pollymc/depends/pack200/src/coding.cpp

/*
 * Copyright (c) 2002, 2009, Oracle and/or its affiliates. All rights reserved.
 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
 *
 * This code is free software; you can redistribute it and/or modify it
 * under the terms of the GNU General Public License version 2 only, as
 * published by the Free Software Foundation.  Oracle designates this
 * particular file as subject to the "Classpath" exception as provided
 * by Oracle in the LICENSE file that accompanied this code.
 *
 * This code is distributed in the hope that it will be useful, but WITHOUT
 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
 * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
 * version 2 for more details (a copy is included in the LICENSE file that
 * accompanied this code).
 *
 * You should have received a copy of the GNU General Public License version
 * 2 along with this work; if not, write to the Free Software Foundation,
 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
 *
 * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
 * or visit www.oracle.com if you need additional information or have any
 * questions.
 */

// -*- C++ -*-
// Small program for unpacking specially compressed Java packages.
// John R. Rose

#include <stdio.h>
#include <string.h>
#include <stdlib.h>
#include <stdarg.h>
#include <assert.h>
#include <stdint.h>

#include "defines.h"
#include "bytes.h"
#include "utils.h"
#include "coding.h"

#include "constants.h"
#include "unpack.h"

extern coding basic_codings[];

// CODING_PRIVATE causes a lot of them
#pragma GCC diagnostic ignored "-Wunused-variable"

#define CODING_PRIVATE(spec)                                                                   \
	int spec_ = spec;                                                                          \
	int B = CODING_B(spec_);                                                                   \
	int H = CODING_H(spec_);                                                                   \
	int L = 256 - H;                                                                           \
	int S = CODING_S(spec_);                                                                   \
	int D = CODING_D(spec_)

#define IS_NEG_CODE(S, codeVal) ((((int)(codeVal) + 1) & ((1 << S) - 1)) == 0)

#define DECODE_SIGN_S1(ux) (((uint32_t)(ux) >> 1) ^ -((int)(ux) & 1))

static int decode_sign(int S, uint32_t ux)
{ // == Coding.decodeSign32
	assert(S > 0);
	uint32_t sigbits = (ux >> S);
	if (IS_NEG_CODE(S, ux))
		return (int)(~sigbits);
	else
		return (int)(ux - sigbits);
	// Note that (int)(ux-sigbits) can be negative, if ux is large enough.
}

coding *coding::init()
{
	if (umax > 0)
		return this;   // already done
	assert(spec != 0); // sanity

	// fill in derived fields
	CODING_PRIVATE(spec);

	// Return nullptr if 'arb(BHSD)' parameter constraints are not met:
	if (B < 1 || B > B_MAX)
		return nullptr;
	if (H < 1 || H > 256)
		return nullptr;
	if (S < 0 || S > 2)
		return nullptr;
	if (D < 0 || D > 1)
		return nullptr;
	if (B == 1 && H != 256)
		return nullptr; // 1-byte coding must be fixed-size
	if (B >= 5 && H == 256)
		return nullptr; // no 5-byte fixed-size coding

	// first compute the range of the coding, in 64 bits
	int64_t range = 0;
	{
		int64_t H_i = 1;
		for (int i = 0; i < B; i++)
		{
			range += H_i;
			H_i *= H;
		}
		range *= L;
		range += H_i;
	}
	assert(range > 0); // no useless codings, please

	int this_umax;

	// now, compute min and max
	if (range >= ((int64_t)1 << 32))
	{
		this_umax = INT_MAX_VALUE;
		this->umin = INT_MIN_VALUE;
		this->max = INT_MAX_VALUE;
		this->min = INT_MIN_VALUE;
	}
	else
	{
		this_umax = (range > INT_MAX_VALUE) ? INT_MAX_VALUE : (int)range - 1;
		this->max = this_umax;
		this->min = this->umin = 0;
		if (S != 0 && range != 0)
		{
			int64_t maxPosCode = range - 1;
			int64_t maxNegCode = range - 1;
			while (IS_NEG_CODE(S, maxPosCode))
				--maxPosCode;
			while (!IS_NEG_CODE(S, maxNegCode))
				--maxNegCode;
			int maxPos = decode_sign(S, (uint32_t)maxPosCode);
			if (maxPos < 0)
				this->max = INT_MAX_VALUE; // 32-bit wraparound
			else
				this->max = maxPos;
			if (maxNegCode < 0)
				this->min = 0; // No negative codings at all.
			else
				this->min = decode_sign(S, (uint32_t)maxNegCode);
		}
	}

	assert(!(isFullRange | isSigned | isSubrange)); // init
	if (min < 0)
		this->isSigned = true;
	if (max < INT_MAX_VALUE && range <= INT_MAX_VALUE)
		this->isSubrange = true;
	if (max == INT_MAX_VALUE && min == INT_MIN_VALUE)
		this->isFullRange = true;

	// do this last, to reduce MT exposure (should have a membar too)
	this->umax = this_umax;

	return this;
}

coding *coding::findBySpec(int spec)
{
	for (coding *scan = &basic_codings[0];; scan++)
	{
		if (scan->spec == spec)
			return scan->init();
		if (scan->spec == 0)
			break;
	}
	coding *ptr = NEW(coding, 1);
	if (!ptr)
		return nullptr;
	coding *c = ptr->initFrom(spec);
	if (c == nullptr)
	{
		::free(ptr);
	}
	else
		// else caller should free it...
		c->isMalloc = true;
	return c;
}

coding *coding::findBySpec(int B, int H, int S, int D)
{
	if (B < 1 || B > B_MAX)
		return nullptr;
	if (H < 1 || H > 256)
		return nullptr;
	if (S < 0 || S > 2)
		return nullptr;
	if (D < 0 || D > 1)
		return nullptr;
	return findBySpec(CODING_SPEC(B, H, S, D));
}

void coding::free()
{
	if (isMalloc)
	{
		::free(this);
	}
}

void coding_method::reset(value_stream *state)
{
	assert(state->rp == state->rplimit); // not in mid-stream, please
	// assert(this == vs0.cm);
	state[0] = vs0;
	if (uValues != nullptr)
	{
		uValues->reset(state->helper());
	}
}

uint32_t coding::parse(byte *&rp, int B, int H)
{
	int L = 256 - H;
	byte *ptr = rp;
	// hand peel the i==0 part of the loop:
	uint32_t b_i = *ptr++ & 0xFF;
	if (B == 1 || b_i < (uint32_t)L)
	{
		rp = ptr;
		return b_i;
	}
	uint32_t sum = b_i;
	uint32_t H_i = H;
	assert(B <= B_MAX);
	for (int i = 2; i <= B_MAX; i++)
	{ // easy for compilers to unroll if desired
		b_i = *ptr++ & 0xFF;
		sum += b_i * H_i;
		if (i == B || b_i < (uint32_t)L)
		{
			rp = ptr;
			return sum;
		}
		H_i *= H;
	}
	assert(false);
	return 0;
}

uint32_t coding::parse_lgH(byte *&rp, int B, int H, int lgH)
{
	assert(H == (1 << lgH));
	int L = 256 - (1 << lgH);
	byte *ptr = rp;
	// hand peel the i==0 part of the loop:
	uint32_t b_i = *ptr++ & 0xFF;
	if (B == 1 || b_i < (uint32_t)L)
	{
		rp = ptr;
		return b_i;
	}
	uint32_t sum = b_i;
	uint32_t lg_H_i = lgH;
	assert(B <= B_MAX);
	for (int i = 2; i <= B_MAX; i++)
	{ // easy for compilers to unroll if desired
		b_i = *ptr++ & 0xFF;
		sum += b_i << lg_H_i;
		if (i == B || b_i < (uint32_t)L)
		{
			rp = ptr;
			return sum;
		}
		lg_H_i += lgH;
	}
	assert(false);
	return 0;
}

static const char ERB[] = "EOF reading band";

void coding::parseMultiple(byte *&rp, int N, byte *limit, int B, int H)
{
	if (N < 0)
	{
		unpack_abort("bad value count");
		return;
	}
	byte *ptr = rp;
	if (B == 1 || H == 256)
	{
		size_t len = (size_t)N * B;
		if (len / B != (size_t)N || ptr + len > limit)
		{
			unpack_abort(ERB);
			return;
		}
		rp = ptr + len;
		return;
	}
	// Note:  We assume rp has enough zero-padding.
	int L = 256 - H;
	int n = B;
	while (N > 0)
	{
		ptr += 1;
		if (--n == 0)
		{
			// end of encoding at B bytes, regardless of byte value
		}
		else
		{
			int b = (ptr[-1] & 0xFF);
			if (b >= L)
			{
				// keep going, unless we find a byte < L
				continue;
			}
		}
		// found the last byte
		N -= 1;
		n = B; // reset length counter
		// do an error check here
		if (ptr > limit)
		{
			unpack_abort(ERB);
			return;
		}
	}
	rp = ptr;
	return;
}

bool value_stream::hasHelper()
{
	// If my coding method is a pop-style method,
	// then I need a second value stream to transmit
	// unfavored values.
	// This can be determined by examining fValues.
	return cm->fValues != nullptr;
}

void value_stream::init(byte *rp_, byte *rplimit_, coding *defc)
{
	rp = rp_;
	rplimit = rplimit_;
	sum = 0;
	cm = nullptr; // no need in the simple case
	setCoding(defc);
}

void value_stream::setCoding(coding *defc)
{
	if (defc == nullptr)
	{
		unpack_abort("bad coding");
		defc = coding::findByIndex(_meta_canon_min); // random pick for recovery
	}

	c = (*defc);

	// choose cmk
	cmk = cmk_ERROR;
	switch (c.spec)
	{
	case BYTE1_spec:
		cmk = cmk_BYTE1;
		break;
	case CHAR3_spec:
		cmk = cmk_CHAR3;
		break;
	case UNSIGNED5_spec:
		cmk = cmk_UNSIGNED5;
		break;
	case DELTA5_spec:
		cmk = cmk_DELTA5;
		break;
	case BCI5_spec:
		cmk = cmk_BCI5;
		break;
	case BRANCH5_spec:
		cmk = cmk_BRANCH5;
		break;
	default:
		if (c.D() == 0)
		{
			switch (c.S())
			{
			case 0:
				cmk = cmk_BHS0;
				break;
			case 1:
				cmk = cmk_BHS1;
				break;
			default:
				cmk = cmk_BHS;
				break;
			}
		}
		else
		{
			if (c.S() == 1)
			{
				if (c.isFullRange)
					cmk = cmk_BHS1D1full;
				if (c.isSubrange)
					cmk = cmk_BHS1D1sub;
			}
			if (cmk == cmk_ERROR)
				cmk = cmk_BHSD1;
		}
	}
}

static int getPopValue(value_stream *self, uint32_t uval)
{
	if (uval > 0)
	{
		// note that the initial parse performed a range check
		assert(uval <= (uint32_t)self->cm->fVlength);
		return self->cm->fValues[uval - 1];
	}
	else
	{
		// take an unfavored value
		return self->helper()->getInt();
	}
}

int coding::sumInUnsignedRange(int x, int y)
{
	assert(isSubrange);
	int range = (int)(umax + 1);
	assert(range > 0);
	x += y;
	if (x != (int)((int64_t)(x - y) + (int64_t)y))
	{
		// 32-bit overflow interferes with range reduction.
		// Back off from the overflow by adding a multiple of range:
		if (x < 0)
		{
			x -= range;
			assert(x >= 0);
		}
		else
		{
			x += range;
			assert(x < 0);
		}
	}
	if (x < 0)
	{
		x += range;
		if (x >= 0)
			return x;
	}
	else if (x >= range)
	{
		x -= range;
		if (x < range)
			return x;
	}
	else
	{
		// in range
		return x;
	}
	// do it the hard way
	x %= range;
	if (x < 0)
		x += range;
	return x;
}

static int getDeltaValue(value_stream *self, uint32_t uval, bool isSubrange)
{
	assert((uint32_t)(self->c.isSubrange) == (uint32_t)isSubrange);
	assert(self->c.isSubrange | self->c.isFullRange);
	if (isSubrange)
		return self->sum = self->c.sumInUnsignedRange(self->sum, (int)uval);
	else
		return self->sum += (int)uval;
}

bool value_stream::hasValue()
{
	if (rp < rplimit)
		return true;
	if (cm == nullptr)
		return false;
	if (cm->next == nullptr)
		return false;
	cm->next->reset(this);
	return hasValue();
}

int value_stream::getInt()
{
	if (rp >= rplimit)
	{
		// Advance to next coding segment.
		if (rp > rplimit || cm == nullptr || cm->next == nullptr)
		{
			// Must perform this check and throw an exception on bad input.
			unpack_abort(ERB);
			return 0;
		}
		cm->next->reset(this);
		return getInt();
	}

	CODING_PRIVATE(c.spec);
	uint32_t uval;
	enum
	{
		B5 = 5,
		B3 = 3,
		H128 = 128,
		H64 = 64,
		H4 = 4
	};
	switch (cmk)
	{
	case cmk_BHS:
		assert(D == 0);
		uval = coding::parse(rp, B, H);
		if (S == 0)
			return (int)uval;
		return decode_sign(S, uval);

	case cmk_BHS0:
		assert(S == 0 && D == 0);
		uval = coding::parse(rp, B, H);
		return (int)uval;

	case cmk_BHS1:
		assert(S == 1 && D == 0);
		uval = coding::parse(rp, B, H);
		return DECODE_SIGN_S1(uval);

	case cmk_BYTE1:
		assert(c.spec == BYTE1_spec);
		assert(B == 1 && H == 256 && S == 0 && D == 0);
		return *rp++ & 0xFF;

	case cmk_CHAR3:
		assert(c.spec == CHAR3_spec);
		assert(B == B3 && H == H128 && S == 0 && D == 0);
		return coding::parse_lgH(rp, B3, H128, 7);

	case cmk_UNSIGNED5:
		assert(c.spec == UNSIGNED5_spec);
		assert(B == B5 && H == H64 && S == 0 && D == 0);
		return coding::parse_lgH(rp, B5, H64, 6);

	case cmk_BHSD1:
		assert(D == 1);
		uval = coding::parse(rp, B, H);
		if (S != 0)
			uval = (uint32_t)decode_sign(S, uval);
		return getDeltaValue(this, uval, (bool)c.isSubrange);

	case cmk_BHS1D1full:
		assert(S == 1 && D == 1 && c.isFullRange);
		uval = coding::parse(rp, B, H);
		uval = (uint32_t)DECODE_SIGN_S1(uval);
		return getDeltaValue(this, uval, false);

	case cmk_BHS1D1sub:
		assert(S == 1 && D == 1 && c.isSubrange);
		uval = coding::parse(rp, B, H);
		uval = (uint32_t)DECODE_SIGN_S1(uval);
		return getDeltaValue(this, uval, true);

	case cmk_DELTA5:
		assert(c.spec == DELTA5_spec);
		assert(B == B5 && H == H64 && S == 1 && D == 1 && c.isFullRange);
		uval = coding::parse_lgH(rp, B5, H64, 6);
		sum += DECODE_SIGN_S1(uval);
		return sum;

	case cmk_BCI5:
		assert(c.spec == BCI5_spec);
		assert(B == B5 && H == H4 && S == 0 && D == 0);
		return coding::parse_lgH(rp, B5, H4, 2);

	case cmk_BRANCH5:
		assert(c.spec == BRANCH5_spec);
		assert(B == B5 && H == H4 && S == 2 && D == 0);
		uval = coding::parse_lgH(rp, B5, H4, 2);
		return decode_sign(S, uval);

	case cmk_pop:
		uval = coding::parse(rp, B, H);
		if (S != 0)
		{
			uval = (uint32_t)decode_sign(S, uval);
		}
		if (D != 0)
		{
			assert(c.isSubrange | c.isFullRange);
			if (c.isSubrange)
				sum = c.sumInUnsignedRange(sum, (int)uval);
			else
				sum += (int)uval;
			uval = (uint32_t)sum;
		}
		return getPopValue(this, uval);

	case cmk_pop_BHS0:
		assert(S == 0 && D == 0);
		uval = coding::parse(rp, B, H);
		return getPopValue(this, uval);

	case cmk_pop_BYTE1:
		assert(c.spec == BYTE1_spec);
		assert(B == 1 && H == 256 && S == 0 && D == 0);
		return getPopValue(this, *rp++ & 0xFF);

	default:
		break;
	}
	assert(false);
	return 0;
}

static int moreCentral(int x, int y)
{ // used to find end of Pop.{F}
	// Suggested implementation from the Pack200 specification:
	uint32_t kx = (x >> 31) ^ (x << 1);
	uint32_t ky = (y >> 31) ^ (y << 1);
	return (kx < ky ? x : y);
}
// static maybe_inline
// int moreCentral2(int x, int y, int min) {
//  // Strict implementation of buggy 150.7 specification.
//  // The bug is that the spec. says absolute-value ties are broken
//  // in favor of positive numbers, but the suggested implementation
//  // (also mentioned in the spec.) breaks ties in favor of negative numbers.
//  if ((x + y) != 0)
//    return min;
//  else
//    // return the other value, which breaks a tie in the positive direction
//    return (x > y)? x: y;
//}

static const byte *no_meta[] = {nullptr};
#define NO_META (*(byte **)no_meta)
enum
{
	POP_FAVORED_N = -2
};

// mode bits
#define DISABLE_RUN 1 // used immediately inside ACodee
#define DISABLE_POP 2 // used recursively in all pop sub-bands

// This function knows all about meta-coding.
void coding_method::init(byte *&band_rp, byte *band_limit, byte *&meta_rp, int mode,
						 coding *defc, int N, intlist *valueSink)
{
	assert(N != 0);

	assert(u != nullptr); // must be pre-initialized
	// if (u == nullptr)  u = unpacker::current();  // expensive

	int op = (meta_rp == nullptr) ? _meta_default : (*meta_rp++ & 0xFF);
	coding *foundc = nullptr;
	coding *to_free = nullptr;

	if (op == _meta_default)
	{
		foundc = defc;
		// and fall through
	}
	else if (op >= _meta_canon_min && op <= _meta_canon_max)
	{
		foundc = coding::findByIndex(op);
		// and fall through
	}
	else if (op == _meta_arb)
	{
		int args = (*meta_rp++ & 0xFF);
		// args = (D:[0..1] + 2*S[0..2] + 8*(B:[1..5]-1))
		int D = ((args >> 0) & 1);
		int S = ((args >> 1) & 3);
		int B = ((args >> 3) & -1) + 1;
		// & (H[1..256]-1)
		int H = (*meta_rp++ & 0xFF) + 1;
		foundc = coding::findBySpec(B, H, S, D);
		to_free = foundc; // findBySpec may dynamically allocate
		if (foundc == nullptr)
		{
			unpack_abort("illegal arbitrary coding");
			return;
		}
		// and fall through
	}
	else if (op >= _meta_run && op < _meta_pop)
	{
		int args = (op - _meta_run);
		// args: KX:[0..3] + 4*(KBFlag:[0..1]) + 8*(ABDef:[0..2])
		int KX = ((args >> 0) & 3);
		int KBFlag = ((args >> 2) & 1);
		int ABDef = ((args >> 3) & -1);
		assert(ABDef <= 2);
		// & KB: one of [0..255] if KBFlag=1
		int KB = (!KBFlag ? 3 : (*meta_rp++ & 0xFF));
		int K = (KB + 1) << (KX * 4);
		int N2 = (N >= 0) ? N - K : N;
		if (N == 0 || (N2 <= 0 && N2 != N))
		{
			unpack_abort("illegal run encoding");
		}
		if ((mode & DISABLE_RUN) != 0)
		{
			unpack_abort("illegal nested run encoding");
		}

		// & Enc{ ACode } if ADef=0  (ABDef != 1)
		// No direct nesting of 'run' in ACode, but in BCode it's OK.
		int disRun = mode | DISABLE_RUN;
		if (ABDef == 1)
		{
			this->init(band_rp, band_limit, NO_META, disRun, defc, K, valueSink);
		}
		else
		{
			this->init(band_rp, band_limit, meta_rp, disRun, defc, K, valueSink);
		}

		// & Enc{ BCode } if BDef=0  (ABDef != 2)
		coding_method *tail = U_NEW(coding_method, 1);
		if (!tail)
			return;
		tail->u = u;

		// The 'run' codings may be nested indirectly via 'pop' codings.
		// This means that this->next may already be filled in, if
		// ACode was of type 'pop' with a 'run' token coding.
		// No problem:  Just chain the upcoming BCode onto the end.
		for (coding_method *self = this;; self = self->next)
		{
			if (self->next == nullptr)
			{
				self->next = tail;
				break;
			}
		}

		if (ABDef == 2)
		{
			tail->init(band_rp, band_limit, NO_META, mode, defc, N2, valueSink);
		}
		else
		{
			tail->init(band_rp, band_limit, meta_rp, mode, defc, N2, valueSink);
		}
		// Note:  The preceding calls to init should be tail-recursive.

		return; // done; no falling through
	}
	else if (op >= _meta_pop && op < _meta_limit)
	{
		int args = (op - _meta_pop);
		// args: (FDef:[0..1]) + 2*UDef:[0..1] + 4*(TDefL:[0..11])
		int FDef = ((args >> 0) & 1);
		int UDef = ((args >> 1) & 1);
		int TDefL = ((args >> 2) & -1);
		assert(TDefL <= 11);
		int TDef = (TDefL > 0);
		int TL = (TDefL <= 6) ? (2 << TDefL) : (256 - (4 << (11 - TDefL)));
		int TH = (256 - TL);
		if (N <= 0)
		{
			unpack_abort("illegal pop encoding");
		}
		if ((mode & DISABLE_POP) != 0)
		{
			unpack_abort("illegal nested pop encoding");
		}

		// No indirect nesting of 'pop', but 'run' is OK.
		int disPop = DISABLE_POP;

		// & Enc{ FCode } if FDef=0
		int FN = POP_FAVORED_N;
		assert(valueSink == nullptr);
		intlist fValueSink;
		fValueSink.init();
		coding_method fval;
		BYTES_OF(fval).clear();
		fval.u = u;
		if (FDef != 0)
		{
			fval.init(band_rp, band_limit, NO_META, disPop, defc, FN, &fValueSink);
		}
		else
		{
			fval.init(band_rp, band_limit, meta_rp, disPop, defc, FN, &fValueSink);
		}
		bytes fvbuf;
		fValues = (u->saveTo(fvbuf, fValueSink.b), (int *)fvbuf.ptr);
		fVlength = fValueSink.length(); // i.e., the parameter K
		fValueSink.free();

		// Skip the first {F} run in all subsequent passes.
		// The next call to this->init(...) will set vs0.rp to point after the {F}.

		// & Enc{ TCode } if TDef=0  (TDefL==0)
		if (TDef != 0)
		{
			coding *tcode = coding::findBySpec(1, 256); // BYTE1
			// find the most narrowly sufficient code:
			for (int B = 2; B <= B_MAX; B++)
			{
				if (fVlength <= tcode->umax)
					break; // found it
				tcode->free();
				tcode = coding::findBySpec(B, TH);
				if (!tcode)
					return;
			}
			if (!(fVlength <= tcode->umax))
			{
				unpack_abort("pop.L value too small");
			}
			this->init(band_rp, band_limit, NO_META, disPop, tcode, N, nullptr);
			tcode->free();
		}
		else
		{
			this->init(band_rp, band_limit, meta_rp, disPop, defc, N, nullptr);
		}

		// Count the number of zero tokens right now.
		// Also verify that they are in bounds.
		int UN = 0; // one {U} for each zero in {T}
		value_stream vs = vs0;
		for (int i = 0; i < N; i++)
		{
			uint32_t val = vs.getInt();
			if (val == 0)
				UN += 1;
			if (!(val <= (uint32_t)fVlength))
			{
				unpack_abort("pop token out of range");
			}
		}
		vs.done();

		// & Enc{ UCode } if UDef=0
		if (UN != 0)
		{
			uValues = U_NEW(coding_method, 1);
			if (uValues == nullptr)
				return;
			uValues->u = u;
			if (UDef != 0)
			{
				uValues->init(band_rp, band_limit, NO_META, disPop, defc, UN, nullptr);
			}
			else
			{
				uValues->init(band_rp, band_limit, meta_rp, disPop, defc, UN, nullptr);
			}
		}
		else
		{
			if (UDef == 0)
			{
				int uop = (*meta_rp++ & 0xFF);
				if (uop > _meta_canon_max)
					// %%% Spec. requires the more strict (uop != _meta_default).
					unpack_abort("bad meta-coding for empty pop/U");
			}
		}

		// Bug fix for 6259542
		// Last of all, adjust vs0.cmk to the 'pop' flavor
		for (coding_method *self = this; self != nullptr; self = self->next)
		{
			coding_method_kind cmk2 = cmk_pop;
			switch (self->vs0.cmk)
			{
			case cmk_BHS0:
				cmk2 = cmk_pop_BHS0;
				break;
			case cmk_BYTE1:
				cmk2 = cmk_pop_BYTE1;
				break;
			default:
				break;
			}
			self->vs0.cmk = cmk2;
			if (self != this)
			{
				assert(self->fValues == nullptr); // no double init
				self->fValues = this->fValues;
				self->fVlength = this->fVlength;
				assert(self->uValues == nullptr); // must stay nullptr
			}
		}

		return; // done; no falling through
	}
	else
	{
		unpack_abort("bad meta-coding");
	}

	// Common code here skips a series of values with one coding.
	assert(foundc != nullptr);

	assert(vs0.cmk == cmk_ERROR);   // no garbage, please
	assert(vs0.rp == nullptr);	  // no garbage, please
	assert(vs0.rplimit == nullptr); // no garbage, please
	assert(vs0.sum == 0);		   // no garbage, please

	vs0.init(band_rp, band_limit, foundc);

	// Done with foundc.  Free if necessary.
	if (to_free != nullptr)
	{
		to_free->free();
		to_free = nullptr;
	}
	foundc = nullptr;

	coding &c = vs0.c;
	CODING_PRIVATE(c.spec);
	// assert sane N
	assert((uint32_t)N < INT_MAX_VALUE || N == POP_FAVORED_N);

	// Look at the values, or at least skip over them quickly.
	if (valueSink == nullptr)
	{
		// Skip and ignore values in the first pass.
		c.parseMultiple(band_rp, N, band_limit, B, H);
	}
	else if (N >= 0)
	{
		// Pop coding, {F} sequence, initial run of values...
		assert((mode & DISABLE_POP) != 0);
		value_stream vs = vs0;
		for (int n = 0; n < N; n++)
		{
			int val = vs.getInt();
			valueSink->add(val);
		}
		band_rp = vs.rp;
	}
	else
	{
		// Pop coding, {F} sequence, final run of values...
		assert((mode & DISABLE_POP) != 0);
		assert(N == POP_FAVORED_N);
		int min = INT_MIN_VALUE; // farthest from the center
		// min2 is based on the buggy specification of centrality in version 150.7
		// no known implementations transmit this value, but just in case...
		// int min2 = INT_MIN_VALUE;
		int last = 0;
		// if there were initial runs, find the potential sentinels in them:
		for (int i = 0; i < valueSink->length(); i++)
		{
			last = valueSink->get(i);
			min = moreCentral(min, last);
			// min2 = moreCentral2(min2, last, min);
		}
		value_stream vs = vs0;
		for (;;)
		{
			int val = vs.getInt();
			if (valueSink->length() > 0 && (val == last || val == min)) //|| val == min2
				break;
			valueSink->add(val);
			last = val;
			min = moreCentral(min, last);
			// min2 = moreCentral2(min2, last, min);
		}
		band_rp = vs.rp;
	}

	// Get an accurate upper limit now.
	vs0.rplimit = band_rp;
	vs0.cm = this;

	return; // success
}

coding basic_codings[] = {
	// This one is not a usable irregular coding, but is used by cp_Utf8_chars.
	CODING_INIT(3, 128, 0, 0),

	// Fixed-length codings:
	CODING_INIT(1, 256, 0, 0), CODING_INIT(1, 256, 1, 0), CODING_INIT(1, 256, 0, 1),
	CODING_INIT(1, 256, 1, 1), CODING_INIT(2, 256, 0, 0), CODING_INIT(2, 256, 1, 0),
	CODING_INIT(2, 256, 0, 1), CODING_INIT(2, 256, 1, 1), CODING_INIT(3, 256, 0, 0),
	CODING_INIT(3, 256, 1, 0), CODING_INIT(3, 256, 0, 1), CODING_INIT(3, 256, 1, 1),
	CODING_INIT(4, 256, 0, 0), CODING_INIT(4, 256, 1, 0), CODING_INIT(4, 256, 0, 1),
	CODING_INIT(4, 256, 1, 1),

	// Full-range variable-length codings:
	CODING_INIT(5, 4, 0, 0),   CODING_INIT(5, 4, 1, 0),   CODING_INIT(5, 4, 2, 0),
	CODING_INIT(5, 16, 0, 0),  CODING_INIT(5, 16, 1, 0),  CODING_INIT(5, 16, 2, 0),
	CODING_INIT(5, 32, 0, 0),  CODING_INIT(5, 32, 1, 0),  CODING_INIT(5, 32, 2, 0),
	CODING_INIT(5, 64, 0, 0),  CODING_INIT(5, 64, 1, 0),  CODING_INIT(5, 64, 2, 0),
	CODING_INIT(5, 128, 0, 0), CODING_INIT(5, 128, 1, 0), CODING_INIT(5, 128, 2, 0),
	CODING_INIT(5, 4, 0, 1),   CODING_INIT(5, 4, 1, 1),   CODING_INIT(5, 4, 2, 1),
	CODING_INIT(5, 16, 0, 1),  CODING_INIT(5, 16, 1, 1),  CODING_INIT(5, 16, 2, 1),
	CODING_INIT(5, 32, 0, 1),  CODING_INIT(5, 32, 1, 1),  CODING_INIT(5, 32, 2, 1),
	CODING_INIT(5, 64, 0, 1),  CODING_INIT(5, 64, 1, 1),  CODING_INIT(5, 64, 2, 1),
	CODING_INIT(5, 128, 0, 1), CODING_INIT(5, 128, 1, 1), CODING_INIT(5, 128, 2, 1),

	// Variable length subrange codings:
	CODING_INIT(2, 192, 0, 0), CODING_INIT(2, 224, 0, 0), CODING_INIT(2, 240, 0, 0),
	CODING_INIT(2, 248, 0, 0), CODING_INIT(2, 252, 0, 0), CODING_INIT(2, 8, 0, 1),
	CODING_INIT(2, 8, 1, 1),   CODING_INIT(2, 16, 0, 1),  CODING_INIT(2, 16, 1, 1),
	CODING_INIT(2, 32, 0, 1),  CODING_INIT(2, 32, 1, 1),  CODING_INIT(2, 64, 0, 1),
	CODING_INIT(2, 64, 1, 1),  CODING_INIT(2, 128, 0, 1), CODING_INIT(2, 128, 1, 1),
	CODING_INIT(2, 192, 0, 1), CODING_INIT(2, 192, 1, 1), CODING_INIT(2, 224, 0, 1),
	CODING_INIT(2, 224, 1, 1), CODING_INIT(2, 240, 0, 1), CODING_INIT(2, 240, 1, 1),
	CODING_INIT(2, 248, 0, 1), CODING_INIT(2, 248, 1, 1), CODING_INIT(3, 192, 0, 0),
	CODING_INIT(3, 224, 0, 0), CODING_INIT(3, 240, 0, 0), CODING_INIT(3, 248, 0, 0),
	CODING_INIT(3, 252, 0, 0), CODING_INIT(3, 8, 0, 1),   CODING_INIT(3, 8, 1, 1),
	CODING_INIT(3, 16, 0, 1),  CODING_INIT(3, 16, 1, 1),  CODING_INIT(3, 32, 0, 1),
	CODING_INIT(3, 32, 1, 1),  CODING_INIT(3, 64, 0, 1),  CODING_INIT(3, 64, 1, 1),
	CODING_INIT(3, 128, 0, 1), CODING_INIT(3, 128, 1, 1), CODING_INIT(3, 192, 0, 1),
	CODING_INIT(3, 192, 1, 1), CODING_INIT(3, 224, 0, 1), CODING_INIT(3, 224, 1, 1),
	CODING_INIT(3, 240, 0, 1), CODING_INIT(3, 240, 1, 1), CODING_INIT(3, 248, 0, 1),
	CODING_INIT(3, 248, 1, 1), CODING_INIT(4, 192, 0, 0), CODING_INIT(4, 224, 0, 0),
	CODING_INIT(4, 240, 0, 0), CODING_INIT(4, 248, 0, 0), CODING_INIT(4, 252, 0, 0),
	CODING_INIT(4, 8, 0, 1),   CODING_INIT(4, 8, 1, 1),   CODING_INIT(4, 16, 0, 1),
	CODING_INIT(4, 16, 1, 1),  CODING_INIT(4, 32, 0, 1),  CODING_INIT(4, 32, 1, 1),
	CODING_INIT(4, 64, 0, 1),  CODING_INIT(4, 64, 1, 1),  CODING_INIT(4, 128, 0, 1),
	CODING_INIT(4, 128, 1, 1), CODING_INIT(4, 192, 0, 1), CODING_INIT(4, 192, 1, 1),
	CODING_INIT(4, 224, 0, 1), CODING_INIT(4, 224, 1, 1), CODING_INIT(4, 240, 0, 1),
	CODING_INIT(4, 240, 1, 1), CODING_INIT(4, 248, 0, 1), CODING_INIT(4, 248, 1, 1),
	CODING_INIT(0, 0, 0, 0)};
#define BASIC_INDEX_LIMIT (int)(sizeof(basic_codings) / sizeof(basic_codings[0]) - 1)

coding *coding::findByIndex(int idx)
{
	int index_limit = BASIC_INDEX_LIMIT;
	assert(_meta_canon_min == 1 && _meta_canon_max + 1 == index_limit);

	if (idx >= _meta_canon_min && idx <= _meta_canon_max)
		return basic_codings[idx].init();
	else
		return nullptr;
}