Files
mozjpeg/jdhuff.c
DRC af618ffe09 Clean up the lossless JPEG feature
- Rename jpeg_simple_lossless() to jpeg_enable_lossless() and modify the
  function so that it stores the lossless parameters directly in the Ss
  and Al fields of jpeg_compress_struct rather than using a scan script.

- Move the cjpeg -lossless switch into "Switches for advanced users".

- Document the libjpeg API and run-time features that are unavailable in
  lossless mode, and ensure that all parameters, functions, and switches
  related to unavailable features are ignored or generate errors in
  lossless mode.

- Defer any action that depends on whether lossless mode is enabled
  until jpeg_start_compress()/jpeg_start_decompress() is called.

- Document the purpose of the point transform value.

- "Codec" stands for coder/decoder, so it is a bit awkward to say
  "lossless compression codec" and "lossless decompression codec".
  Use "lossless compressor" and "lossless decompressor" instead.

- Restore backward API/ABI compatibility with libjpeg v6b:

  * Move the new 'lossless' field from the exposed jpeg_compress_struct
    and jpeg_decompress_struct structures into the opaque
    jpeg_comp_master and jpeg_decomp_master structures, and allocate the
    master structures in the body of jpeg_create_compress() and
    jpeg_create_decompress().

  * Remove the new 'process' field from jpeg_compress_struct and
    jpeg_decompress_struct and replace it with the old
    'progressive_mode' field and the new 'lossless' field.

  * Remove the new 'data_unit' field from jpeg_compress_struct and
    jpeg_decompress_struct and replace it with a locally-computed
    data unit variable.

  * Restore the names of macros and fields that refer to DCT blocks, and
    document that they have a different meaning in lossless mode.  (Most
    of them aren't very meaningful in lossless mode anyhow.)

  * Remove the new alloc_darray() method from jpeg_memory_mgr and
    replace it with an internal macro that wraps the alloc_sarray()
    method.

  * Move the JDIFF* data types from jpeglib.h and jmorecfg.h into
    jpegint.h.

  * Remove the new 'codec' field from jpeg_compress_struct and
    jpeg_decompress_struct and instead reuse the existing internal
    coefficient control, forward/inverse DCT, and entropy
    encoding/decoding structures for lossless compression/decompression.

  * Repurpose existing error codes rather than introducing new ones.
    (The new JERR_BAD_RESTART and JWRN_MUST_DOWNSCALE codes remain,
    although JWRN_MUST_DOWNSCALE will probably be removed in
    libjpeg-turbo, since we have a different way of handling multiple
    data precisions.)

- Automatically enable lossless mode when a scan script with parameters
  that are only valid for lossless mode is detected, and document the
  use of scan scripts to generate lossless JPEG images.

- Move the sequential and shared Huffman routines back into jchuff.c and
  jdhuff.c, and document that those routines are shared with jclhuff.c
  and jdlhuff.c as well as with jcphuff.c and jdphuff.c.

- Move MAX_DIFF_BITS from jchuff.h into jclhuff.c, the only place where
  it is used.

- Move the predictor and scaler code into jclossls.c and jdlossls.c.

- Streamline register usage in the [un]differencers (inspired by similar
  optimizations in the color [de]converters.)

- Restructure the logic in a few places to reduce duplicated code.

- Ensure that all lossless-specific code is guarded by
  C_LOSSLESS_SUPPORTED or D_LOSSLESS_SUPPORTED and that the library can
  be built successfully if either or both of those macros is undefined.

- Remove all short forms of external names introduced by the lossless
  JPEG patch.  (These will not be needed by libjpeg-turbo, so there is
  no use cleaning them up.)

- Various wordsmithing, formatting, and punctuation tweaks

- Eliminate various compiler warnings.
2022-11-16 11:27:18 -06:00

837 lines
26 KiB
C

/*
* jdhuff.c
*
* This file was part of the Independent JPEG Group's software:
* Copyright (C) 1991-1997, Thomas G. Lane.
* Lossless JPEG Modifications:
* Copyright (C) 1999, Ken Murchison.
* libjpeg-turbo Modifications:
* Copyright (C) 2009-2011, 2016, 2018-2019, 2022, D. R. Commander.
* Copyright (C) 2018, Matthias Räncker.
* For conditions of distribution and use, see the accompanying README.ijg
* file.
*
* This file contains Huffman entropy decoding routines.
*
* Much of the complexity here has to do with supporting input suspension.
* If the data source module demands suspension, we want to be able to back
* up to the start of the current MCU. To do this, we copy state variables
* into local working storage, and update them back to the permanent
* storage only upon successful completion of an MCU.
*
* NOTE: All referenced figures are from
* Recommendation ITU-T T.81 (1992) | ISO/IEC 10918-1:1994.
*/
#define JPEG_INTERNALS
#include "jinclude.h"
#include "jpeglib.h"
#include "jdhuff.h" /* Declarations shared with jd*huff.c */
#include "jpegapicomp.h"
#include "jstdhuff.c"
/*
* Expanded entropy decoder object for Huffman decoding.
*
* The savable_state subrecord contains fields that change within an MCU,
* but must not be updated permanently until we complete the MCU.
*/
typedef struct {
int last_dc_val[MAX_COMPS_IN_SCAN]; /* last DC coef for each component */
} savable_state;
typedef struct {
struct jpeg_entropy_decoder pub; /* public fields */
/* These fields are loaded into local variables at start of each MCU.
* In case of suspension, we exit WITHOUT updating them.
*/
bitread_perm_state bitstate; /* Bit buffer at start of MCU */
savable_state saved; /* Other state at start of MCU */
/* These fields are NOT loaded into local working state. */
unsigned int restarts_to_go; /* MCUs left in this restart interval */
/* Pointers to derived tables (these workspaces have image lifespan) */
d_derived_tbl *dc_derived_tbls[NUM_HUFF_TBLS];
d_derived_tbl *ac_derived_tbls[NUM_HUFF_TBLS];
/* Precalculated info set up by start_pass for use in decode_mcu: */
/* Pointers to derived tables to be used for each block within an MCU */
d_derived_tbl *dc_cur_tbls[D_MAX_BLOCKS_IN_MCU];
d_derived_tbl *ac_cur_tbls[D_MAX_BLOCKS_IN_MCU];
/* Whether we care about the DC and AC coefficient values for each block */
boolean dc_needed[D_MAX_BLOCKS_IN_MCU];
boolean ac_needed[D_MAX_BLOCKS_IN_MCU];
} huff_entropy_decoder;
typedef huff_entropy_decoder *huff_entropy_ptr;
/*
* Initialize for a Huffman-compressed scan.
*/
METHODDEF(void)
start_pass_huff_decoder(j_decompress_ptr cinfo)
{
huff_entropy_ptr entropy = (huff_entropy_ptr)cinfo->entropy;
int ci, blkn, dctbl, actbl;
d_derived_tbl **pdtbl;
jpeg_component_info *compptr;
/* Check that the scan parameters Ss, Se, Ah/Al are OK for sequential JPEG.
* This ought to be an error condition, but we make it a warning because
* there are some baseline files out there with all zeroes in these bytes.
*/
if (cinfo->Ss != 0 || cinfo->Se != DCTSIZE2 - 1 ||
cinfo->Ah != 0 || cinfo->Al != 0)
WARNMS(cinfo, JWRN_NOT_SEQUENTIAL);
for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
compptr = cinfo->cur_comp_info[ci];
dctbl = compptr->dc_tbl_no;
actbl = compptr->ac_tbl_no;
/* Compute derived values for Huffman tables */
/* We may do this more than once for a table, but it's not expensive */
pdtbl = (d_derived_tbl **)(entropy->dc_derived_tbls) + dctbl;
jpeg_make_d_derived_tbl(cinfo, TRUE, dctbl, pdtbl);
pdtbl = (d_derived_tbl **)(entropy->ac_derived_tbls) + actbl;
jpeg_make_d_derived_tbl(cinfo, FALSE, actbl, pdtbl);
/* Initialize DC predictions to 0 */
entropy->saved.last_dc_val[ci] = 0;
}
/* Precalculate decoding info for each block in an MCU of this scan */
for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
ci = cinfo->MCU_membership[blkn];
compptr = cinfo->cur_comp_info[ci];
/* Precalculate which table to use for each block */
entropy->dc_cur_tbls[blkn] = entropy->dc_derived_tbls[compptr->dc_tbl_no];
entropy->ac_cur_tbls[blkn] = entropy->ac_derived_tbls[compptr->ac_tbl_no];
/* Decide whether we really care about the coefficient values */
if (compptr->component_needed) {
entropy->dc_needed[blkn] = TRUE;
/* we don't need the ACs if producing a 1/8th-size image */
entropy->ac_needed[blkn] = (compptr->_DCT_scaled_size > 1);
} else {
entropy->dc_needed[blkn] = entropy->ac_needed[blkn] = FALSE;
}
}
/* Initialize bitread state variables */
entropy->bitstate.bits_left = 0;
entropy->bitstate.get_buffer = 0; /* unnecessary, but keeps Purify quiet */
entropy->pub.insufficient_data = FALSE;
/* Initialize restart counter */
entropy->restarts_to_go = cinfo->restart_interval;
}
/*
* Compute the derived values for a Huffman table.
* This routine also performs some validation checks on the table.
*
* Note this is also used by jdphuff.c and jdlhuff.c.
*/
GLOBAL(void)
jpeg_make_d_derived_tbl(j_decompress_ptr cinfo, boolean isDC, int tblno,
d_derived_tbl **pdtbl)
{
JHUFF_TBL *htbl;
d_derived_tbl *dtbl;
int p, i, l, si, numsymbols;
int lookbits, ctr;
char huffsize[257];
unsigned int huffcode[257];
unsigned int code;
/* Note that huffsize[] and huffcode[] are filled in code-length order,
* paralleling the order of the symbols themselves in htbl->huffval[].
*/
/* Find the input Huffman table */
if (tblno < 0 || tblno >= NUM_HUFF_TBLS)
ERREXIT1(cinfo, JERR_NO_HUFF_TABLE, tblno);
htbl =
isDC ? cinfo->dc_huff_tbl_ptrs[tblno] : cinfo->ac_huff_tbl_ptrs[tblno];
if (htbl == NULL)
ERREXIT1(cinfo, JERR_NO_HUFF_TABLE, tblno);
/* Allocate a workspace if we haven't already done so. */
if (*pdtbl == NULL)
*pdtbl = (d_derived_tbl *)
(*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
sizeof(d_derived_tbl));
dtbl = *pdtbl;
dtbl->pub = htbl; /* fill in back link */
/* Figure C.1: make table of Huffman code length for each symbol */
p = 0;
for (l = 1; l <= 16; l++) {
i = (int)htbl->bits[l];
if (i < 0 || p + i > 256) /* protect against table overrun */
ERREXIT(cinfo, JERR_BAD_HUFF_TABLE);
while (i--)
huffsize[p++] = (char)l;
}
huffsize[p] = 0;
numsymbols = p;
/* Figure C.2: generate the codes themselves */
/* We also validate that the counts represent a legal Huffman code tree. */
code = 0;
si = huffsize[0];
p = 0;
while (huffsize[p]) {
while (((int)huffsize[p]) == si) {
huffcode[p++] = code;
code++;
}
/* code is now 1 more than the last code used for codelength si; but
* it must still fit in si bits, since no code is allowed to be all ones.
*/
if (((JLONG)code) >= (((JLONG)1) << si))
ERREXIT(cinfo, JERR_BAD_HUFF_TABLE);
code <<= 1;
si++;
}
/* Figure F.15: generate decoding tables for bit-sequential decoding */
p = 0;
for (l = 1; l <= 16; l++) {
if (htbl->bits[l]) {
/* valoffset[l] = huffval[] index of 1st symbol of code length l,
* minus the minimum code of length l
*/
dtbl->valoffset[l] = (JLONG)p - (JLONG)huffcode[p];
p += htbl->bits[l];
dtbl->maxcode[l] = huffcode[p - 1]; /* maximum code of length l */
} else {
dtbl->maxcode[l] = -1; /* -1 if no codes of this length */
}
}
dtbl->valoffset[17] = 0;
dtbl->maxcode[17] = 0xFFFFFL; /* ensures jpeg_huff_decode terminates */
/* Compute lookahead tables to speed up decoding.
* First we set all the table entries to 0, indicating "too long";
* then we iterate through the Huffman codes that are short enough and
* fill in all the entries that correspond to bit sequences starting
* with that code.
*/
for (i = 0; i < (1 << HUFF_LOOKAHEAD); i++)
dtbl->lookup[i] = (HUFF_LOOKAHEAD + 1) << HUFF_LOOKAHEAD;
p = 0;
for (l = 1; l <= HUFF_LOOKAHEAD; l++) {
for (i = 1; i <= (int)htbl->bits[l]; i++, p++) {
/* l = current code's length, p = its index in huffcode[] & huffval[]. */
/* Generate left-justified code followed by all possible bit sequences */
lookbits = huffcode[p] << (HUFF_LOOKAHEAD - l);
for (ctr = 1 << (HUFF_LOOKAHEAD - l); ctr > 0; ctr--) {
dtbl->lookup[lookbits] = (l << HUFF_LOOKAHEAD) | htbl->huffval[p];
lookbits++;
}
}
}
/* Validate symbols as being reasonable.
* For AC tables, we make no check, but accept all byte values 0..255.
* For DC tables, we require the symbols to be in range 0..15 in lossy mode
* and 0..16 in lossless mode. (Tighter bounds could be applied depending on
* the data depth and mode, but this is sufficient to ensure safe decoding.)
*/
if (isDC) {
for (i = 0; i < numsymbols; i++) {
int sym = htbl->huffval[i];
if (sym < 0 || sym > (cinfo->master->lossless ? 16 : 15))
ERREXIT(cinfo, JERR_BAD_HUFF_TABLE);
}
}
}
/*
* Out-of-line code for bit fetching (shared with jdphuff.c and jdlhuff.c).
* See jdhuff.h for info about usage.
* Note: current values of get_buffer and bits_left are passed as parameters,
* but are returned in the corresponding fields of the state struct.
*
* On most machines MIN_GET_BITS should be 25 to allow the full 32-bit width
* of get_buffer to be used. (On machines with wider words, an even larger
* buffer could be used.) However, on some machines 32-bit shifts are
* quite slow and take time proportional to the number of places shifted.
* (This is true with most PC compilers, for instance.) In this case it may
* be a win to set MIN_GET_BITS to the minimum value of 15. This reduces the
* average shift distance at the cost of more calls to jpeg_fill_bit_buffer.
*/
#ifdef SLOW_SHIFT_32
#define MIN_GET_BITS 15 /* minimum allowable value */
#else
#define MIN_GET_BITS (BIT_BUF_SIZE - 7)
#endif
GLOBAL(boolean)
jpeg_fill_bit_buffer(bitread_working_state *state,
register bit_buf_type get_buffer, register int bits_left,
int nbits)
/* Load up the bit buffer to a depth of at least nbits */
{
/* Copy heavily used state fields into locals (hopefully registers) */
register const JOCTET *next_input_byte = state->next_input_byte;
register size_t bytes_in_buffer = state->bytes_in_buffer;
j_decompress_ptr cinfo = state->cinfo;
/* Attempt to load at least MIN_GET_BITS bits into get_buffer. */
/* (It is assumed that no request will be for more than that many bits.) */
/* We fail to do so only if we hit a marker or are forced to suspend. */
if (cinfo->unread_marker == 0) { /* cannot advance past a marker */
while (bits_left < MIN_GET_BITS) {
register int c;
/* Attempt to read a byte */
if (bytes_in_buffer == 0) {
if (!(*cinfo->src->fill_input_buffer) (cinfo))
return FALSE;
next_input_byte = cinfo->src->next_input_byte;
bytes_in_buffer = cinfo->src->bytes_in_buffer;
}
bytes_in_buffer--;
c = *next_input_byte++;
/* If it's 0xFF, check and discard stuffed zero byte */
if (c == 0xFF) {
/* Loop here to discard any padding FF's on terminating marker,
* so that we can save a valid unread_marker value. NOTE: we will
* accept multiple FF's followed by a 0 as meaning a single FF data
* byte. This data pattern is not valid according to the standard.
*/
do {
if (bytes_in_buffer == 0) {
if (!(*cinfo->src->fill_input_buffer) (cinfo))
return FALSE;
next_input_byte = cinfo->src->next_input_byte;
bytes_in_buffer = cinfo->src->bytes_in_buffer;
}
bytes_in_buffer--;
c = *next_input_byte++;
} while (c == 0xFF);
if (c == 0) {
/* Found FF/00, which represents an FF data byte */
c = 0xFF;
} else {
/* Oops, it's actually a marker indicating end of compressed data.
* Save the marker code for later use.
* Fine point: it might appear that we should save the marker into
* bitread working state, not straight into permanent state. But
* once we have hit a marker, we cannot need to suspend within the
* current MCU, because we will read no more bytes from the data
* source. So it is OK to update permanent state right away.
*/
cinfo->unread_marker = c;
/* See if we need to insert some fake zero bits. */
goto no_more_bytes;
}
}
/* OK, load c into get_buffer */
get_buffer = (get_buffer << 8) | c;
bits_left += 8;
} /* end while */
} else {
no_more_bytes:
/* We get here if we've read the marker that terminates the compressed
* data segment. There should be enough bits in the buffer register
* to satisfy the request; if so, no problem.
*/
if (nbits > bits_left) {
/* Uh-oh. Report corrupted data to user and stuff zeroes into
* the data stream, so that we can produce some kind of image.
* We use a nonvolatile flag to ensure that only one warning message
* appears per data segment.
*/
if (!cinfo->entropy->insufficient_data) {
WARNMS(cinfo, JWRN_HIT_MARKER);
cinfo->entropy->insufficient_data = TRUE;
}
/* Fill the buffer with zero bits */
get_buffer <<= MIN_GET_BITS - bits_left;
bits_left = MIN_GET_BITS;
}
}
/* Unload the local registers */
state->next_input_byte = next_input_byte;
state->bytes_in_buffer = bytes_in_buffer;
state->get_buffer = get_buffer;
state->bits_left = bits_left;
return TRUE;
}
/* Macro version of the above, which performs much better but does not
handle markers. We have to hand off any blocks with markers to the
slower routines. */
#define GET_BYTE { \
register int c0, c1; \
c0 = *buffer++; \
c1 = *buffer; \
/* Pre-execute most common case */ \
get_buffer = (get_buffer << 8) | c0; \
bits_left += 8; \
if (c0 == 0xFF) { \
/* Pre-execute case of FF/00, which represents an FF data byte */ \
buffer++; \
if (c1 != 0) { \
/* Oops, it's actually a marker indicating end of compressed data. */ \
cinfo->unread_marker = c1; \
/* Back out pre-execution and fill the buffer with zero bits */ \
buffer -= 2; \
get_buffer &= ~0xFF; \
} \
} \
}
#if SIZEOF_SIZE_T == 8 || defined(_WIN64) || (defined(__x86_64__) && defined(__ILP32__))
/* Pre-fetch 48 bytes, because the holding register is 64-bit */
#define FILL_BIT_BUFFER_FAST \
if (bits_left <= 16) { \
GET_BYTE GET_BYTE GET_BYTE GET_BYTE GET_BYTE GET_BYTE \
}
#else
/* Pre-fetch 16 bytes, because the holding register is 32-bit */
#define FILL_BIT_BUFFER_FAST \
if (bits_left <= 16) { \
GET_BYTE GET_BYTE \
}
#endif
/*
* Out-of-line code for Huffman code decoding.
* See jdhuff.h for info about usage.
*/
GLOBAL(int)
jpeg_huff_decode(bitread_working_state *state,
register bit_buf_type get_buffer, register int bits_left,
d_derived_tbl *htbl, int min_bits)
{
register int l = min_bits;
register JLONG code;
/* HUFF_DECODE has determined that the code is at least min_bits */
/* bits long, so fetch that many bits in one swoop. */
CHECK_BIT_BUFFER(*state, l, return -1);
code = GET_BITS(l);
/* Collect the rest of the Huffman code one bit at a time. */
/* This is per Figure F.16. */
while (code > htbl->maxcode[l]) {
code <<= 1;
CHECK_BIT_BUFFER(*state, 1, return -1);
code |= GET_BITS(1);
l++;
}
/* Unload the local registers */
state->get_buffer = get_buffer;
state->bits_left = bits_left;
/* With garbage input we may reach the sentinel value l = 17. */
if (l > 16) {
WARNMS(state->cinfo, JWRN_HUFF_BAD_CODE);
return 0; /* fake a zero as the safest result */
}
return htbl->pub->huffval[(int)(code + htbl->valoffset[l])];
}
/*
* Figure F.12: extend sign bit.
* On some machines, a shift and add will be faster than a table lookup.
*/
#define AVOID_TABLES
#ifdef AVOID_TABLES
#define NEG_1 ((unsigned int)-1)
#define HUFF_EXTEND(x, s) \
((x) + ((((x) - (1 << ((s) - 1))) >> 31) & (((NEG_1) << (s)) + 1)))
#else
#define HUFF_EXTEND(x, s) \
((x) < extend_test[s] ? (x) + extend_offset[s] : (x))
static const int extend_test[16] = { /* entry n is 2**(n-1) */
0, 0x0001, 0x0002, 0x0004, 0x0008, 0x0010, 0x0020, 0x0040, 0x0080,
0x0100, 0x0200, 0x0400, 0x0800, 0x1000, 0x2000, 0x4000
};
static const int extend_offset[16] = { /* entry n is (-1 << n) + 1 */
0, ((-1) << 1) + 1, ((-1) << 2) + 1, ((-1) << 3) + 1, ((-1) << 4) + 1,
((-1) << 5) + 1, ((-1) << 6) + 1, ((-1) << 7) + 1, ((-1) << 8) + 1,
((-1) << 9) + 1, ((-1) << 10) + 1, ((-1) << 11) + 1, ((-1) << 12) + 1,
((-1) << 13) + 1, ((-1) << 14) + 1, ((-1) << 15) + 1
};
#endif /* AVOID_TABLES */
/*
* Check for a restart marker & resynchronize decoder.
* Returns FALSE if must suspend.
*/
LOCAL(boolean)
process_restart(j_decompress_ptr cinfo)
{
huff_entropy_ptr entropy = (huff_entropy_ptr)cinfo->entropy;
int ci;
/* Throw away any unused bits remaining in bit buffer; */
/* include any full bytes in next_marker's count of discarded bytes */
cinfo->marker->discarded_bytes += entropy->bitstate.bits_left / 8;
entropy->bitstate.bits_left = 0;
/* Advance past the RSTn marker */
if (!(*cinfo->marker->read_restart_marker) (cinfo))
return FALSE;
/* Re-initialize DC predictions to 0 */
for (ci = 0; ci < cinfo->comps_in_scan; ci++)
entropy->saved.last_dc_val[ci] = 0;
/* Reset restart counter */
entropy->restarts_to_go = cinfo->restart_interval;
/* Reset out-of-data flag, unless read_restart_marker left us smack up
* against a marker. In that case we will end up treating the next data
* segment as empty, and we can avoid producing bogus output pixels by
* leaving the flag set.
*/
if (cinfo->unread_marker == 0)
entropy->pub.insufficient_data = FALSE;
return TRUE;
}
#if defined(__has_feature)
#if __has_feature(undefined_behavior_sanitizer)
__attribute__((no_sanitize("signed-integer-overflow"),
no_sanitize("unsigned-integer-overflow")))
#endif
#endif
LOCAL(boolean)
decode_mcu_slow(j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
{
huff_entropy_ptr entropy = (huff_entropy_ptr)cinfo->entropy;
BITREAD_STATE_VARS;
int blkn;
savable_state state;
/* Outer loop handles each block in the MCU */
/* Load up working state */
BITREAD_LOAD_STATE(cinfo, entropy->bitstate);
state = entropy->saved;
for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
JBLOCKROW block = MCU_data ? MCU_data[blkn] : NULL;
d_derived_tbl *dctbl = entropy->dc_cur_tbls[blkn];
d_derived_tbl *actbl = entropy->ac_cur_tbls[blkn];
register int s, k, r;
/* Decode a single block's worth of coefficients */
/* Section F.2.2.1: decode the DC coefficient difference */
HUFF_DECODE(s, br_state, dctbl, return FALSE, label1);
if (s) {
CHECK_BIT_BUFFER(br_state, s, return FALSE);
r = GET_BITS(s);
s = HUFF_EXTEND(r, s);
}
if (entropy->dc_needed[blkn]) {
/* Convert DC difference to actual value, update last_dc_val */
int ci = cinfo->MCU_membership[blkn];
/* Certain malformed JPEG images produce repeated DC coefficient
* differences of 2047 or -2047, which causes state.last_dc_val[ci] to
* grow until it overflows or underflows a 32-bit signed integer. This
* behavior is, to the best of our understanding, innocuous, and it is
* unclear how to work around it without potentially affecting
* performance. Thus, we (hopefully temporarily) suppress UBSan integer
* overflow errors for this function and decode_mcu_fast().
*/
s += state.last_dc_val[ci];
state.last_dc_val[ci] = s;
if (block) {
/* Output the DC coefficient (assumes jpeg_natural_order[0] = 0) */
(*block)[0] = (JCOEF)s;
}
}
if (entropy->ac_needed[blkn] && block) {
/* Section F.2.2.2: decode the AC coefficients */
/* Since zeroes are skipped, output area must be cleared beforehand */
for (k = 1; k < DCTSIZE2; k++) {
HUFF_DECODE(s, br_state, actbl, return FALSE, label2);
r = s >> 4;
s &= 15;
if (s) {
k += r;
CHECK_BIT_BUFFER(br_state, s, return FALSE);
r = GET_BITS(s);
s = HUFF_EXTEND(r, s);
/* Output coefficient in natural (dezigzagged) order.
* Note: the extra entries in jpeg_natural_order[] will save us
* if k >= DCTSIZE2, which could happen if the data is corrupted.
*/
(*block)[jpeg_natural_order[k]] = (JCOEF)s;
} else {
if (r != 15)
break;
k += 15;
}
}
} else {
/* Section F.2.2.2: decode the AC coefficients */
/* In this path we just discard the values */
for (k = 1; k < DCTSIZE2; k++) {
HUFF_DECODE(s, br_state, actbl, return FALSE, label3);
r = s >> 4;
s &= 15;
if (s) {
k += r;
CHECK_BIT_BUFFER(br_state, s, return FALSE);
DROP_BITS(s);
} else {
if (r != 15)
break;
k += 15;
}
}
}
}
/* Completed MCU, so update state */
BITREAD_SAVE_STATE(cinfo, entropy->bitstate);
entropy->saved = state;
return TRUE;
}
#if defined(__has_feature)
#if __has_feature(undefined_behavior_sanitizer)
__attribute__((no_sanitize("signed-integer-overflow"),
no_sanitize("unsigned-integer-overflow")))
#endif
#endif
LOCAL(boolean)
decode_mcu_fast(j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
{
huff_entropy_ptr entropy = (huff_entropy_ptr)cinfo->entropy;
BITREAD_STATE_VARS;
JOCTET *buffer;
int blkn;
savable_state state;
/* Outer loop handles each block in the MCU */
/* Load up working state */
BITREAD_LOAD_STATE(cinfo, entropy->bitstate);
buffer = (JOCTET *)br_state.next_input_byte;
state = entropy->saved;
for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
JBLOCKROW block = MCU_data ? MCU_data[blkn] : NULL;
d_derived_tbl *dctbl = entropy->dc_cur_tbls[blkn];
d_derived_tbl *actbl = entropy->ac_cur_tbls[blkn];
register int s, k, r, l;
HUFF_DECODE_FAST(s, l, dctbl);
if (s) {
FILL_BIT_BUFFER_FAST
r = GET_BITS(s);
s = HUFF_EXTEND(r, s);
}
if (entropy->dc_needed[blkn]) {
int ci = cinfo->MCU_membership[blkn];
/* Refer to the comment in decode_mcu_slow() regarding the supression of
* a UBSan integer overflow error in this line of code.
*/
s += state.last_dc_val[ci];
state.last_dc_val[ci] = s;
if (block)
(*block)[0] = (JCOEF)s;
}
if (entropy->ac_needed[blkn] && block) {
for (k = 1; k < DCTSIZE2; k++) {
HUFF_DECODE_FAST(s, l, actbl);
r = s >> 4;
s &= 15;
if (s) {
k += r;
FILL_BIT_BUFFER_FAST
r = GET_BITS(s);
s = HUFF_EXTEND(r, s);
(*block)[jpeg_natural_order[k]] = (JCOEF)s;
} else {
if (r != 15) break;
k += 15;
}
}
} else {
for (k = 1; k < DCTSIZE2; k++) {
HUFF_DECODE_FAST(s, l, actbl);
r = s >> 4;
s &= 15;
if (s) {
k += r;
FILL_BIT_BUFFER_FAST
DROP_BITS(s);
} else {
if (r != 15) break;
k += 15;
}
}
}
}
if (cinfo->unread_marker != 0) {
cinfo->unread_marker = 0;
return FALSE;
}
br_state.bytes_in_buffer -= (buffer - br_state.next_input_byte);
br_state.next_input_byte = buffer;
BITREAD_SAVE_STATE(cinfo, entropy->bitstate);
entropy->saved = state;
return TRUE;
}
/*
* Decode and return one MCU's worth of Huffman-compressed coefficients.
* The coefficients are reordered from zigzag order into natural array order,
* but are not dequantized.
*
* The i'th block of the MCU is stored into the block pointed to by
* MCU_data[i]. WE ASSUME THIS AREA HAS BEEN ZEROED BY THE CALLER.
* (Wholesale zeroing is usually a little faster than retail...)
*
* Returns FALSE if data source requested suspension. In that case no
* changes have been made to permanent state. (Exception: some output
* coefficients may already have been assigned. This is harmless for
* this module, since we'll just re-assign them on the next call.)
*/
#define BUFSIZE (DCTSIZE2 * 8)
METHODDEF(boolean)
decode_mcu(j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
{
huff_entropy_ptr entropy = (huff_entropy_ptr)cinfo->entropy;
int usefast = 1;
/* Process restart marker if needed; may have to suspend */
if (cinfo->restart_interval) {
if (entropy->restarts_to_go == 0)
if (!process_restart(cinfo))
return FALSE;
usefast = 0;
}
if (cinfo->src->bytes_in_buffer < BUFSIZE * (size_t)cinfo->blocks_in_MCU ||
cinfo->unread_marker != 0)
usefast = 0;
/* If we've run out of data, just leave the MCU set to zeroes.
* This way, we return uniform gray for the remainder of the segment.
*/
if (!entropy->pub.insufficient_data) {
if (usefast) {
if (!decode_mcu_fast(cinfo, MCU_data)) goto use_slow;
} else {
use_slow:
if (!decode_mcu_slow(cinfo, MCU_data)) return FALSE;
}
}
/* Account for restart interval (no-op if not using restarts) */
if (cinfo->restart_interval)
entropy->restarts_to_go--;
return TRUE;
}
/*
* Module initialization routine for Huffman entropy decoding.
*/
GLOBAL(void)
jinit_huff_decoder(j_decompress_ptr cinfo)
{
huff_entropy_ptr entropy;
int i;
/* Motion JPEG frames typically do not include the Huffman tables if they
are the default tables. Thus, if the tables are not set by the time
the Huffman decoder is initialized (usually within the body of
jpeg_start_decompress()), we set them to default values. */
std_huff_tables((j_common_ptr)cinfo);
entropy = (huff_entropy_ptr)
(*cinfo->mem->alloc_small) ((j_common_ptr)cinfo, JPOOL_IMAGE,
sizeof(huff_entropy_decoder));
cinfo->entropy = (struct jpeg_entropy_decoder *)entropy;
entropy->pub.start_pass = start_pass_huff_decoder;
entropy->pub.decode_mcu = decode_mcu;
/* Mark tables unallocated */
for (i = 0; i < NUM_HUFF_TBLS; i++) {
entropy->dc_derived_tbls[i] = entropy->ac_derived_tbls[i] = NULL;
}
}