if (format->samplesperblock == 1 || blockdatasamples < format->samplesperblock - 2) {

return SDL_SetError("Invalid number of samples per MS ADPCM block (wSamplesPerBlock)");

}

if (MS_ADPCM_CalculateSampleFrames(file, datalength) < 0) {

return -1;

}

return 0;

}

static Sint16 MS_ADPCM_ProcessNibble(MS_ADPCM_ChannelState *cstate, Sint32 sample1, Sint32 sample2, Uint8 nybble)

{

const Sint32 max_audioval = 32767;

const Sint32 min_audioval = -32768;

const Uint16 max_deltaval = 65535;

const Uint16 adaptive[] = {

230, 230, 230, 230, 307, 409, 512, 614,

768, 614, 512, 409, 307, 230, 230, 230

};

Sint32 new_sample;

Sint32 errordelta;

Uint32 delta = cstate->delta;

new_sample = (sample1 * cstate->coeff1 + sample2 * cstate->coeff2) / 256;

/* The nibble is a signed 4-bit error delta. */

errordelta = (Sint32)nybble - (nybble >= 0x08 ? 0x10 : 0);

new_sample += (Sint32)delta * errordelta;

if (new_sample < min_audioval) {

new_sample = min_audioval;

} else if (new_sample > max_audioval) {

new_sample = max_audioval;

}

delta = (delta * adaptive[nybble]) / 256;

if (delta < 16) {

delta = 16;

} else if (delta > max_deltaval) {

/* This issue is not described in the Standards Update and therefore

* undefined. It seems sensible to prevent overflows with a limit.

*/

delta = max_deltaval;

}

cstate->delta = (Uint16)delta;

return (Sint16)new_sample;

}

static int MS_ADPCM_DecodeBlockHeader(ADPCM_DecoderState *state)

{

Uint8 coeffindex;

const Uint32 channels = state->channels;

Sint32 sample;

Uint32 c;

MS_ADPCM_ChannelState *cstate = (MS_ADPCM_ChannelState *)state->cstate;

MS_ADPCM_CoeffData *ddata = (MS_ADPCM_CoeffData *)state->ddata;

for (c = 0; c < channels; c++) {

size_t o = c;

/* Load the coefficient pair into the channel state. */

coeffindex = state->block.data[o];

if (coeffindex > ddata->coeffcount) {

return SDL_SetError("Invalid MS ADPCM coefficient index in block header");

}

cstate[c].coeff1 = ddata->coeff[coeffindex * 2];

cstate[c].coeff2 = ddata->coeff[coeffindex * 2 + 1];

/* Initial delta value. */

o = (size_t)channels + c * 2;

cstate[c].delta = state->block.data[o] | ((Uint16)state->block.data[o + 1] << 8);

/* Load the samples from the header. Interestingly, the sample later in

* the output stream comes first.

*/

o = (size_t)channels * 3 + c * 2;

sample = state->block.data[o] | ((Sint32)state->block.data[o + 1] << 8);

if (sample >= 0x8000) {

sample -= 0x10000;

}

state->output.data[state->output.pos + channels] = (Sint16)sample;

o = (size_t)channels * 5 + c * 2;

sample = state->block.data[o] | ((Sint32)state->block.data[o + 1] << 8);

if (sample >= 0x8000) {

sample -= 0x10000;

}

state->output.data[state->output.pos] = (Sint16)sample;

state->output.pos++;

}

state->block.pos += state->blockheadersize;

/* Skip second sample frame that came from the header. */

state->output.pos += state->channels;

/* Header provided two sample frames. */

state->framesleft -= 2;

return 0;

}

/* Decodes the data of the MS ADPCM block. Decoding will stop if a block is too

* short, returning with none or partially decoded data. The partial data

* will always contain full sample frames (same sample count for each channel).

* Incomplete sample frames are discarded.

*/

static int MS_ADPCM_DecodeBlockData(ADPCM_DecoderState *state)

{

Uint16 nybble = 0;

Sint16 sample1, sample2;

const Uint32 channels = state->channels;

Uint32 c;

MS_ADPCM_ChannelState *cstate = (MS_ADPCM_ChannelState *)state->cstate;

size_t blockpos = state->block.pos;

size_t blocksize = state->block.size;

size_t outpos = state->output.pos;

Sint64 blockframesleft = state->samplesperblock - 2;

if (blockframesleft > state->framesleft) {

blockframesleft = state->framesleft;

}

while (blockframesleft > 0) {

for (c = 0; c < channels; c++) {

if (nybble & 0x4000) {

nybble <<= 4;

} else if (blockpos < blocksize) {

nybble = state->block.data[blockpos++] | 0x4000;

} else {

/* Out of input data. Drop the incomplete frame and return. */

state->output.pos = outpos - c;

return -1;

}

/* Load previous samples which may come from the block header. */

sample1 = state->output.data[outpos - channels];

sample2 = state->output.data[outpos - channels * 2];

sample1 = MS_ADPCM_ProcessNibble(cstate + c, sample1, sample2, (nybble >> 4) & 0x0f);

state->output.data[outpos++] = sample1;

}

state->framesleft--;

blockframesleft--;

}

state->output.pos = outpos;

return 0;

}

static int MS_ADPCM_Decode(WaveFile *file, Uint8 **audio_buf, Uint32 *audio_len)

{

int result;

size_t bytesleft, outputsize;

WaveChunk *chunk = &file->chunk;

ADPCM_DecoderState state;

MS_ADPCM_ChannelState cstate[2];

SDL_zero(state);

SDL_zeroa(cstate);

if (chunk->size != chunk->length) {

/* Could not read everything. Recalculate number of sample frames. */

if (MS_ADPCM_CalculateSampleFrames(file, chunk->size) < 0) {

return -1;

}

}

/* Nothing to decode, nothing to return. */

if (file->sampleframes == 0) {

*audio_buf = NULL;

*audio_len = 0;

return 0;

}

state.blocksize = file->format.blockalign;

state.channels = file->format.channels;

state.blockheadersize = (size_t)state.channels * 7;

state.samplesperblock = file->format.samplesperblock;

state.framesize = state.channels * sizeof(Sint16);

state.ddata = file->decoderdata;

state.framestotal = file->sampleframes;

state.framesleft = state.framestotal;

state.input.data = chunk->data;

state.input.size = chunk->size;

state.input.pos = 0;

/* The output size in bytes. May get modified if data is truncated. */

outputsize = (size_t)state.framestotal;

if (SafeMult(&outputsize, state.framesize)) {

return SDL_OutOfMemory();

} else if (outputsize > SDL_MAX_UINT32 || state.framestotal > SIZE_MAX) {

return SDL_SetError("WAVE file too big");

}

state.output.pos = 0;

state.output.size = outputsize / sizeof(Sint16);

state.output.data = (Sint16 *)SDL_calloc(1, outputsize);

if (!state.output.data) {

return SDL_OutOfMemory();

}

state.cstate = cstate;

/* Decode block by block. A truncated block will stop the decoding. */

bytesleft = state.input.size - state.input.pos;

while (state.framesleft > 0 && bytesleft >= state.blockheadersize) {

state.block.data = state.input.data + state.input.pos;

state.block.size = bytesleft < state.blocksize ? bytesleft : state.blocksize;

state.block.pos = 0;

if (state.output.size - state.output.pos < (Uint64)state.framesleft * state.channels) {

/* Somehow didn't allocate enough space for the output. */

SDL_free(state.output.data);

return SDL_SetError("Unexpected overflow in MS ADPCM decoder");

}

/* Initialize decoder with the values from the block header. */

result = MS_ADPCM_DecodeBlockHeader(&state);

if (result == -1) {

SDL_free(state.output.data);

return -1;

}

/* Decode the block data. It stores the samples directly in the output. */

result = MS_ADPCM_DecodeBlockData(&state);

if (result == -1) {

/* Unexpected end. Stop decoding and return partial data if necessary. */

if (file->trunchint == TruncVeryStrict || file->trunchint == TruncStrict) {

SDL_free(state.output.data);

return SDL_SetError("Truncated data chunk");

} else if (file->trunchint != TruncDropFrame) {

state.output.pos -= state.output.pos % (state.samplesperblock * state.channels);

}

outputsize = state.output.pos * sizeof(Sint16); /* Can't overflow, is always smaller. */

break;

}

state.input.pos += state.block.size;

bytesleft = state.input.size - state.input.pos;

}

*audio_buf = (Uint8 *)state.output.data;

*audio_len = (Uint32)outputsize;

return 0;

}

static int IMA_ADPCM_CalculateSampleFrames(WaveFile *file, size_t datalength)

{

WaveFormat *format = &file->format;

const size_t blockheadersize = (size_t)format->channels * 4;

const size_t subblockframesize = (size_t)format->channels * 4;

const size_t availableblocks = datalength / format->blockalign;

const size_t trailingdata = datalength % format->blockalign;

if (file->trunchint == TruncVeryStrict || file->trunchint == TruncStrict) {

/* The size of the data chunk must be a multiple of the block size. */

if (datalength < blockheadersize || trailingdata > 0) {

return SDL_SetError("Truncated IMA ADPCM block");

}

}

/* Calculate number of sample frames that will be decoded. */

file->sampleframes = (Uint64)availableblocks * format->samplesperblock;

if (trailingdata > 0) {

/* The last block is truncated. Check if we can get any samples out of it. */

if (file->trunchint == TruncDropFrame && trailingdata > blockheadersize - 2) {

/* The sample frame in the header of the truncated block is present.

* Drop incomplete sample frames.

*/

size_t trailingsamples = 1;

if (trailingdata > blockheadersize) {

/* More data following after the header. */

const size_t trailingblockdata = trailingdata - blockheadersize;

const size_t trailingsubblockdata = trailingblockdata % subblockframesize;

trailingsamples += (trailingblockdata / subblockframesize) * 8;

/* Due to the interleaved sub-blocks, the last 4 bytes determine

* how many samples of the truncated sub-block are lost.

*/

if (trailingsubblockdata > subblockframesize - 4) {

trailingsamples += (trailingsubblockdata % 4) * 2;

}

}

if (trailingsamples > format->samplesperblock) {

trailingsamples = format->samplesperblock;

}

file->sampleframes += trailingsamples;

}

}

file->sampleframes = WaveAdjustToFactValue(file, file->sampleframes);

if (file->sampleframes < 0) {

return -1;

}

return 0;

}

static int IMA_ADPCM_Init(WaveFile *file, size_t datalength)

{

WaveFormat *format = &file->format;

WaveChunk *chunk = &file->chunk;

const size_t blockheadersize = (size_t)format->channels * 4;

const size_t blockdatasize = (size_t)format->blockalign - blockheadersize;

const size_t blockframebitsize = (size_t)format->bitspersample * format->channels;

const size_t blockdatasamples = (blockdatasize * 8) / blockframebitsize;

/* Sanity checks. */

/* IMA ADPCM can also have 3-bit samples, but it's not supported by SDL at this time. */

if (format->bitspersample == 3) {

return SDL_SetError("3-bit IMA ADPCM currently not supported");

} else if (format->bitspersample != 4) {

return SDL_SetError("Invalid IMA ADPCM bits per sample of %u", (unsigned int)format->bitspersample);

}

/* The block size is required to be a multiple of 4 and it must be able to

* hold a block header.

*/

if (format->blockalign < blockheadersize || format->blockalign % 4) {

return SDL_SetError("Invalid IMA ADPCM block size (nBlockAlign)");

}

if (format->formattag == EXTENSIBLE_CODE) {

/* There's no specification for this, but it's basically the same

* format because the extensible header has wSampePerBlocks too.

*/

} else {

/* The Standards Update says there 'should' be 2 bytes for wSamplesPerBlock. */

if (chunk->size >= 20 && format->extsize >= 2) {

format->samplesperblock = chunk->data[18] | ((Uint16)chunk->data[19] << 8);

}

}

if (format->samplesperblock == 0) {

/* Field zero? No problem. We just assume the encoder packed the block.

* The specification calculates it this way:

*

* x = Block size (in bits) minus header size (in bits)

* y = Bit depth multiplied by channel count

* z = Number of samples per channel in header

* wSamplesPerBlock = x / y + z

*/

format->samplesperblock = (Uint32)blockdatasamples + 1;

}

/* nBlockAlign can be in conflict with wSamplesPerBlock. For example, if

* the number of samples doesn't fit into the block. The Standards Update

* also describes wSamplesPerBlock with a formula that makes it necessary

* to always fill the block with the maximum amount of samples, but this is

* not enforced here as there are no compatibility issues.

*/

if (blockdatasamples < format->samplesperblock - 1) {

return SDL_SetError("Invalid number of samples per IMA ADPCM block (wSamplesPerBlock)");

}

if (IMA_ADPCM_CalculateSampleFrames(file, datalength) < 0) {

return -1;

}

return 0;

}

static Sint16 IMA_ADPCM_ProcessNibble(Sint8 *cindex, Sint16 lastsample, Uint8 nybble)

{

const Sint32 max_audioval = 32767;

const Sint32 min_audioval = -32768;

const Sint8 index_table_4b[16] = {

-1, -1, -1, -1,

2, 4, 6, 8,

-1, -1, -1, -1,

2, 4, 6, 8

};

const Uint16 step_table[89] = {

7, 8, 9, 10, 11, 12, 13, 14, 16, 17, 19, 21, 23, 25, 28, 31,

34, 37, 41, 45, 50, 55, 60, 66, 73, 80, 88, 97, 107, 118, 130,

143, 157, 173, 190, 209, 230, 253, 279, 307, 337, 371, 408,

449, 494, 544, 598, 658, 724, 796, 876, 963, 1060, 1166, 1282,

1411, 1552, 1707, 1878, 2066, 2272, 2499, 2749, 3024, 3327,

3660, 4026, 4428, 4871, 5358, 5894, 6484, 7132, 7845, 8630,

9493, 10442, 11487, 12635, 13899, 15289, 16818, 18500, 20350,

22385, 24623, 27086, 29794, 32767

};

Uint32 step;

Sint32 sample, delta;

Sint8 index = *cindex;

/* Clamp index into valid range. */

if (index > 88) {

index = 88;

} else if (index < 0) {

index = 0;

}

/* explicit cast to avoid gcc warning about using 'char' as array index */

step = step_table[(size_t)index];

/* Update index value */

*cindex = index + index_table_4b[nybble];

/* This calculation uses shifts and additions because multiplications were

* much slower back then. Sadly, this can't just be replaced with an actual

* multiplication now as the old algorithm drops some bits. The closest

* approximation I could find is something like this:

* (nybble & 0x8 ? -1 : 1) * ((nybble & 0x7) * step / 4 + step / 8)

*/

delta = step >> 3;

if (nybble & 0x04) {

delta += step;

}

if (nybble & 0x02) {

delta += step >> 1;

}

if (nybble & 0x01) {

delta += step >> 2;

}

if (nybble & 0x08) {

delta = -delta;

}

sample = lastsample + delta;

/* Clamp output sample */

if (sample > max_audioval) {

sample = max_audioval;

} else if (sample < min_audioval) {

sample = min_audioval;

}

return (Sint16)sample;

}

static int IMA_ADPCM_DecodeBlockHeader(ADPCM_DecoderState *state)

{

Sint16 step;

Uint32 c;

Uint8 *cstate = (Uint8 *)state->cstate;

for (c = 0; c < state->channels; c++) {

size_t o = state->block.pos + c * 4;

/* Extract the sample from the header. */

Sint32 sample = state->block.data[o] | ((Sint32)state->block.data[o + 1] << 8);

if (sample >= 0x8000) {

sample -= 0x10000;

}

state->output.data[state->output.pos++] = (Sint16)sample;

/* Channel step index. */

step = (Sint16)state->block.data[o + 2];

cstate[c] = (Sint8)(step > 0x80 ? step - 0x100 : step);

/* Reserved byte in block header, should be 0. */

if (state->block.data[o + 3] != 0) {

/* Uh oh, corrupt data? Buggy code? */;

}

}

state->block.pos += state->blockheadersize;

/* Header provided one sample frame. */

state->framesleft--;

return 0;

}

/* Decodes the data of the IMA ADPCM block. Decoding will stop if a block is too

* short, returning with none or partially decoded data. The partial data always

* contains full sample frames (same sample count for each channel).

* Incomplete sample frames are discarded.

*/

static int IMA_ADPCM_DecodeBlockData(ADPCM_DecoderState *state)

{

size_t i;

int retval = 0;

const Uint32 channels = state->channels;

const size_t subblockframesize = (size_t)channels * 4;

Uint64 bytesrequired;

Uint32 c;

size_t blockpos = state->block.pos;

size_t blocksize = state->block.size;

size_t blockleft = blocksize - blockpos;

size_t outpos = state->output.pos;

Sint64 blockframesleft = state->samplesperblock - 1;

if (blockframesleft > state->framesleft) {

blockframesleft = state->framesleft;

}

bytesrequired = (blockframesleft + 7) / 8 * subblockframesize;

if (blockleft < bytesrequired) {

/* Data truncated. Calculate how many samples we can get out if it. */

const size_t guaranteedframes = blockleft / subblockframesize;

const size_t remainingbytes = blockleft % subblockframesize;

blockframesleft = guaranteedframes;

if (remainingbytes > subblockframesize - 4) {

blockframesleft += (remainingbytes % 4) * 2;

}

/* Signal the truncation. */

retval = -1;

}

/* Each channel has their nibbles packed into 32-bit blocks. These blocks

* are interleaved and make up the data part of the ADPCM block. This loop

* decodes the samples as they come from the input data and puts them at

* the appropriate places in the output data.

*/