xref: /freebsd/contrib/bc/src/num.c (revision 7fdf597e96a02165cfe22ff357b857d5fa15ed8a)
1 /*
2  * *****************************************************************************
3  *
4  * SPDX-License-Identifier: BSD-2-Clause
5  *
6  * Copyright (c) 2018-2024 Gavin D. Howard and contributors.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions are met:
10  *
11  * * Redistributions of source code must retain the above copyright notice, this
12  *   list of conditions and the following disclaimer.
13  *
14  * * Redistributions in binary form must reproduce the above copyright notice,
15  *   this list of conditions and the following disclaimer in the documentation
16  *   and/or other materials provided with the distribution.
17  *
18  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
19  * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
20  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
21  * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
22  * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
23  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
24  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
25  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
26  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
27  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
28  * POSSIBILITY OF SUCH DAMAGE.
29  *
30  * *****************************************************************************
31  *
32  * Code for the number type.
33  *
34  */
35 
36 #include <assert.h>
37 #include <ctype.h>
38 #include <stdbool.h>
39 #include <stdlib.h>
40 #include <string.h>
41 #include <setjmp.h>
42 #include <limits.h>
43 
44 #include <num.h>
45 #include <rand.h>
46 #include <vm.h>
47 #if BC_ENABLE_LIBRARY
48 #include <library.h>
49 #endif // BC_ENABLE_LIBRARY
50 
51 // Before you try to understand this code, see the development manual
52 // (manuals/development.md#numbers).
53 
54 static void
55 bc_num_m(BcNum* a, BcNum* b, BcNum* restrict c, size_t scale);
56 
57 /**
58  * Multiply two numbers and throw a math error if they overflow.
59  * @param a  The first operand.
60  * @param b  The second operand.
61  * @return   The product of the two operands.
62  */
63 static inline size_t
64 bc_num_mulOverflow(size_t a, size_t b)
65 {
66 	size_t res = a * b;
67 	if (BC_ERR(BC_VM_MUL_OVERFLOW(a, b, res))) bc_err(BC_ERR_MATH_OVERFLOW);
68 	return res;
69 }
70 
71 /**
72  * Conditionally negate @a n based on @a neg. Algorithm taken from
73  * https://graphics.stanford.edu/~seander/bithacks.html#ConditionalNegate .
74  * @param n    The value to turn into a signed value and negate.
75  * @param neg  The condition to negate or not.
76  */
77 static inline ssize_t
78 bc_num_neg(size_t n, bool neg)
79 {
80 	return (((ssize_t) n) ^ -((ssize_t) neg)) + neg;
81 }
82 
83 /**
84  * Compare a BcNum against zero.
85  * @param n  The number to compare.
86  * @return   -1 if the number is less than 0, 1 if greater, and 0 if equal.
87  */
88 ssize_t
89 bc_num_cmpZero(const BcNum* n)
90 {
91 	return bc_num_neg((n)->len != 0, BC_NUM_NEG(n));
92 }
93 
94 /**
95  * Return the number of integer limbs in a BcNum. This is the opposite of rdx.
96  * @param n  The number to return the amount of integer limbs for.
97  * @return   The amount of integer limbs in @a n.
98  */
99 static inline size_t
100 bc_num_int(const BcNum* n)
101 {
102 	return n->len ? n->len - BC_NUM_RDX_VAL(n) : 0;
103 }
104 
105 /**
106  * Expand a number's allocation capacity to at least req limbs.
107  * @param n    The number to expand.
108  * @param req  The number limbs to expand the allocation capacity to.
109  */
110 static void
111 bc_num_expand(BcNum* restrict n, size_t req)
112 {
113 	assert(n != NULL);
114 
115 	req = req >= BC_NUM_DEF_SIZE ? req : BC_NUM_DEF_SIZE;
116 
117 	if (req > n->cap)
118 	{
119 		BC_SIG_LOCK;
120 
121 		n->num = bc_vm_realloc(n->num, BC_NUM_SIZE(req));
122 		n->cap = req;
123 
124 		BC_SIG_UNLOCK;
125 	}
126 }
127 
128 /**
129  * Set a number to 0 with the specified scale.
130  * @param n      The number to set to zero.
131  * @param scale  The scale to set the number to.
132  */
133 static inline void
134 bc_num_setToZero(BcNum* restrict n, size_t scale)
135 {
136 	assert(n != NULL);
137 	n->scale = scale;
138 	n->len = n->rdx = 0;
139 }
140 
141 void
142 bc_num_zero(BcNum* restrict n)
143 {
144 	bc_num_setToZero(n, 0);
145 }
146 
147 void
148 bc_num_one(BcNum* restrict n)
149 {
150 	bc_num_zero(n);
151 	n->len = 1;
152 	n->num[0] = 1;
153 }
154 
155 /**
156  * "Cleans" a number, which means reducing the length if the most significant
157  * limbs are zero.
158  * @param n  The number to clean.
159  */
160 static void
161 bc_num_clean(BcNum* restrict n)
162 {
163 	// Reduce the length.
164 	while (BC_NUM_NONZERO(n) && !n->num[n->len - 1])
165 	{
166 		n->len -= 1;
167 	}
168 
169 	// Special cases.
170 	if (BC_NUM_ZERO(n)) n->rdx = 0;
171 	else
172 	{
173 		// len must be at least as much as rdx.
174 		size_t rdx = BC_NUM_RDX_VAL(n);
175 		if (n->len < rdx) n->len = rdx;
176 	}
177 }
178 
179 /**
180  * Returns the log base 10 of @a i. I could have done this with floating-point
181  * math, and in fact, I originally did. However, that was the only
182  * floating-point code in the entire codebase, and I decided I didn't want any.
183  * This is fast enough. Also, it might handle larger numbers better.
184  * @param i  The number to return the log base 10 of.
185  * @return   The log base 10 of @a i.
186  */
187 static size_t
188 bc_num_log10(size_t i)
189 {
190 	size_t len;
191 
192 	for (len = 1; i; i /= BC_BASE, ++len)
193 	{
194 		continue;
195 	}
196 
197 	assert(len - 1 <= BC_BASE_DIGS + 1);
198 
199 	return len - 1;
200 }
201 
202 /**
203  * Returns the number of decimal digits in a limb that are zero starting at the
204  * most significant digits. This basically returns how much of the limb is used.
205  * @param n  The number.
206  * @return   The number of decimal digits that are 0 starting at the most
207  *           significant digits.
208  */
209 static inline size_t
210 bc_num_zeroDigits(const BcDig* n)
211 {
212 	assert(*n >= 0);
213 	assert(((size_t) *n) < BC_BASE_POW);
214 	return BC_BASE_DIGS - bc_num_log10((size_t) *n);
215 }
216 
217 /**
218  * Returns the power of 10 that the least significant limb should be multiplied
219  * by to put its digits in the right place. For example, if the scale only
220  * reaches 8 places into the limb, this will return 1 (because it should be
221  * multiplied by 10^1) to put the number in the correct place.
222  * @param scale  The scale.
223  * @return       The power of 10 that the least significant limb should be
224  *               multiplied by
225  */
226 static inline size_t
227 bc_num_leastSigPow(size_t scale)
228 {
229 	size_t digs;
230 
231 	digs = scale % BC_BASE_DIGS;
232 	digs = digs != 0 ? BC_BASE_DIGS - digs : 0;
233 
234 	return bc_num_pow10[digs];
235 }
236 
237 /**
238  * Return the total number of integer digits in a number. This is the opposite
239  * of scale, like bc_num_int() is the opposite of rdx.
240  * @param n  The number.
241  * @return   The number of integer digits in @a n.
242  */
243 static size_t
244 bc_num_intDigits(const BcNum* n)
245 {
246 	size_t digits = bc_num_int(n) * BC_BASE_DIGS;
247 	if (digits > 0) digits -= bc_num_zeroDigits(n->num + n->len - 1);
248 	return digits;
249 }
250 
251 /**
252  * Returns the number of limbs of a number that are non-zero starting at the
253  * most significant limbs. This expects that there are *no* integer limbs in the
254  * number because it is specifically to figure out how many zero limbs after the
255  * decimal place to ignore. If there are zero limbs after non-zero limbs, they
256  * are counted as non-zero limbs.
257  * @param n  The number.
258  * @return   The number of non-zero limbs after the decimal point.
259  */
260 static size_t
261 bc_num_nonZeroLen(const BcNum* restrict n)
262 {
263 	size_t i, len = n->len;
264 
265 	assert(len == BC_NUM_RDX_VAL(n));
266 
267 	for (i = len - 1; i < len && !n->num[i]; --i)
268 	{
269 		continue;
270 	}
271 
272 	assert(i + 1 > 0);
273 
274 	return i + 1;
275 }
276 
277 #if BC_ENABLE_EXTRA_MATH
278 
279 /**
280  * Returns the power of 10 that a number with an absolute value less than 1
281  * needs to be multiplied by in order to be greater than 1 or less than -1.
282  * @param n  The number.
283  * @return   The power of 10 that a number greater than 1 and less than -1 must
284  *           be multiplied by to be greater than 1 or less than -1.
285  */
286 static size_t
287 bc_num_negPow10(const BcNum* restrict n)
288 {
289 	// Figure out how many limbs after the decimal point is zero.
290 	size_t i, places, idx = bc_num_nonZeroLen(n) - 1;
291 
292 	places = 1;
293 
294 	// Figure out how much in the last limb is zero.
295 	for (i = BC_BASE_DIGS - 1; i < BC_BASE_DIGS; --i)
296 	{
297 		if (bc_num_pow10[i] > (BcBigDig) n->num[idx]) places += 1;
298 		else break;
299 	}
300 
301 	// Calculate the combination of zero limbs and zero digits in the last
302 	// limb.
303 	return places + (BC_NUM_RDX_VAL(n) - (idx + 1)) * BC_BASE_DIGS;
304 }
305 
306 #endif // BC_ENABLE_EXTRA_MATH
307 
308 /**
309  * Performs a one-limb add with a carry.
310  * @param a      The first limb.
311  * @param b      The second limb.
312  * @param carry  An in/out parameter; the carry in from the previous add and the
313  *               carry out from this add.
314  * @return       The resulting limb sum.
315  */
316 static BcDig
317 bc_num_addDigits(BcDig a, BcDig b, bool* carry)
318 {
319 	assert(((BcBigDig) BC_BASE_POW) * 2 == ((BcDig) BC_BASE_POW) * 2);
320 	assert(a < BC_BASE_POW && a >= 0);
321 	assert(b < BC_BASE_POW && b >= 0);
322 
323 	a += b + *carry;
324 	*carry = (a >= BC_BASE_POW);
325 	if (*carry) a -= BC_BASE_POW;
326 
327 	assert(a >= 0);
328 	assert(a < BC_BASE_POW);
329 
330 	return a;
331 }
332 
333 /**
334  * Performs a one-limb subtract with a carry.
335  * @param a      The first limb.
336  * @param b      The second limb.
337  * @param carry  An in/out parameter; the carry in from the previous subtract
338  *               and the carry out from this subtract.
339  * @return       The resulting limb difference.
340  */
341 static BcDig
342 bc_num_subDigits(BcDig a, BcDig b, bool* carry)
343 {
344 	assert(a < BC_BASE_POW && a >= 0);
345 	assert(b < BC_BASE_POW && b >= 0);
346 
347 	b += *carry;
348 	*carry = (a < b);
349 	if (*carry) a += BC_BASE_POW;
350 
351 	assert(a - b >= 0);
352 	assert(a - b < BC_BASE_POW);
353 
354 	return a - b;
355 }
356 
357 /**
358  * Add two BcDig arrays and store the result in the first array.
359  * @param a    The first operand and out array.
360  * @param b    The second operand.
361  * @param len  The length of @a b.
362  */
363 static void
364 bc_num_addArrays(BcDig* restrict a, const BcDig* restrict b, size_t len)
365 {
366 	size_t i;
367 	bool carry = false;
368 
369 	for (i = 0; i < len; ++i)
370 	{
371 		a[i] = bc_num_addDigits(a[i], b[i], &carry);
372 	}
373 
374 	// Take care of the extra limbs in the bigger array.
375 	for (; carry; ++i)
376 	{
377 		a[i] = bc_num_addDigits(a[i], 0, &carry);
378 	}
379 }
380 
381 /**
382  * Subtract two BcDig arrays and store the result in the first array.
383  * @param a    The first operand and out array.
384  * @param b    The second operand.
385  * @param len  The length of @a b.
386  */
387 static void
388 bc_num_subArrays(BcDig* restrict a, const BcDig* restrict b, size_t len)
389 {
390 	size_t i;
391 	bool carry = false;
392 
393 	for (i = 0; i < len; ++i)
394 	{
395 		a[i] = bc_num_subDigits(a[i], b[i], &carry);
396 	}
397 
398 	// Take care of the extra limbs in the bigger array.
399 	for (; carry; ++i)
400 	{
401 		a[i] = bc_num_subDigits(a[i], 0, &carry);
402 	}
403 }
404 
405 /**
406  * Multiply a BcNum array by a one-limb number. This is a faster version of
407  * multiplication for when we can use it.
408  * @param a  The BcNum to multiply by the one-limb number.
409  * @param b  The one limb of the one-limb number.
410  * @param c  The return parameter.
411  */
412 static void
413 bc_num_mulArray(const BcNum* restrict a, BcBigDig b, BcNum* restrict c)
414 {
415 	size_t i;
416 	BcBigDig carry = 0;
417 
418 	assert(b <= BC_BASE_POW);
419 
420 	// Make sure the return parameter is big enough.
421 	if (a->len + 1 > c->cap) bc_num_expand(c, a->len + 1);
422 
423 	// We want the entire return parameter to be zero for cleaning later.
424 	// NOLINTNEXTLINE
425 	memset(c->num, 0, BC_NUM_SIZE(c->cap));
426 
427 	// Actual multiplication loop.
428 	for (i = 0; i < a->len; ++i)
429 	{
430 		BcBigDig in = ((BcBigDig) a->num[i]) * b + carry;
431 		c->num[i] = in % BC_BASE_POW;
432 		carry = in / BC_BASE_POW;
433 	}
434 
435 	assert(carry < BC_BASE_POW);
436 
437 	// Finishing touches.
438 	c->num[i] = (BcDig) carry;
439 	assert(c->num[i] >= 0 && c->num[i] < BC_BASE_POW);
440 	c->len = a->len;
441 	c->len += (carry != 0);
442 
443 	bc_num_clean(c);
444 
445 	// Postconditions.
446 	assert(!BC_NUM_NEG(c) || BC_NUM_NONZERO(c));
447 	assert(BC_NUM_RDX_VAL(c) <= c->len || !c->len);
448 	assert(!c->len || c->num[c->len - 1] || BC_NUM_RDX_VAL(c) == c->len);
449 }
450 
451 /**
452  * Divide a BcNum array by a one-limb number. This is a faster version of divide
453  * for when we can use it.
454  * @param a    The BcNum to multiply by the one-limb number.
455  * @param b    The one limb of the one-limb number.
456  * @param c    The return parameter for the quotient.
457  * @param rem  The return parameter for the remainder.
458  */
459 static void
460 bc_num_divArray(const BcNum* restrict a, BcBigDig b, BcNum* restrict c,
461                 BcBigDig* rem)
462 {
463 	size_t i;
464 	BcBigDig carry = 0;
465 
466 	assert(c->cap >= a->len);
467 
468 	// Actual division loop.
469 	for (i = a->len - 1; i < a->len; --i)
470 	{
471 		BcBigDig in = ((BcBigDig) a->num[i]) + carry * BC_BASE_POW;
472 		assert(in / b < BC_BASE_POW);
473 		c->num[i] = (BcDig) (in / b);
474 		assert(c->num[i] >= 0 && c->num[i] < BC_BASE_POW);
475 		carry = in % b;
476 	}
477 
478 	// Finishing touches.
479 	c->len = a->len;
480 	bc_num_clean(c);
481 	*rem = carry;
482 
483 	// Postconditions.
484 	assert(!BC_NUM_NEG(c) || BC_NUM_NONZERO(c));
485 	assert(BC_NUM_RDX_VAL(c) <= c->len || !c->len);
486 	assert(!c->len || c->num[c->len - 1] || BC_NUM_RDX_VAL(c) == c->len);
487 }
488 
489 /**
490  * Compare two BcDig arrays and return >0 if @a b is greater, <0 if @a b is
491  * less, and 0 if equal. Both @a a and @a b must have the same length.
492  * @param a    The first array.
493  * @param b    The second array.
494  * @param len  The minimum length of the arrays.
495  */
496 static ssize_t
497 bc_num_compare(const BcDig* restrict a, const BcDig* restrict b, size_t len)
498 {
499 	size_t i;
500 	BcDig c = 0;
501 	for (i = len - 1; i < len && !(c = a[i] - b[i]); --i)
502 	{
503 		continue;
504 	}
505 	return bc_num_neg(i + 1, c < 0);
506 }
507 
508 ssize_t
509 bc_num_cmp(const BcNum* a, const BcNum* b)
510 {
511 	size_t i, min, a_int, b_int, diff, ardx, brdx;
512 	BcDig* max_num;
513 	BcDig* min_num;
514 	bool a_max, neg = false;
515 	ssize_t cmp;
516 
517 	assert(a != NULL && b != NULL);
518 
519 	// Same num? Equal.
520 	if (a == b) return 0;
521 
522 	// Easy cases.
523 	if (BC_NUM_ZERO(a)) return bc_num_neg(b->len != 0, !BC_NUM_NEG(b));
524 	if (BC_NUM_ZERO(b)) return bc_num_cmpZero(a);
525 	if (BC_NUM_NEG(a))
526 	{
527 		if (BC_NUM_NEG(b)) neg = true;
528 		else return -1;
529 	}
530 	else if (BC_NUM_NEG(b)) return 1;
531 
532 	// Get the number of int limbs in each number and get the difference.
533 	a_int = bc_num_int(a);
534 	b_int = bc_num_int(b);
535 	a_int -= b_int;
536 
537 	// If there's a difference, then just return the comparison.
538 	if (a_int) return neg ? -((ssize_t) a_int) : (ssize_t) a_int;
539 
540 	// Get the rdx's and figure out the max.
541 	ardx = BC_NUM_RDX_VAL(a);
542 	brdx = BC_NUM_RDX_VAL(b);
543 	a_max = (ardx > brdx);
544 
545 	// Set variables based on the above.
546 	if (a_max)
547 	{
548 		min = brdx;
549 		diff = ardx - brdx;
550 		max_num = a->num + diff;
551 		min_num = b->num;
552 	}
553 	else
554 	{
555 		min = ardx;
556 		diff = brdx - ardx;
557 		max_num = b->num + diff;
558 		min_num = a->num;
559 	}
560 
561 	// Do a full limb-by-limb comparison.
562 	cmp = bc_num_compare(max_num, min_num, b_int + min);
563 
564 	// If we found a difference, return it based on state.
565 	if (cmp) return bc_num_neg((size_t) cmp, !a_max == !neg);
566 
567 	// If there was no difference, then the final step is to check which number
568 	// has greater or lesser limbs beyond the other's.
569 	for (max_num -= diff, i = diff - 1; i < diff; --i)
570 	{
571 		if (max_num[i]) return bc_num_neg(1, !a_max == !neg);
572 	}
573 
574 	return 0;
575 }
576 
577 void
578 bc_num_truncate(BcNum* restrict n, size_t places)
579 {
580 	size_t nrdx, places_rdx;
581 
582 	if (!places) return;
583 
584 	// Grab these needed values; places_rdx is the rdx equivalent to places like
585 	// rdx is to scale.
586 	nrdx = BC_NUM_RDX_VAL(n);
587 	places_rdx = nrdx ? nrdx - BC_NUM_RDX(n->scale - places) : 0;
588 
589 	// We cannot truncate more places than we have.
590 	assert(places <= n->scale && (BC_NUM_ZERO(n) || places_rdx <= n->len));
591 
592 	n->scale -= places;
593 	BC_NUM_RDX_SET(n, nrdx - places_rdx);
594 
595 	// Only when the number is nonzero do we need to do the hard stuff.
596 	if (BC_NUM_NONZERO(n))
597 	{
598 		size_t pow;
599 
600 		// This calculates how many decimal digits are in the least significant
601 		// limb, then gets the power for that.
602 		pow = bc_num_leastSigPow(n->scale);
603 
604 		n->len -= places_rdx;
605 
606 		// We have to move limbs to maintain invariants. The limbs must begin at
607 		// the beginning of the BcNum array.
608 		// NOLINTNEXTLINE
609 		memmove(n->num, n->num + places_rdx, BC_NUM_SIZE(n->len));
610 
611 		// Clear the lower part of the last digit.
612 		if (BC_NUM_NONZERO(n)) n->num[0] -= n->num[0] % (BcDig) pow;
613 
614 		bc_num_clean(n);
615 	}
616 }
617 
618 void
619 bc_num_extend(BcNum* restrict n, size_t places)
620 {
621 	size_t nrdx, places_rdx;
622 
623 	if (!places) return;
624 
625 	// Easy case with zero; set the scale.
626 	if (BC_NUM_ZERO(n))
627 	{
628 		n->scale += places;
629 		return;
630 	}
631 
632 	// Grab these needed values; places_rdx is the rdx equivalent to places like
633 	// rdx is to scale.
634 	nrdx = BC_NUM_RDX_VAL(n);
635 	places_rdx = BC_NUM_RDX(places + n->scale) - nrdx;
636 
637 	// This is the hard case. We need to expand the number, move the limbs, and
638 	// set the limbs that were just cleared.
639 	if (places_rdx)
640 	{
641 		bc_num_expand(n, bc_vm_growSize(n->len, places_rdx));
642 		// NOLINTNEXTLINE
643 		memmove(n->num + places_rdx, n->num, BC_NUM_SIZE(n->len));
644 		// NOLINTNEXTLINE
645 		memset(n->num, 0, BC_NUM_SIZE(places_rdx));
646 	}
647 
648 	// Finally, set scale and rdx.
649 	BC_NUM_RDX_SET(n, nrdx + places_rdx);
650 	n->scale += places;
651 	n->len += places_rdx;
652 
653 	assert(BC_NUM_RDX_VAL(n) == BC_NUM_RDX(n->scale));
654 }
655 
656 /**
657  * Retires (finishes) a multiplication or division operation.
658  */
659 static void
660 bc_num_retireMul(BcNum* restrict n, size_t scale, bool neg1, bool neg2)
661 {
662 	// Make sure scale is correct.
663 	if (n->scale < scale) bc_num_extend(n, scale - n->scale);
664 	else bc_num_truncate(n, n->scale - scale);
665 
666 	bc_num_clean(n);
667 
668 	// We need to ensure rdx is correct.
669 	if (BC_NUM_NONZERO(n)) n->rdx = BC_NUM_NEG_VAL(n, !neg1 != !neg2);
670 }
671 
672 /**
673  * Splits a number into two BcNum's. This is used in Karatsuba.
674  * @param n    The number to split.
675  * @param idx  The index to split at.
676  * @param a    An out parameter; the low part of @a n.
677  * @param b    An out parameter; the high part of @a n.
678  */
679 static void
680 bc_num_split(const BcNum* restrict n, size_t idx, BcNum* restrict a,
681              BcNum* restrict b)
682 {
683 	// We want a and b to be clear.
684 	assert(BC_NUM_ZERO(a));
685 	assert(BC_NUM_ZERO(b));
686 
687 	// The usual case.
688 	if (idx < n->len)
689 	{
690 		// Set the fields first.
691 		b->len = n->len - idx;
692 		a->len = idx;
693 		a->scale = b->scale = 0;
694 		BC_NUM_RDX_SET(a, 0);
695 		BC_NUM_RDX_SET(b, 0);
696 
697 		assert(a->cap >= a->len);
698 		assert(b->cap >= b->len);
699 
700 		// Copy the arrays. This is not necessary for safety, but it is faster,
701 		// for some reason.
702 		// NOLINTNEXTLINE
703 		memcpy(b->num, n->num + idx, BC_NUM_SIZE(b->len));
704 		// NOLINTNEXTLINE
705 		memcpy(a->num, n->num, BC_NUM_SIZE(idx));
706 
707 		bc_num_clean(b);
708 	}
709 	// If the index is weird, just skip the split.
710 	else bc_num_copy(a, n);
711 
712 	bc_num_clean(a);
713 }
714 
715 /**
716  * Copies a number into another, but shifts the rdx so that the result number
717  * only sees the integer part of the source number.
718  * @param n  The number to copy.
719  * @param r  The result number with a shifted rdx, len, and num.
720  */
721 static void
722 bc_num_shiftRdx(const BcNum* restrict n, BcNum* restrict r)
723 {
724 	size_t rdx = BC_NUM_RDX_VAL(n);
725 
726 	r->len = n->len - rdx;
727 	r->cap = n->cap - rdx;
728 	r->num = n->num + rdx;
729 
730 	BC_NUM_RDX_SET_NEG(r, 0, BC_NUM_NEG(n));
731 	r->scale = 0;
732 }
733 
734 /**
735  * Shifts a number so that all of the least significant limbs of the number are
736  * skipped. This must be undone by bc_num_unshiftZero().
737  * @param n  The number to shift.
738  */
739 static size_t
740 bc_num_shiftZero(BcNum* restrict n)
741 {
742 	// This is volatile to quiet a GCC warning about longjmp() clobbering.
743 	volatile size_t i;
744 
745 	// If we don't have an integer, that is a problem, but it's also a bug
746 	// because the caller should have set everything up right.
747 	assert(!BC_NUM_RDX_VAL(n) || BC_NUM_ZERO(n));
748 
749 	for (i = 0; i < n->len && !n->num[i]; ++i)
750 	{
751 		continue;
752 	}
753 
754 	n->len -= i;
755 	n->num += i;
756 
757 	return i;
758 }
759 
760 /**
761  * Undo the damage done by bc_num_unshiftZero(). This must be called like all
762  * cleanup functions: after a label used by setjmp() and longjmp().
763  * @param n           The number to unshift.
764  * @param places_rdx  The amount the number was originally shift.
765  */
766 static void
767 bc_num_unshiftZero(BcNum* restrict n, size_t places_rdx)
768 {
769 	n->len += places_rdx;
770 	n->num -= places_rdx;
771 }
772 
773 /**
774  * Shifts the digits right within a number by no more than BC_BASE_DIGS - 1.
775  * This is the final step on shifting numbers with the shift operators. It
776  * depends on the caller to shift the limbs properly because all it does is
777  * ensure that the rdx point is realigned to be between limbs.
778  * @param n    The number to shift digits in.
779  * @param dig  The number of places to shift right.
780  */
781 static void
782 bc_num_shift(BcNum* restrict n, BcBigDig dig)
783 {
784 	size_t i, len = n->len;
785 	BcBigDig carry = 0, pow;
786 	BcDig* ptr = n->num;
787 
788 	assert(dig < BC_BASE_DIGS);
789 
790 	// Figure out the parameters for division.
791 	pow = bc_num_pow10[dig];
792 	dig = bc_num_pow10[BC_BASE_DIGS - dig];
793 
794 	// Run a series of divisions and mods with carries across the entire number
795 	// array. This effectively shifts everything over.
796 	for (i = len - 1; i < len; --i)
797 	{
798 		BcBigDig in, temp;
799 		in = ((BcBigDig) ptr[i]);
800 		temp = carry * dig;
801 		carry = in % pow;
802 		ptr[i] = ((BcDig) (in / pow)) + (BcDig) temp;
803 		assert(ptr[i] >= 0 && ptr[i] < BC_BASE_POW);
804 	}
805 
806 	assert(!carry);
807 }
808 
809 /**
810  * Shift a number left by a certain number of places. This is the workhorse of
811  * the left shift operator.
812  * @param n       The number to shift left.
813  * @param places  The amount of places to shift @a n left by.
814  */
815 static void
816 bc_num_shiftLeft(BcNum* restrict n, size_t places)
817 {
818 	BcBigDig dig;
819 	size_t places_rdx;
820 	bool shift;
821 
822 	if (!places) return;
823 
824 	// Make sure to grow the number if necessary.
825 	if (places > n->scale)
826 	{
827 		size_t size = bc_vm_growSize(BC_NUM_RDX(places - n->scale), n->len);
828 		if (size > SIZE_MAX - 1) bc_err(BC_ERR_MATH_OVERFLOW);
829 	}
830 
831 	// If zero, we can just set the scale and bail.
832 	if (BC_NUM_ZERO(n))
833 	{
834 		if (n->scale >= places) n->scale -= places;
835 		else n->scale = 0;
836 		return;
837 	}
838 
839 	// When I changed bc to have multiple digits per limb, this was the hardest
840 	// code to change. This and shift right. Make sure you understand this
841 	// before attempting anything.
842 
843 	// This is how many limbs we will shift.
844 	dig = (BcBigDig) (places % BC_BASE_DIGS);
845 	shift = (dig != 0);
846 
847 	// Convert places to a rdx value.
848 	places_rdx = BC_NUM_RDX(places);
849 
850 	// If the number is not an integer, we need special care. The reason an
851 	// integer doesn't is because left shift would only extend the integer,
852 	// whereas a non-integer might have its fractional part eliminated or only
853 	// partially eliminated.
854 	if (n->scale)
855 	{
856 		size_t nrdx = BC_NUM_RDX_VAL(n);
857 
858 		// If the number's rdx is bigger, that's the hard case.
859 		if (nrdx >= places_rdx)
860 		{
861 			size_t mod = n->scale % BC_BASE_DIGS, revdig;
862 
863 			// We want mod to be in the range [1, BC_BASE_DIGS], not
864 			// [0, BC_BASE_DIGS).
865 			mod = mod ? mod : BC_BASE_DIGS;
866 
867 			// We need to reverse dig to get the actual number of digits.
868 			revdig = dig ? BC_BASE_DIGS - dig : 0;
869 
870 			// If the two overflow BC_BASE_DIGS, we need to move an extra place.
871 			if (mod + revdig > BC_BASE_DIGS) places_rdx = 1;
872 			else places_rdx = 0;
873 		}
874 		else places_rdx -= nrdx;
875 	}
876 
877 	// If this is non-zero, we need an extra place, so expand, move, and set.
878 	if (places_rdx)
879 	{
880 		bc_num_expand(n, bc_vm_growSize(n->len, places_rdx));
881 		// NOLINTNEXTLINE
882 		memmove(n->num + places_rdx, n->num, BC_NUM_SIZE(n->len));
883 		// NOLINTNEXTLINE
884 		memset(n->num, 0, BC_NUM_SIZE(places_rdx));
885 		n->len += places_rdx;
886 	}
887 
888 	// Set the scale appropriately.
889 	if (places > n->scale)
890 	{
891 		n->scale = 0;
892 		BC_NUM_RDX_SET(n, 0);
893 	}
894 	else
895 	{
896 		n->scale -= places;
897 		BC_NUM_RDX_SET(n, BC_NUM_RDX(n->scale));
898 	}
899 
900 	// Finally, shift within limbs.
901 	if (shift) bc_num_shift(n, BC_BASE_DIGS - dig);
902 
903 	bc_num_clean(n);
904 }
905 
906 void
907 bc_num_shiftRight(BcNum* restrict n, size_t places)
908 {
909 	BcBigDig dig;
910 	size_t places_rdx, scale, scale_mod, int_len, expand;
911 	bool shift;
912 
913 	if (!places) return;
914 
915 	// If zero, we can just set the scale and bail.
916 	if (BC_NUM_ZERO(n))
917 	{
918 		n->scale += places;
919 		bc_num_expand(n, BC_NUM_RDX(n->scale));
920 		return;
921 	}
922 
923 	// Amount within a limb we have to shift by.
924 	dig = (BcBigDig) (places % BC_BASE_DIGS);
925 	shift = (dig != 0);
926 
927 	scale = n->scale;
928 
929 	// Figure out how the scale is affected.
930 	scale_mod = scale % BC_BASE_DIGS;
931 	scale_mod = scale_mod ? scale_mod : BC_BASE_DIGS;
932 
933 	// We need to know the int length and rdx for places.
934 	int_len = bc_num_int(n);
935 	places_rdx = BC_NUM_RDX(places);
936 
937 	// If we are going to shift past a limb boundary or not, set accordingly.
938 	if (scale_mod + dig > BC_BASE_DIGS)
939 	{
940 		expand = places_rdx - 1;
941 		places_rdx = 1;
942 	}
943 	else
944 	{
945 		expand = places_rdx;
946 		places_rdx = 0;
947 	}
948 
949 	// Clamp expanding.
950 	if (expand > int_len) expand -= int_len;
951 	else expand = 0;
952 
953 	// Extend, expand, and zero.
954 	bc_num_extend(n, places_rdx * BC_BASE_DIGS);
955 	bc_num_expand(n, bc_vm_growSize(expand, n->len));
956 	// NOLINTNEXTLINE
957 	memset(n->num + n->len, 0, BC_NUM_SIZE(expand));
958 
959 	// Set the fields.
960 	n->len += expand;
961 	n->scale = 0;
962 	BC_NUM_RDX_SET(n, 0);
963 
964 	// Finally, shift within limbs.
965 	if (shift) bc_num_shift(n, dig);
966 
967 	n->scale = scale + places;
968 	BC_NUM_RDX_SET(n, BC_NUM_RDX(n->scale));
969 
970 	bc_num_clean(n);
971 
972 	assert(BC_NUM_RDX_VAL(n) <= n->len && n->len <= n->cap);
973 	assert(BC_NUM_RDX_VAL(n) == BC_NUM_RDX(n->scale));
974 }
975 
976 /**
977  * Tests if a number is a integer with scale or not. Returns true if the number
978  * is not an integer. If it is, its integer shifted form is copied into the
979  * result parameter for use where only integers are allowed.
980  * @param n  The integer to test and shift.
981  * @param r  The number to store the shifted result into. This number should
982  *           *not* be allocated.
983  * @return   True if the number is a non-integer, false otherwise.
984  */
985 static bool
986 bc_num_nonInt(const BcNum* restrict n, BcNum* restrict r)
987 {
988 	bool zero;
989 	size_t i, rdx = BC_NUM_RDX_VAL(n);
990 
991 	if (!rdx)
992 	{
993 		// NOLINTNEXTLINE
994 		memcpy(r, n, sizeof(BcNum));
995 		return false;
996 	}
997 
998 	zero = true;
999 
1000 	for (i = 0; zero && i < rdx; ++i)
1001 	{
1002 		zero = (n->num[i] == 0);
1003 	}
1004 
1005 	if (BC_ERR(!zero)) return true;
1006 
1007 	bc_num_shiftRdx(n, r);
1008 
1009 	return false;
1010 }
1011 
1012 #if BC_ENABLE_EXTRA_MATH
1013 
1014 /**
1015  * Execute common code for an operater that needs an integer for the second
1016  * operand and return the integer operand as a BcBigDig.
1017  * @param a  The first operand.
1018  * @param b  The second operand.
1019  * @param c  The result operand.
1020  * @return   The second operand as a hardware integer.
1021  */
1022 static BcBigDig
1023 bc_num_intop(const BcNum* a, const BcNum* b, BcNum* restrict c)
1024 {
1025 	BcNum temp;
1026 
1027 #if BC_GCC
1028 	temp.len = 0;
1029 	temp.rdx = 0;
1030 	temp.num = NULL;
1031 #endif // BC_GCC
1032 
1033 	if (BC_ERR(bc_num_nonInt(b, &temp))) bc_err(BC_ERR_MATH_NON_INTEGER);
1034 
1035 	bc_num_copy(c, a);
1036 
1037 	return bc_num_bigdig(&temp);
1038 }
1039 #endif // BC_ENABLE_EXTRA_MATH
1040 
1041 /**
1042  * This is the actual implementation of add *and* subtract. Since this function
1043  * doesn't need to use scale (per the bc spec), I am hijacking it to say whether
1044  * it's doing an add or a subtract. And then I convert substraction to addition
1045  * of negative second operand. This is a BcNumBinOp function.
1046  * @param a    The first operand.
1047  * @param b    The second operand.
1048  * @param c    The return parameter.
1049  * @param sub  Non-zero for a subtract, zero for an add.
1050  */
1051 static void
1052 bc_num_as(BcNum* a, BcNum* b, BcNum* restrict c, size_t sub)
1053 {
1054 	BcDig* ptr_c;
1055 	BcDig* ptr_l;
1056 	BcDig* ptr_r;
1057 	size_t i, min_rdx, max_rdx, diff, a_int, b_int, min_len, max_len, max_int;
1058 	size_t len_l, len_r, ardx, brdx;
1059 	bool b_neg, do_sub, do_rev_sub, carry, c_neg;
1060 
1061 	if (BC_NUM_ZERO(b))
1062 	{
1063 		bc_num_copy(c, a);
1064 		return;
1065 	}
1066 
1067 	if (BC_NUM_ZERO(a))
1068 	{
1069 		bc_num_copy(c, b);
1070 		c->rdx = BC_NUM_NEG_VAL(c, BC_NUM_NEG(b) != sub);
1071 		return;
1072 	}
1073 
1074 	// Invert sign of b if it is to be subtracted. This operation must
1075 	// precede the tests for any of the operands being zero.
1076 	b_neg = (BC_NUM_NEG(b) != sub);
1077 
1078 	// Figure out if we will actually add the numbers if their signs are equal
1079 	// or subtract.
1080 	do_sub = (BC_NUM_NEG(a) != b_neg);
1081 
1082 	a_int = bc_num_int(a);
1083 	b_int = bc_num_int(b);
1084 	max_int = BC_MAX(a_int, b_int);
1085 
1086 	// Figure out which number will have its last limbs copied (for addition) or
1087 	// subtracted (for subtraction).
1088 	ardx = BC_NUM_RDX_VAL(a);
1089 	brdx = BC_NUM_RDX_VAL(b);
1090 	min_rdx = BC_MIN(ardx, brdx);
1091 	max_rdx = BC_MAX(ardx, brdx);
1092 	diff = max_rdx - min_rdx;
1093 
1094 	max_len = max_int + max_rdx;
1095 
1096 	if (do_sub)
1097 	{
1098 		// Check whether b has to be subtracted from a or a from b.
1099 		if (a_int != b_int) do_rev_sub = (a_int < b_int);
1100 		else if (ardx > brdx)
1101 		{
1102 			do_rev_sub = (bc_num_compare(a->num + diff, b->num, b->len) < 0);
1103 		}
1104 		else do_rev_sub = (bc_num_compare(a->num, b->num + diff, a->len) <= 0);
1105 	}
1106 	else
1107 	{
1108 		// The result array of the addition might come out one element
1109 		// longer than the bigger of the operand arrays.
1110 		max_len += 1;
1111 		do_rev_sub = (a_int < b_int);
1112 	}
1113 
1114 	assert(max_len <= c->cap);
1115 
1116 	// Cache values for simple code later.
1117 	if (do_rev_sub)
1118 	{
1119 		ptr_l = b->num;
1120 		ptr_r = a->num;
1121 		len_l = b->len;
1122 		len_r = a->len;
1123 	}
1124 	else
1125 	{
1126 		ptr_l = a->num;
1127 		ptr_r = b->num;
1128 		len_l = a->len;
1129 		len_r = b->len;
1130 	}
1131 
1132 	ptr_c = c->num;
1133 	carry = false;
1134 
1135 	// This is true if the numbers have a different number of limbs after the
1136 	// decimal point.
1137 	if (diff)
1138 	{
1139 		// If the rdx values of the operands do not match, the result will
1140 		// have low end elements that are the positive or negative trailing
1141 		// elements of the operand with higher rdx value.
1142 		if ((ardx > brdx) != do_rev_sub)
1143 		{
1144 			// !do_rev_sub && ardx > brdx || do_rev_sub && brdx > ardx
1145 			// The left operand has BcDig values that need to be copied,
1146 			// either from a or from b (in case of a reversed subtraction).
1147 			// NOLINTNEXTLINE
1148 			memcpy(ptr_c, ptr_l, BC_NUM_SIZE(diff));
1149 			ptr_l += diff;
1150 			len_l -= diff;
1151 		}
1152 		else
1153 		{
1154 			// The right operand has BcDig values that need to be copied
1155 			// or subtracted from zero (in case of a subtraction).
1156 			if (do_sub)
1157 			{
1158 				// do_sub (do_rev_sub && ardx > brdx ||
1159 				// !do_rev_sub && brdx > ardx)
1160 				for (i = 0; i < diff; i++)
1161 				{
1162 					ptr_c[i] = bc_num_subDigits(0, ptr_r[i], &carry);
1163 				}
1164 			}
1165 			else
1166 			{
1167 				// !do_sub && brdx > ardx
1168 				// NOLINTNEXTLINE
1169 				memcpy(ptr_c, ptr_r, BC_NUM_SIZE(diff));
1170 			}
1171 
1172 			// Future code needs to ignore the limbs we just did.
1173 			ptr_r += diff;
1174 			len_r -= diff;
1175 		}
1176 
1177 		// The return value pointer needs to ignore what we just did.
1178 		ptr_c += diff;
1179 	}
1180 
1181 	// This is the length that can be directly added/subtracted.
1182 	min_len = BC_MIN(len_l, len_r);
1183 
1184 	// After dealing with possible low array elements that depend on only one
1185 	// operand above, the actual add or subtract can be performed as if the rdx
1186 	// of both operands was the same.
1187 	//
1188 	// Inlining takes care of eliminating constant zero arguments to
1189 	// addDigit/subDigit (checked in disassembly of resulting bc binary
1190 	// compiled with gcc and clang).
1191 	if (do_sub)
1192 	{
1193 		// Actual subtraction.
1194 		for (i = 0; i < min_len; ++i)
1195 		{
1196 			ptr_c[i] = bc_num_subDigits(ptr_l[i], ptr_r[i], &carry);
1197 		}
1198 
1199 		// Finishing the limbs beyond the direct subtraction.
1200 		for (; i < len_l; ++i)
1201 		{
1202 			ptr_c[i] = bc_num_subDigits(ptr_l[i], 0, &carry);
1203 		}
1204 	}
1205 	else
1206 	{
1207 		// Actual addition.
1208 		for (i = 0; i < min_len; ++i)
1209 		{
1210 			ptr_c[i] = bc_num_addDigits(ptr_l[i], ptr_r[i], &carry);
1211 		}
1212 
1213 		// Finishing the limbs beyond the direct addition.
1214 		for (; i < len_l; ++i)
1215 		{
1216 			ptr_c[i] = bc_num_addDigits(ptr_l[i], 0, &carry);
1217 		}
1218 
1219 		// Addition can create an extra limb. We take care of that here.
1220 		ptr_c[i] = bc_num_addDigits(0, 0, &carry);
1221 	}
1222 
1223 	assert(carry == false);
1224 
1225 	// The result has the same sign as a, unless the operation was a
1226 	// reverse subtraction (b - a).
1227 	c_neg = BC_NUM_NEG(a) != (do_sub && do_rev_sub);
1228 	BC_NUM_RDX_SET_NEG(c, max_rdx, c_neg);
1229 	c->len = max_len;
1230 	c->scale = BC_MAX(a->scale, b->scale);
1231 
1232 	bc_num_clean(c);
1233 }
1234 
1235 /**
1236  * The simple multiplication that karatsuba dishes out to when the length of the
1237  * numbers gets low enough. This doesn't use scale because it treats the
1238  * operands as though they are integers.
1239  * @param a  The first operand.
1240  * @param b  The second operand.
1241  * @param c  The return parameter.
1242  */
1243 static void
1244 bc_num_m_simp(const BcNum* a, const BcNum* b, BcNum* restrict c)
1245 {
1246 	size_t i, alen = a->len, blen = b->len, clen;
1247 	BcDig* ptr_a = a->num;
1248 	BcDig* ptr_b = b->num;
1249 	BcDig* ptr_c;
1250 	BcBigDig sum = 0, carry = 0;
1251 
1252 	assert(sizeof(sum) >= sizeof(BcDig) * 2);
1253 	assert(!BC_NUM_RDX_VAL(a) && !BC_NUM_RDX_VAL(b));
1254 
1255 	// Make sure c is big enough.
1256 	clen = bc_vm_growSize(alen, blen);
1257 	bc_num_expand(c, bc_vm_growSize(clen, 1));
1258 
1259 	// If we don't memset, then we might have uninitialized data use later.
1260 	ptr_c = c->num;
1261 	// NOLINTNEXTLINE
1262 	memset(ptr_c, 0, BC_NUM_SIZE(c->cap));
1263 
1264 	// This is the actual multiplication loop. It uses the lattice form of long
1265 	// multiplication (see the explanation on the web page at
1266 	// https://knilt.arcc.albany.edu/What_is_Lattice_Multiplication or the
1267 	// explanation at Wikipedia).
1268 	for (i = 0; i < clen; ++i)
1269 	{
1270 		ssize_t sidx = (ssize_t) (i - blen + 1);
1271 		size_t j, k;
1272 
1273 		// These are the start indices.
1274 		j = (size_t) BC_MAX(0, sidx);
1275 		k = BC_MIN(i, blen - 1);
1276 
1277 		// On every iteration of this loop, a multiplication happens, then the
1278 		// sum is automatically calculated.
1279 		for (; j < alen && k < blen; ++j, --k)
1280 		{
1281 			sum += ((BcBigDig) ptr_a[j]) * ((BcBigDig) ptr_b[k]);
1282 
1283 			if (sum >= ((BcBigDig) BC_BASE_POW) * BC_BASE_POW)
1284 			{
1285 				carry += sum / BC_BASE_POW;
1286 				sum %= BC_BASE_POW;
1287 			}
1288 		}
1289 
1290 		// Calculate the carry.
1291 		if (sum >= BC_BASE_POW)
1292 		{
1293 			carry += sum / BC_BASE_POW;
1294 			sum %= BC_BASE_POW;
1295 		}
1296 
1297 		// Store and set up for next iteration.
1298 		ptr_c[i] = (BcDig) sum;
1299 		assert(ptr_c[i] < BC_BASE_POW);
1300 		sum = carry;
1301 		carry = 0;
1302 	}
1303 
1304 	// This should always be true because there should be no carry on the last
1305 	// digit; multiplication never goes above the sum of both lengths.
1306 	assert(!sum);
1307 
1308 	c->len = clen;
1309 }
1310 
1311 /**
1312  * Does a shifted add or subtract for Karatsuba below. This calls either
1313  * bc_num_addArrays() or bc_num_subArrays().
1314  * @param n      An in/out parameter; the first operand and return parameter.
1315  * @param a      The second operand.
1316  * @param shift  The amount to shift @a n by when adding/subtracting.
1317  * @param op     The function to call, either bc_num_addArrays() or
1318  *               bc_num_subArrays().
1319  */
1320 static void
1321 bc_num_shiftAddSub(BcNum* restrict n, const BcNum* restrict a, size_t shift,
1322                    BcNumShiftAddOp op)
1323 {
1324 	assert(n->len >= shift + a->len);
1325 	assert(!BC_NUM_RDX_VAL(n) && !BC_NUM_RDX_VAL(a));
1326 	op(n->num + shift, a->num, a->len);
1327 }
1328 
1329 /**
1330  * Implements the Karatsuba algorithm.
1331  */
1332 static void
1333 bc_num_k(const BcNum* a, const BcNum* b, BcNum* restrict c)
1334 {
1335 	size_t max, max2, total;
1336 	BcNum l1, h1, l2, h2, m2, m1, z0, z1, z2, temp;
1337 	BcDig* digs;
1338 	BcDig* dig_ptr;
1339 	BcNumShiftAddOp op;
1340 	bool aone = BC_NUM_ONE(a);
1341 #if BC_ENABLE_LIBRARY
1342 	BcVm* vm = bcl_getspecific();
1343 #endif // BC_ENABLE_LIBRARY
1344 
1345 	assert(BC_NUM_ZERO(c));
1346 
1347 	if (BC_NUM_ZERO(a) || BC_NUM_ZERO(b)) return;
1348 
1349 	if (aone || BC_NUM_ONE(b))
1350 	{
1351 		bc_num_copy(c, aone ? b : a);
1352 		if ((aone && BC_NUM_NEG(a)) || BC_NUM_NEG(b)) BC_NUM_NEG_TGL(c);
1353 		return;
1354 	}
1355 
1356 	// Shell out to the simple algorithm with certain conditions.
1357 	if (a->len < BC_NUM_KARATSUBA_LEN || b->len < BC_NUM_KARATSUBA_LEN)
1358 	{
1359 		bc_num_m_simp(a, b, c);
1360 		return;
1361 	}
1362 
1363 	// We need to calculate the max size of the numbers that can result from the
1364 	// operations.
1365 	max = BC_MAX(a->len, b->len);
1366 	max = BC_MAX(max, BC_NUM_DEF_SIZE);
1367 	max2 = (max + 1) / 2;
1368 
1369 	// Calculate the space needed for all of the temporary allocations. We do
1370 	// this to just allocate once.
1371 	total = bc_vm_arraySize(BC_NUM_KARATSUBA_ALLOCS, max);
1372 
1373 	BC_SIG_LOCK;
1374 
1375 	// Allocate space for all of the temporaries.
1376 	digs = dig_ptr = bc_vm_malloc(BC_NUM_SIZE(total));
1377 
1378 	// Set up the temporaries.
1379 	bc_num_setup(&l1, dig_ptr, max);
1380 	dig_ptr += max;
1381 	bc_num_setup(&h1, dig_ptr, max);
1382 	dig_ptr += max;
1383 	bc_num_setup(&l2, dig_ptr, max);
1384 	dig_ptr += max;
1385 	bc_num_setup(&h2, dig_ptr, max);
1386 	dig_ptr += max;
1387 	bc_num_setup(&m1, dig_ptr, max);
1388 	dig_ptr += max;
1389 	bc_num_setup(&m2, dig_ptr, max);
1390 
1391 	// Some temporaries need the ability to grow, so we allocate them
1392 	// separately.
1393 	max = bc_vm_growSize(max, 1);
1394 	bc_num_init(&z0, max);
1395 	bc_num_init(&z1, max);
1396 	bc_num_init(&z2, max);
1397 	max = bc_vm_growSize(max, max) + 1;
1398 	bc_num_init(&temp, max);
1399 
1400 	BC_SETJMP_LOCKED(vm, err);
1401 
1402 	BC_SIG_UNLOCK;
1403 
1404 	// First, set up c.
1405 	bc_num_expand(c, max);
1406 	c->len = max;
1407 	// NOLINTNEXTLINE
1408 	memset(c->num, 0, BC_NUM_SIZE(c->len));
1409 
1410 	// Split the parameters.
1411 	bc_num_split(a, max2, &l1, &h1);
1412 	bc_num_split(b, max2, &l2, &h2);
1413 
1414 	// Do the subtraction.
1415 	bc_num_sub(&h1, &l1, &m1, 0);
1416 	bc_num_sub(&l2, &h2, &m2, 0);
1417 
1418 	// The if statements below are there for efficiency reasons. The best way to
1419 	// understand them is to understand the Karatsuba algorithm because now that
1420 	// the ollocations and splits are done, the algorithm is pretty
1421 	// straightforward.
1422 
1423 	if (BC_NUM_NONZERO(&h1) && BC_NUM_NONZERO(&h2))
1424 	{
1425 		assert(BC_NUM_RDX_VALID_NP(h1));
1426 		assert(BC_NUM_RDX_VALID_NP(h2));
1427 
1428 		bc_num_m(&h1, &h2, &z2, 0);
1429 		bc_num_clean(&z2);
1430 
1431 		bc_num_shiftAddSub(c, &z2, max2 * 2, bc_num_addArrays);
1432 		bc_num_shiftAddSub(c, &z2, max2, bc_num_addArrays);
1433 	}
1434 
1435 	if (BC_NUM_NONZERO(&l1) && BC_NUM_NONZERO(&l2))
1436 	{
1437 		assert(BC_NUM_RDX_VALID_NP(l1));
1438 		assert(BC_NUM_RDX_VALID_NP(l2));
1439 
1440 		bc_num_m(&l1, &l2, &z0, 0);
1441 		bc_num_clean(&z0);
1442 
1443 		bc_num_shiftAddSub(c, &z0, max2, bc_num_addArrays);
1444 		bc_num_shiftAddSub(c, &z0, 0, bc_num_addArrays);
1445 	}
1446 
1447 	if (BC_NUM_NONZERO(&m1) && BC_NUM_NONZERO(&m2))
1448 	{
1449 		assert(BC_NUM_RDX_VALID_NP(m1));
1450 		assert(BC_NUM_RDX_VALID_NP(m1));
1451 
1452 		bc_num_m(&m1, &m2, &z1, 0);
1453 		bc_num_clean(&z1);
1454 
1455 		op = (BC_NUM_NEG_NP(m1) != BC_NUM_NEG_NP(m2)) ?
1456 		         bc_num_subArrays :
1457 		         bc_num_addArrays;
1458 		bc_num_shiftAddSub(c, &z1, max2, op);
1459 	}
1460 
1461 err:
1462 	BC_SIG_MAYLOCK;
1463 	free(digs);
1464 	bc_num_free(&temp);
1465 	bc_num_free(&z2);
1466 	bc_num_free(&z1);
1467 	bc_num_free(&z0);
1468 	BC_LONGJMP_CONT(vm);
1469 }
1470 
1471 /**
1472  * Does checks for Karatsuba. It also changes things to ensure that the
1473  * Karatsuba and simple multiplication can treat the numbers as integers. This
1474  * is a BcNumBinOp function.
1475  * @param a      The first operand.
1476  * @param b      The second operand.
1477  * @param c      The return parameter.
1478  * @param scale  The current scale.
1479  */
1480 static void
1481 bc_num_m(BcNum* a, BcNum* b, BcNum* restrict c, size_t scale)
1482 {
1483 	BcNum cpa, cpb;
1484 	size_t ascale, bscale, ardx, brdx, zero, len, rscale;
1485 	// These are meant to quiet warnings on GCC about longjmp() clobbering.
1486 	// The problem is real here.
1487 	size_t scale1, scale2, realscale;
1488 	// These are meant to quiet the GCC longjmp() clobbering, even though it
1489 	// does not apply here.
1490 	volatile size_t azero;
1491 	volatile size_t bzero;
1492 #if BC_ENABLE_LIBRARY
1493 	BcVm* vm = bcl_getspecific();
1494 #endif // BC_ENABLE_LIBRARY
1495 
1496 	assert(BC_NUM_RDX_VALID(a));
1497 	assert(BC_NUM_RDX_VALID(b));
1498 
1499 	bc_num_zero(c);
1500 
1501 	ascale = a->scale;
1502 	bscale = b->scale;
1503 
1504 	// This sets the final scale according to the bc spec.
1505 	scale1 = BC_MAX(scale, ascale);
1506 	scale2 = BC_MAX(scale1, bscale);
1507 	rscale = ascale + bscale;
1508 	realscale = BC_MIN(rscale, scale2);
1509 
1510 	// If this condition is true, we can use bc_num_mulArray(), which would be
1511 	// much faster.
1512 	if ((a->len == 1 || b->len == 1) && !a->rdx && !b->rdx)
1513 	{
1514 		BcNum* operand;
1515 		BcBigDig dig;
1516 
1517 		// Set the correct operands.
1518 		if (a->len == 1)
1519 		{
1520 			dig = (BcBigDig) a->num[0];
1521 			operand = b;
1522 		}
1523 		else
1524 		{
1525 			dig = (BcBigDig) b->num[0];
1526 			operand = a;
1527 		}
1528 
1529 		bc_num_mulArray(operand, dig, c);
1530 
1531 		// Need to make sure the sign is correct.
1532 		if (BC_NUM_NONZERO(c))
1533 		{
1534 			c->rdx = BC_NUM_NEG_VAL(c, BC_NUM_NEG(a) != BC_NUM_NEG(b));
1535 		}
1536 
1537 		return;
1538 	}
1539 
1540 	assert(BC_NUM_RDX_VALID(a));
1541 	assert(BC_NUM_RDX_VALID(b));
1542 
1543 	BC_SIG_LOCK;
1544 
1545 	// We need copies because of all of the mutation needed to make Karatsuba
1546 	// think the numbers are integers.
1547 	bc_num_init(&cpa, a->len + BC_NUM_RDX_VAL(a));
1548 	bc_num_init(&cpb, b->len + BC_NUM_RDX_VAL(b));
1549 
1550 	BC_SETJMP_LOCKED(vm, init_err);
1551 
1552 	BC_SIG_UNLOCK;
1553 
1554 	bc_num_copy(&cpa, a);
1555 	bc_num_copy(&cpb, b);
1556 
1557 	assert(BC_NUM_RDX_VALID_NP(cpa));
1558 	assert(BC_NUM_RDX_VALID_NP(cpb));
1559 
1560 	BC_NUM_NEG_CLR_NP(cpa);
1561 	BC_NUM_NEG_CLR_NP(cpb);
1562 
1563 	assert(BC_NUM_RDX_VALID_NP(cpa));
1564 	assert(BC_NUM_RDX_VALID_NP(cpb));
1565 
1566 	// These are what makes them appear like integers.
1567 	ardx = BC_NUM_RDX_VAL_NP(cpa) * BC_BASE_DIGS;
1568 	bc_num_shiftLeft(&cpa, ardx);
1569 
1570 	brdx = BC_NUM_RDX_VAL_NP(cpb) * BC_BASE_DIGS;
1571 	bc_num_shiftLeft(&cpb, brdx);
1572 
1573 	// We need to reset the jump here because azero and bzero are used in the
1574 	// cleanup, and local variables are not guaranteed to be the same after a
1575 	// jump.
1576 	BC_SIG_LOCK;
1577 
1578 	BC_UNSETJMP(vm);
1579 
1580 	// We want to ignore zero limbs.
1581 	azero = bc_num_shiftZero(&cpa);
1582 	bzero = bc_num_shiftZero(&cpb);
1583 
1584 	BC_SETJMP_LOCKED(vm, err);
1585 
1586 	BC_SIG_UNLOCK;
1587 
1588 	bc_num_clean(&cpa);
1589 	bc_num_clean(&cpb);
1590 
1591 	bc_num_k(&cpa, &cpb, c);
1592 
1593 	// The return parameter needs to have its scale set. This is the start. It
1594 	// also needs to be shifted by the same amount as a and b have limbs after
1595 	// the decimal point.
1596 	zero = bc_vm_growSize(azero, bzero);
1597 	len = bc_vm_growSize(c->len, zero);
1598 
1599 	bc_num_expand(c, len);
1600 
1601 	// Shift c based on the limbs after the decimal point in a and b.
1602 	bc_num_shiftLeft(c, (len - c->len) * BC_BASE_DIGS);
1603 	bc_num_shiftRight(c, ardx + brdx);
1604 
1605 	bc_num_retireMul(c, realscale, BC_NUM_NEG(a), BC_NUM_NEG(b));
1606 
1607 err:
1608 	BC_SIG_MAYLOCK;
1609 	bc_num_unshiftZero(&cpb, bzero);
1610 	bc_num_unshiftZero(&cpa, azero);
1611 init_err:
1612 	BC_SIG_MAYLOCK;
1613 	bc_num_free(&cpb);
1614 	bc_num_free(&cpa);
1615 	BC_LONGJMP_CONT(vm);
1616 }
1617 
1618 /**
1619  * Returns true if the BcDig array has non-zero limbs, false otherwise.
1620  * @param a    The array to test.
1621  * @param len  The length of the array.
1622  * @return     True if @a has any non-zero limbs, false otherwise.
1623  */
1624 static bool
1625 bc_num_nonZeroDig(BcDig* restrict a, size_t len)
1626 {
1627 	size_t i;
1628 
1629 	for (i = len - 1; i < len; --i)
1630 	{
1631 		if (a[i] != 0) return true;
1632 	}
1633 
1634 	return false;
1635 }
1636 
1637 /**
1638  * Compares a BcDig array against a BcNum. This is especially suited for
1639  * division. Returns >0 if @a a is greater than @a b, <0 if it is less, and =0
1640  * if they are equal.
1641  * @param a    The array.
1642  * @param b    The number.
1643  * @param len  The length to assume the arrays are. This is always less than the
1644  *             actual length because of how this is implemented.
1645  */
1646 static ssize_t
1647 bc_num_divCmp(const BcDig* a, const BcNum* b, size_t len)
1648 {
1649 	ssize_t cmp;
1650 
1651 	if (b->len > len && a[len]) cmp = bc_num_compare(a, b->num, len + 1);
1652 	else if (b->len <= len)
1653 	{
1654 		if (a[len]) cmp = 1;
1655 		else cmp = bc_num_compare(a, b->num, len);
1656 	}
1657 	else cmp = -1;
1658 
1659 	return cmp;
1660 }
1661 
1662 /**
1663  * Extends the two operands of a division by BC_BASE_DIGS minus the number of
1664  * digits in the divisor estimate. In other words, it is shifting the numbers in
1665  * order to force the divisor estimate to fill the limb.
1666  * @param a        The first operand.
1667  * @param b        The second operand.
1668  * @param divisor  The divisor estimate.
1669  */
1670 static void
1671 bc_num_divExtend(BcNum* restrict a, BcNum* restrict b, BcBigDig divisor)
1672 {
1673 	size_t pow;
1674 
1675 	assert(divisor < BC_BASE_POW);
1676 
1677 	pow = BC_BASE_DIGS - bc_num_log10((size_t) divisor);
1678 
1679 	bc_num_shiftLeft(a, pow);
1680 	bc_num_shiftLeft(b, pow);
1681 }
1682 
1683 /**
1684  * Actually does division. This is a rewrite of my original code by Stefan Esser
1685  * from FreeBSD.
1686  * @param a      The first operand.
1687  * @param b      The second operand.
1688  * @param c      The return parameter.
1689  * @param scale  The current scale.
1690  */
1691 static void
1692 bc_num_d_long(BcNum* restrict a, BcNum* restrict b, BcNum* restrict c,
1693               size_t scale)
1694 {
1695 	BcBigDig divisor;
1696 	size_t i, rdx;
1697 	// This is volatile and len 2 and reallen exist to quiet the GCC warning
1698 	// about clobbering on longjmp(). This one is possible, I think.
1699 	volatile size_t len;
1700 	size_t len2, reallen;
1701 	// This is volatile and realend exists to quiet the GCC warning about
1702 	// clobbering on longjmp(). This one is possible, I think.
1703 	volatile size_t end;
1704 	size_t realend;
1705 	BcNum cpb;
1706 	// This is volatile and realnonzero exists to quiet the GCC warning about
1707 	// clobbering on longjmp(). This one is possible, I think.
1708 	volatile bool nonzero;
1709 	bool realnonzero;
1710 #if BC_ENABLE_LIBRARY
1711 	BcVm* vm = bcl_getspecific();
1712 #endif // BC_ENABLE_LIBRARY
1713 
1714 	assert(b->len < a->len);
1715 
1716 	len = b->len;
1717 	end = a->len - len;
1718 
1719 	assert(len >= 1);
1720 
1721 	// This is a final time to make sure c is big enough and that its array is
1722 	// properly zeroed.
1723 	bc_num_expand(c, a->len);
1724 	// NOLINTNEXTLINE
1725 	memset(c->num, 0, c->cap * sizeof(BcDig));
1726 
1727 	// Setup.
1728 	BC_NUM_RDX_SET(c, BC_NUM_RDX_VAL(a));
1729 	c->scale = a->scale;
1730 	c->len = a->len;
1731 
1732 	// This is pulling the most significant limb of b in order to establish a
1733 	// good "estimate" for the actual divisor.
1734 	divisor = (BcBigDig) b->num[len - 1];
1735 
1736 	// The entire bit of code in this if statement is to tighten the estimate of
1737 	// the divisor. The condition asks if b has any other non-zero limbs.
1738 	if (len > 1 && bc_num_nonZeroDig(b->num, len - 1))
1739 	{
1740 		// This takes a little bit of understanding. The "10*BC_BASE_DIGS/6+1"
1741 		// results in either 16 for 64-bit 9-digit limbs or 7 for 32-bit 4-digit
1742 		// limbs. Then it shifts a 1 by that many, which in both cases, puts the
1743 		// result above *half* of the max value a limb can store. Basically,
1744 		// this quickly calculates if the divisor is greater than half the max
1745 		// of a limb.
1746 		nonzero = (divisor > 1 << ((10 * BC_BASE_DIGS) / 6 + 1));
1747 
1748 		// If the divisor is *not* greater than half the limb...
1749 		if (!nonzero)
1750 		{
1751 			// Extend the parameters by the number of missing digits in the
1752 			// divisor.
1753 			bc_num_divExtend(a, b, divisor);
1754 
1755 			// Check bc_num_d(). In there, we grow a again and again. We do it
1756 			// again here; we *always* want to be sure it is big enough.
1757 			len2 = BC_MAX(a->len, b->len);
1758 			bc_num_expand(a, len2 + 1);
1759 
1760 			// Make a have a zero most significant limb to match the len.
1761 			if (len2 + 1 > a->len) a->len = len2 + 1;
1762 
1763 			// Grab the new divisor estimate, new because the shift has made it
1764 			// different.
1765 			reallen = b->len;
1766 			realend = a->len - reallen;
1767 			divisor = (BcBigDig) b->num[reallen - 1];
1768 
1769 			realnonzero = bc_num_nonZeroDig(b->num, reallen - 1);
1770 		}
1771 		else
1772 		{
1773 			realend = end;
1774 			realnonzero = nonzero;
1775 		}
1776 	}
1777 	else
1778 	{
1779 		realend = end;
1780 		realnonzero = false;
1781 	}
1782 
1783 	// If b has other nonzero limbs, we want the divisor to be one higher, so
1784 	// that it is an upper bound.
1785 	divisor += realnonzero;
1786 
1787 	// Make sure c can fit the new length.
1788 	bc_num_expand(c, a->len);
1789 	// NOLINTNEXTLINE
1790 	memset(c->num, 0, BC_NUM_SIZE(c->cap));
1791 
1792 	assert(c->scale >= scale);
1793 	rdx = BC_NUM_RDX_VAL(c) - BC_NUM_RDX(scale);
1794 
1795 	BC_SIG_LOCK;
1796 
1797 	bc_num_init(&cpb, len + 1);
1798 
1799 	BC_SETJMP_LOCKED(vm, err);
1800 
1801 	BC_SIG_UNLOCK;
1802 
1803 	// This is the actual division loop.
1804 	for (i = realend - 1; i < realend && i >= rdx && BC_NUM_NONZERO(a); --i)
1805 	{
1806 		ssize_t cmp;
1807 		BcDig* n;
1808 		BcBigDig result;
1809 
1810 		n = a->num + i;
1811 		assert(n >= a->num);
1812 		result = 0;
1813 
1814 		cmp = bc_num_divCmp(n, b, len);
1815 
1816 		// This is true if n is greater than b, which means that division can
1817 		// proceed, so this inner loop is the part that implements one instance
1818 		// of the division.
1819 		while (cmp >= 0)
1820 		{
1821 			BcBigDig n1, dividend, quotient;
1822 
1823 			// These should be named obviously enough. Just imagine that it's a
1824 			// division of one limb. Because that's what it is.
1825 			n1 = (BcBigDig) n[len];
1826 			dividend = n1 * BC_BASE_POW + (BcBigDig) n[len - 1];
1827 			quotient = (dividend / divisor);
1828 
1829 			// If this is true, then we can just subtract. Remember: setting
1830 			// quotient to 1 is not bad because we already know that n is
1831 			// greater than b.
1832 			if (quotient <= 1)
1833 			{
1834 				quotient = 1;
1835 				bc_num_subArrays(n, b->num, len);
1836 			}
1837 			else
1838 			{
1839 				assert(quotient <= BC_BASE_POW);
1840 
1841 				// We need to multiply and subtract for a quotient above 1.
1842 				bc_num_mulArray(b, (BcBigDig) quotient, &cpb);
1843 				bc_num_subArrays(n, cpb.num, cpb.len);
1844 			}
1845 
1846 			// The result is the *real* quotient, by the way, but it might take
1847 			// multiple trips around this loop to get it.
1848 			result += quotient;
1849 			assert(result <= BC_BASE_POW);
1850 
1851 			// And here's why it might take multiple trips: n might *still* be
1852 			// greater than b. So we have to loop again. That's what this is
1853 			// setting up for: the condition of the while loop.
1854 			if (realnonzero) cmp = bc_num_divCmp(n, b, len);
1855 			else cmp = -1;
1856 		}
1857 
1858 		assert(result < BC_BASE_POW);
1859 
1860 		// Store the actual limb quotient.
1861 		c->num[i] = (BcDig) result;
1862 	}
1863 
1864 err:
1865 	BC_SIG_MAYLOCK;
1866 	bc_num_free(&cpb);
1867 	BC_LONGJMP_CONT(vm);
1868 }
1869 
1870 /**
1871  * Implements division. This is a BcNumBinOp function.
1872  * @param a      The first operand.
1873  * @param b      The second operand.
1874  * @param c      The return parameter.
1875  * @param scale  The current scale.
1876  */
1877 static void
1878 bc_num_d(BcNum* a, BcNum* b, BcNum* restrict c, size_t scale)
1879 {
1880 	size_t len, cpardx;
1881 	BcNum cpa, cpb;
1882 #if BC_ENABLE_LIBRARY
1883 	BcVm* vm = bcl_getspecific();
1884 #endif // BC_ENABLE_LIBRARY
1885 
1886 	if (BC_NUM_ZERO(b)) bc_err(BC_ERR_MATH_DIVIDE_BY_ZERO);
1887 
1888 	if (BC_NUM_ZERO(a))
1889 	{
1890 		bc_num_setToZero(c, scale);
1891 		return;
1892 	}
1893 
1894 	if (BC_NUM_ONE(b))
1895 	{
1896 		bc_num_copy(c, a);
1897 		bc_num_retireMul(c, scale, BC_NUM_NEG(a), BC_NUM_NEG(b));
1898 		return;
1899 	}
1900 
1901 	// If this is true, we can use bc_num_divArray(), which would be faster.
1902 	if (!BC_NUM_RDX_VAL(a) && !BC_NUM_RDX_VAL(b) && b->len == 1 && !scale)
1903 	{
1904 		BcBigDig rem;
1905 		bc_num_divArray(a, (BcBigDig) b->num[0], c, &rem);
1906 		bc_num_retireMul(c, scale, BC_NUM_NEG(a), BC_NUM_NEG(b));
1907 		return;
1908 	}
1909 
1910 	len = bc_num_divReq(a, b, scale);
1911 
1912 	BC_SIG_LOCK;
1913 
1914 	// Initialize copies of the parameters. We want the length of the first
1915 	// operand copy to be as big as the result because of the way the division
1916 	// is implemented.
1917 	bc_num_init(&cpa, len);
1918 	bc_num_copy(&cpa, a);
1919 	bc_num_createCopy(&cpb, b);
1920 
1921 	BC_SETJMP_LOCKED(vm, err);
1922 
1923 	BC_SIG_UNLOCK;
1924 
1925 	len = b->len;
1926 
1927 	// Like the above comment, we want the copy of the first parameter to be
1928 	// larger than the second parameter.
1929 	if (len > cpa.len)
1930 	{
1931 		bc_num_expand(&cpa, bc_vm_growSize(len, 2));
1932 		bc_num_extend(&cpa, (len - cpa.len) * BC_BASE_DIGS);
1933 	}
1934 
1935 	cpardx = BC_NUM_RDX_VAL_NP(cpa);
1936 	cpa.scale = cpardx * BC_BASE_DIGS;
1937 
1938 	// This is just setting up the scale in preparation for the division.
1939 	bc_num_extend(&cpa, b->scale);
1940 	cpardx = BC_NUM_RDX_VAL_NP(cpa) - BC_NUM_RDX(b->scale);
1941 	BC_NUM_RDX_SET_NP(cpa, cpardx);
1942 	cpa.scale = cpardx * BC_BASE_DIGS;
1943 
1944 	// Once again, just setting things up, this time to match scale.
1945 	if (scale > cpa.scale)
1946 	{
1947 		bc_num_extend(&cpa, scale);
1948 		cpardx = BC_NUM_RDX_VAL_NP(cpa);
1949 		cpa.scale = cpardx * BC_BASE_DIGS;
1950 	}
1951 
1952 	// Grow if necessary.
1953 	if (cpa.cap == cpa.len) bc_num_expand(&cpa, bc_vm_growSize(cpa.len, 1));
1954 
1955 	// We want an extra zero in front to make things simpler.
1956 	cpa.num[cpa.len++] = 0;
1957 
1958 	// Still setting things up. Why all of these things are needed is not
1959 	// something that can be easily explained, but it has to do with making the
1960 	// actual algorithm easier to understand because it can assume a lot of
1961 	// things. Thus, you should view all of this setup code as establishing
1962 	// assumptions for bc_num_d_long(), where the actual division happens.
1963 	//
1964 	// But in short, this setup makes it so bc_num_d_long() can pretend the
1965 	// numbers are integers.
1966 	if (cpardx == cpa.len) cpa.len = bc_num_nonZeroLen(&cpa);
1967 	if (BC_NUM_RDX_VAL_NP(cpb) == cpb.len) cpb.len = bc_num_nonZeroLen(&cpb);
1968 	cpb.scale = 0;
1969 	BC_NUM_RDX_SET_NP(cpb, 0);
1970 
1971 	bc_num_d_long(&cpa, &cpb, c, scale);
1972 
1973 	bc_num_retireMul(c, scale, BC_NUM_NEG(a), BC_NUM_NEG(b));
1974 
1975 err:
1976 	BC_SIG_MAYLOCK;
1977 	bc_num_free(&cpb);
1978 	bc_num_free(&cpa);
1979 	BC_LONGJMP_CONT(vm);
1980 }
1981 
1982 /**
1983  * Implements divmod. This is the actual modulus function; since modulus
1984  * requires a division anyway, this returns the quotient and modulus. Either can
1985  * be thrown out as desired.
1986  * @param a      The first operand.
1987  * @param b      The second operand.
1988  * @param c      The return parameter for the quotient.
1989  * @param d      The return parameter for the modulus.
1990  * @param scale  The current scale.
1991  * @param ts     The scale that the operation should be done to. Yes, it's not
1992  *               necessarily the same as scale, per the bc spec.
1993  */
1994 static void
1995 bc_num_r(BcNum* a, BcNum* b, BcNum* restrict c, BcNum* restrict d, size_t scale,
1996          size_t ts)
1997 {
1998 	BcNum temp;
1999 	// realscale is meant to quiet a warning on GCC about longjmp() clobbering.
2000 	// This one is real.
2001 	size_t realscale;
2002 	bool neg;
2003 #if BC_ENABLE_LIBRARY
2004 	BcVm* vm = bcl_getspecific();
2005 #endif // BC_ENABLE_LIBRARY
2006 
2007 	if (BC_NUM_ZERO(b)) bc_err(BC_ERR_MATH_DIVIDE_BY_ZERO);
2008 
2009 	if (BC_NUM_ZERO(a))
2010 	{
2011 		bc_num_setToZero(c, ts);
2012 		bc_num_setToZero(d, ts);
2013 		return;
2014 	}
2015 
2016 	BC_SIG_LOCK;
2017 
2018 	bc_num_init(&temp, d->cap);
2019 
2020 	BC_SETJMP_LOCKED(vm, err);
2021 
2022 	BC_SIG_UNLOCK;
2023 
2024 	// Division.
2025 	bc_num_d(a, b, c, scale);
2026 
2027 	// We want an extra digit so we can safely truncate.
2028 	if (scale) realscale = ts + 1;
2029 	else realscale = scale;
2030 
2031 	assert(BC_NUM_RDX_VALID(c));
2032 	assert(BC_NUM_RDX_VALID(b));
2033 
2034 	// Implement the rest of the (a - (a / b) * b) formula.
2035 	bc_num_m(c, b, &temp, realscale);
2036 	bc_num_sub(a, &temp, d, realscale);
2037 
2038 	// Extend if necessary.
2039 	if (ts > d->scale && BC_NUM_NONZERO(d)) bc_num_extend(d, ts - d->scale);
2040 
2041 	neg = BC_NUM_NEG(d);
2042 	bc_num_retireMul(d, ts, BC_NUM_NEG(a), BC_NUM_NEG(b));
2043 	d->rdx = BC_NUM_NEG_VAL(d, BC_NUM_NONZERO(d) ? neg : false);
2044 
2045 err:
2046 	BC_SIG_MAYLOCK;
2047 	bc_num_free(&temp);
2048 	BC_LONGJMP_CONT(vm);
2049 }
2050 
2051 /**
2052  * Implements modulus/remainder. (Yes, I know they are different, but not in the
2053  * context of bc.) This is a BcNumBinOp function.
2054  * @param a      The first operand.
2055  * @param b      The second operand.
2056  * @param c      The return parameter.
2057  * @param scale  The current scale.
2058  */
2059 static void
2060 bc_num_rem(BcNum* a, BcNum* b, BcNum* restrict c, size_t scale)
2061 {
2062 	BcNum c1;
2063 	size_t ts;
2064 #if BC_ENABLE_LIBRARY
2065 	BcVm* vm = bcl_getspecific();
2066 #endif // BC_ENABLE_LIBRARY
2067 
2068 	ts = bc_vm_growSize(scale, b->scale);
2069 	ts = BC_MAX(ts, a->scale);
2070 
2071 	BC_SIG_LOCK;
2072 
2073 	// Need a temp for the quotient.
2074 	bc_num_init(&c1, bc_num_mulReq(a, b, ts));
2075 
2076 	BC_SETJMP_LOCKED(vm, err);
2077 
2078 	BC_SIG_UNLOCK;
2079 
2080 	bc_num_r(a, b, &c1, c, scale, ts);
2081 
2082 err:
2083 	BC_SIG_MAYLOCK;
2084 	bc_num_free(&c1);
2085 	BC_LONGJMP_CONT(vm);
2086 }
2087 
2088 /**
2089  * Implements power (exponentiation). This is a BcNumBinOp function.
2090  * @param a      The first operand.
2091  * @param b      The second operand.
2092  * @param c      The return parameter.
2093  * @param scale  The current scale.
2094  */
2095 static void
2096 bc_num_p(BcNum* a, BcNum* b, BcNum* restrict c, size_t scale)
2097 {
2098 	BcNum copy, btemp;
2099 	BcBigDig exp;
2100 	// realscale is meant to quiet a warning on GCC about longjmp() clobbering.
2101 	// This one is real.
2102 	size_t powrdx, resrdx, realscale;
2103 	bool neg;
2104 #if BC_ENABLE_LIBRARY
2105 	BcVm* vm = bcl_getspecific();
2106 #endif // BC_ENABLE_LIBRARY
2107 
2108 	// This is here to silence a warning from GCC.
2109 #if BC_GCC
2110 	btemp.len = 0;
2111 	btemp.rdx = 0;
2112 	btemp.num = NULL;
2113 #endif // BC_GCC
2114 
2115 	if (BC_ERR(bc_num_nonInt(b, &btemp))) bc_err(BC_ERR_MATH_NON_INTEGER);
2116 
2117 	assert(btemp.len == 0 || btemp.num != NULL);
2118 
2119 	if (BC_NUM_ZERO(&btemp))
2120 	{
2121 		bc_num_one(c);
2122 		return;
2123 	}
2124 
2125 	if (BC_NUM_ZERO(a))
2126 	{
2127 		if (BC_NUM_NEG_NP(btemp)) bc_err(BC_ERR_MATH_DIVIDE_BY_ZERO);
2128 		bc_num_setToZero(c, scale);
2129 		return;
2130 	}
2131 
2132 	if (BC_NUM_ONE(&btemp))
2133 	{
2134 		if (!BC_NUM_NEG_NP(btemp)) bc_num_copy(c, a);
2135 		else bc_num_inv(a, c, scale);
2136 		return;
2137 	}
2138 
2139 	neg = BC_NUM_NEG_NP(btemp);
2140 	BC_NUM_NEG_CLR_NP(btemp);
2141 
2142 	exp = bc_num_bigdig(&btemp);
2143 
2144 	BC_SIG_LOCK;
2145 
2146 	bc_num_createCopy(&copy, a);
2147 
2148 	BC_SETJMP_LOCKED(vm, err);
2149 
2150 	BC_SIG_UNLOCK;
2151 
2152 	// If this is true, then we do not have to do a division, and we need to
2153 	// set scale accordingly.
2154 	if (!neg)
2155 	{
2156 		size_t max = BC_MAX(scale, a->scale), scalepow;
2157 		scalepow = bc_num_mulOverflow(a->scale, exp);
2158 		realscale = BC_MIN(scalepow, max);
2159 	}
2160 	else realscale = scale;
2161 
2162 	// This is only implementing the first exponentiation by squaring, until it
2163 	// reaches the first time where the square is actually used.
2164 	for (powrdx = a->scale; !(exp & 1); exp >>= 1)
2165 	{
2166 		powrdx <<= 1;
2167 		assert(BC_NUM_RDX_VALID_NP(copy));
2168 		bc_num_mul(&copy, &copy, &copy, powrdx);
2169 	}
2170 
2171 	// Make c a copy of copy for the purpose of saving the squares that should
2172 	// be saved.
2173 	bc_num_copy(c, &copy);
2174 	resrdx = powrdx;
2175 
2176 	// Now finish the exponentiation by squaring, this time saving the squares
2177 	// as necessary.
2178 	while (exp >>= 1)
2179 	{
2180 		powrdx <<= 1;
2181 		assert(BC_NUM_RDX_VALID_NP(copy));
2182 		bc_num_mul(&copy, &copy, &copy, powrdx);
2183 
2184 		// If this is true, we want to save that particular square. This does
2185 		// that by multiplying c with copy.
2186 		if (exp & 1)
2187 		{
2188 			resrdx += powrdx;
2189 			assert(BC_NUM_RDX_VALID(c));
2190 			assert(BC_NUM_RDX_VALID_NP(copy));
2191 			bc_num_mul(c, &copy, c, resrdx);
2192 		}
2193 	}
2194 
2195 	// Invert if necessary.
2196 	if (neg) bc_num_inv(c, c, realscale);
2197 
2198 	// Truncate if necessary.
2199 	if (c->scale > realscale) bc_num_truncate(c, c->scale - realscale);
2200 
2201 	bc_num_clean(c);
2202 
2203 err:
2204 	BC_SIG_MAYLOCK;
2205 	bc_num_free(&copy);
2206 	BC_LONGJMP_CONT(vm);
2207 }
2208 
2209 #if BC_ENABLE_EXTRA_MATH
2210 /**
2211  * Implements the places operator. This is a BcNumBinOp function.
2212  * @param a      The first operand.
2213  * @param b      The second operand.
2214  * @param c      The return parameter.
2215  * @param scale  The current scale.
2216  */
2217 static void
2218 bc_num_place(BcNum* a, BcNum* b, BcNum* restrict c, size_t scale)
2219 {
2220 	BcBigDig val;
2221 
2222 	BC_UNUSED(scale);
2223 
2224 	val = bc_num_intop(a, b, c);
2225 
2226 	// Just truncate or extend as appropriate.
2227 	if (val < c->scale) bc_num_truncate(c, c->scale - val);
2228 	else if (val > c->scale) bc_num_extend(c, val - c->scale);
2229 }
2230 
2231 /**
2232  * Implements the left shift operator. This is a BcNumBinOp function.
2233  */
2234 static void
2235 bc_num_left(BcNum* a, BcNum* b, BcNum* restrict c, size_t scale)
2236 {
2237 	BcBigDig val;
2238 
2239 	BC_UNUSED(scale);
2240 
2241 	val = bc_num_intop(a, b, c);
2242 
2243 	bc_num_shiftLeft(c, (size_t) val);
2244 }
2245 
2246 /**
2247  * Implements the right shift operator. This is a BcNumBinOp function.
2248  */
2249 static void
2250 bc_num_right(BcNum* a, BcNum* b, BcNum* restrict c, size_t scale)
2251 {
2252 	BcBigDig val;
2253 
2254 	BC_UNUSED(scale);
2255 
2256 	val = bc_num_intop(a, b, c);
2257 
2258 	if (BC_NUM_ZERO(c)) return;
2259 
2260 	bc_num_shiftRight(c, (size_t) val);
2261 }
2262 #endif // BC_ENABLE_EXTRA_MATH
2263 
2264 /**
2265  * Prepares for, and calls, a binary operator function. This is probably the
2266  * most important function in the entire file because it establishes assumptions
2267  * that make the rest of the code so easy. Those assumptions include:
2268  *
2269  * - a is not the same pointer as c.
2270  * - b is not the same pointer as c.
2271  * - there is enough room in c for the result.
2272  *
2273  * Without these, this whole function would basically have to be duplicated for
2274  * *all* binary operators.
2275  *
2276  * @param a      The first operand.
2277  * @param b      The second operand.
2278  * @param c      The return parameter.
2279  * @param scale  The current scale.
2280  * @param req    The number of limbs needed to fit the result.
2281  */
2282 static void
2283 bc_num_binary(BcNum* a, BcNum* b, BcNum* c, size_t scale, BcNumBinOp op,
2284               size_t req)
2285 {
2286 	BcNum* ptr_a;
2287 	BcNum* ptr_b;
2288 	BcNum num2;
2289 #if BC_ENABLE_LIBRARY
2290 	BcVm* vm = NULL;
2291 #endif // BC_ENABLE_LIBRARY
2292 
2293 	assert(a != NULL && b != NULL && c != NULL && op != NULL);
2294 
2295 	assert(BC_NUM_RDX_VALID(a));
2296 	assert(BC_NUM_RDX_VALID(b));
2297 
2298 	BC_SIG_LOCK;
2299 
2300 	ptr_a = c == a ? &num2 : a;
2301 	ptr_b = c == b ? &num2 : b;
2302 
2303 	// Actually reallocate. If we don't reallocate, we want to expand at the
2304 	// very least.
2305 	if (c == a || c == b)
2306 	{
2307 #if BC_ENABLE_LIBRARY
2308 		vm = bcl_getspecific();
2309 #endif // BC_ENABLE_LIBRARY
2310 
2311 		// NOLINTNEXTLINE
2312 		memcpy(&num2, c, sizeof(BcNum));
2313 
2314 		bc_num_init(c, req);
2315 
2316 		// Must prepare for cleanup. We want this here so that locals that got
2317 		// set stay set since a longjmp() is not guaranteed to preserve locals.
2318 		BC_SETJMP_LOCKED(vm, err);
2319 		BC_SIG_UNLOCK;
2320 	}
2321 	else
2322 	{
2323 		BC_SIG_UNLOCK;
2324 		bc_num_expand(c, req);
2325 	}
2326 
2327 	// It is okay for a and b to be the same. If a binary operator function does
2328 	// need them to be different, the binary operator function is responsible
2329 	// for that.
2330 
2331 	// Call the actual binary operator function.
2332 	op(ptr_a, ptr_b, c, scale);
2333 
2334 	assert(!BC_NUM_NEG(c) || BC_NUM_NONZERO(c));
2335 	assert(BC_NUM_RDX_VAL(c) <= c->len || !c->len);
2336 	assert(BC_NUM_RDX_VALID(c));
2337 	assert(!c->len || c->num[c->len - 1] || BC_NUM_RDX_VAL(c) == c->len);
2338 
2339 err:
2340 	// Cleanup only needed if we initialized c to a new number.
2341 	if (c == a || c == b)
2342 	{
2343 		BC_SIG_MAYLOCK;
2344 		bc_num_free(&num2);
2345 		BC_LONGJMP_CONT(vm);
2346 	}
2347 }
2348 
2349 /**
2350  * Tests a number string for validity. This function has a history; I originally
2351  * wrote it because I did not trust my parser. Over time, however, I came to
2352  * trust it, so I was able to relegate this function to debug builds only, and I
2353  * used it in assert()'s. But then I created the library, and well, I can't
2354  * trust users, so I reused this for yelling at users.
2355  * @param val  The string to check to see if it's a valid number string.
2356  * @return     True if the string is a valid number string, false otherwise.
2357  */
2358 bool
2359 bc_num_strValid(const char* restrict val)
2360 {
2361 	bool radix = false;
2362 	size_t i, len = strlen(val);
2363 
2364 	// Notice that I don't check if there is a negative sign. That is not part
2365 	// of a valid number, except in the library. The library-specific code takes
2366 	// care of that part.
2367 
2368 	// Nothing in the string is okay.
2369 	if (!len) return true;
2370 
2371 	// Loop through the characters.
2372 	for (i = 0; i < len; ++i)
2373 	{
2374 		BcDig c = val[i];
2375 
2376 		// If we have found a radix point...
2377 		if (c == '.')
2378 		{
2379 			// We don't allow two radices.
2380 			if (radix) return false;
2381 
2382 			radix = true;
2383 			continue;
2384 		}
2385 
2386 		// We only allow digits and uppercase letters.
2387 		if (!(isdigit(c) || isupper(c))) return false;
2388 	}
2389 
2390 	return true;
2391 }
2392 
2393 /**
2394  * Parses one character and returns the digit that corresponds to that
2395  * character according to the base.
2396  * @param c     The character to parse.
2397  * @param base  The base.
2398  * @return      The character as a digit.
2399  */
2400 static BcBigDig
2401 bc_num_parseChar(char c, size_t base)
2402 {
2403 	assert(isupper(c) || isdigit(c));
2404 
2405 	// If a letter...
2406 	if (isupper(c))
2407 	{
2408 #if BC_ENABLE_LIBRARY
2409 		BcVm* vm = bcl_getspecific();
2410 #endif // BC_ENABLE_LIBRARY
2411 
2412 		// This returns the digit that directly corresponds with the letter.
2413 		c = BC_NUM_NUM_LETTER(c);
2414 
2415 		// If the digit is greater than the base, we clamp.
2416 		if (BC_DIGIT_CLAMP)
2417 		{
2418 			c = ((size_t) c) >= base ? (char) base - 1 : c;
2419 		}
2420 	}
2421 	// Straight convert the digit to a number.
2422 	else c -= '0';
2423 
2424 	return (BcBigDig) (uchar) c;
2425 }
2426 
2427 /**
2428  * Parses a string as a decimal number. This is separate because it's going to
2429  * be the most used, and it can be heavily optimized for decimal only.
2430  * @param n    The number to parse into and return. Must be preallocated.
2431  * @param val  The string to parse.
2432  */
2433 static void
2434 bc_num_parseDecimal(BcNum* restrict n, const char* restrict val)
2435 {
2436 	size_t len, i, temp, mod;
2437 	const char* ptr;
2438 	bool zero = true, rdx;
2439 #if BC_ENABLE_LIBRARY
2440 	BcVm* vm = bcl_getspecific();
2441 #endif // BC_ENABLE_LIBRARY
2442 
2443 	// Eat leading zeroes.
2444 	for (i = 0; val[i] == '0'; ++i)
2445 	{
2446 		continue;
2447 	}
2448 
2449 	val += i;
2450 	assert(!val[0] || isalnum(val[0]) || val[0] == '.');
2451 
2452 	// All 0's. We can just return, since this procedure expects a virgin
2453 	// (already 0) BcNum.
2454 	if (!val[0]) return;
2455 
2456 	// The length of the string is the length of the number, except it might be
2457 	// one bigger because of a decimal point.
2458 	len = strlen(val);
2459 
2460 	// Find the location of the decimal point.
2461 	ptr = strchr(val, '.');
2462 	rdx = (ptr != NULL);
2463 
2464 	// We eat leading zeroes again. These leading zeroes are different because
2465 	// they will come after the decimal point if they exist, and since that's
2466 	// the case, they must be preserved.
2467 	for (i = 0; i < len && (zero = (val[i] == '0' || val[i] == '.')); ++i)
2468 	{
2469 		continue;
2470 	}
2471 
2472 	// Set the scale of the number based on the location of the decimal point.
2473 	// The casts to uintptr_t is to ensure that bc does not hit undefined
2474 	// behavior when doing math on the values.
2475 	n->scale = (size_t) (rdx *
2476 	                     (((uintptr_t) (val + len)) - (((uintptr_t) ptr) + 1)));
2477 
2478 	// Set rdx.
2479 	BC_NUM_RDX_SET(n, BC_NUM_RDX(n->scale));
2480 
2481 	// Calculate length. First, the length of the integer, then the number of
2482 	// digits in the last limb, then the length.
2483 	i = len - (ptr == val ? 0 : i) - rdx;
2484 	temp = BC_NUM_ROUND_POW(i);
2485 	mod = n->scale % BC_BASE_DIGS;
2486 	i = mod ? BC_BASE_DIGS - mod : 0;
2487 	n->len = ((temp + i) / BC_BASE_DIGS);
2488 
2489 	// Expand and zero. The plus extra is in case the lack of clamping causes
2490 	// the number to overflow the original bounds.
2491 	bc_num_expand(n, n->len + !BC_DIGIT_CLAMP);
2492 	// NOLINTNEXTLINE
2493 	memset(n->num, 0, BC_NUM_SIZE(n->len + !BC_DIGIT_CLAMP));
2494 
2495 	if (zero)
2496 	{
2497 		// I think I can set rdx directly to zero here because n should be a
2498 		// new number with sign set to false.
2499 		n->len = n->rdx = 0;
2500 	}
2501 	else
2502 	{
2503 		// There is actually stuff to parse if we make it here. Yay...
2504 		BcBigDig exp, pow;
2505 
2506 		assert(i <= BC_NUM_BIGDIG_MAX);
2507 
2508 		// The exponent and power.
2509 		exp = (BcBigDig) i;
2510 		pow = bc_num_pow10[exp];
2511 
2512 		// Parse loop. We parse backwards because numbers are stored little
2513 		// endian.
2514 		for (i = len - 1; i < len; --i, ++exp)
2515 		{
2516 			char c = val[i];
2517 
2518 			// Skip the decimal point.
2519 			if (c == '.') exp -= 1;
2520 			else
2521 			{
2522 				// The index of the limb.
2523 				size_t idx = exp / BC_BASE_DIGS;
2524 				BcBigDig dig;
2525 
2526 				if (isupper(c))
2527 				{
2528 					// Clamp for the base.
2529 					if (!BC_DIGIT_CLAMP) c = BC_NUM_NUM_LETTER(c);
2530 					else c = 9;
2531 				}
2532 				else c -= '0';
2533 
2534 				// Add the digit to the limb. This takes care of overflow from
2535 				// lack of clamping.
2536 				dig = ((BcBigDig) n->num[idx]) + ((BcBigDig) c) * pow;
2537 				if (dig >= BC_BASE_POW)
2538 				{
2539 					// We cannot go over BC_BASE_POW with clamping.
2540 					assert(!BC_DIGIT_CLAMP);
2541 
2542 					n->num[idx + 1] = (BcDig) (dig / BC_BASE_POW);
2543 					n->num[idx] = (BcDig) (dig % BC_BASE_POW);
2544 					assert(n->num[idx] >= 0 && n->num[idx] < BC_BASE_POW);
2545 					assert(n->num[idx + 1] >= 0 &&
2546 					       n->num[idx + 1] < BC_BASE_POW);
2547 				}
2548 				else
2549 				{
2550 					n->num[idx] = (BcDig) dig;
2551 					assert(n->num[idx] >= 0 && n->num[idx] < BC_BASE_POW);
2552 				}
2553 
2554 				// Adjust the power and exponent.
2555 				if ((exp + 1) % BC_BASE_DIGS == 0) pow = 1;
2556 				else pow *= BC_BASE;
2557 			}
2558 		}
2559 	}
2560 
2561 	// Make sure to add one to the length if needed from lack of clamping.
2562 	n->len += (!BC_DIGIT_CLAMP && n->num[n->len] != 0);
2563 }
2564 
2565 /**
2566  * Parse a number in any base (besides decimal).
2567  * @param n     The number to parse into and return. Must be preallocated.
2568  * @param val   The string to parse.
2569  * @param base  The base to parse as.
2570  */
2571 static void
2572 bc_num_parseBase(BcNum* restrict n, const char* restrict val, BcBigDig base)
2573 {
2574 	BcNum temp, mult1, mult2, result1, result2;
2575 	BcNum* m1;
2576 	BcNum* m2;
2577 	BcNum* ptr;
2578 	char c = 0;
2579 	bool zero = true;
2580 	BcBigDig v;
2581 	size_t digs, len = strlen(val);
2582 	// This is volatile to quiet a warning on GCC about longjmp() clobbering.
2583 	volatile size_t i;
2584 #if BC_ENABLE_LIBRARY
2585 	BcVm* vm = bcl_getspecific();
2586 #endif // BC_ENABLE_LIBRARY
2587 
2588 	// If zero, just return because the number should be virgin (already 0).
2589 	for (i = 0; zero && i < len; ++i)
2590 	{
2591 		zero = (val[i] == '.' || val[i] == '0');
2592 	}
2593 	if (zero) return;
2594 
2595 	BC_SIG_LOCK;
2596 
2597 	bc_num_init(&temp, BC_NUM_BIGDIG_LOG10);
2598 	bc_num_init(&mult1, BC_NUM_BIGDIG_LOG10);
2599 
2600 	BC_SETJMP_LOCKED(vm, int_err);
2601 
2602 	BC_SIG_UNLOCK;
2603 
2604 	// We split parsing into parsing the integer and parsing the fractional
2605 	// part.
2606 
2607 	// Parse the integer part. This is the easy part because we just multiply
2608 	// the number by the base, then add the digit.
2609 	for (i = 0; i < len && (c = val[i]) && c != '.'; ++i)
2610 	{
2611 		// Convert the character to a digit.
2612 		v = bc_num_parseChar(c, base);
2613 
2614 		// Multiply the number.
2615 		bc_num_mulArray(n, base, &mult1);
2616 
2617 		// Convert the digit to a number and add.
2618 		bc_num_bigdig2num(&temp, v);
2619 		bc_num_add(&mult1, &temp, n, 0);
2620 	}
2621 
2622 	// If this condition is true, then we are done. We still need to do cleanup
2623 	// though.
2624 	if (i == len && !val[i]) goto int_err;
2625 
2626 	// If we get here, we *must* be at the radix point.
2627 	assert(val[i] == '.');
2628 
2629 	BC_SIG_LOCK;
2630 
2631 	// Unset the jump to reset in for these new initializations.
2632 	BC_UNSETJMP(vm);
2633 
2634 	bc_num_init(&mult2, BC_NUM_BIGDIG_LOG10);
2635 	bc_num_init(&result1, BC_NUM_DEF_SIZE);
2636 	bc_num_init(&result2, BC_NUM_DEF_SIZE);
2637 	bc_num_one(&mult1);
2638 
2639 	BC_SETJMP_LOCKED(vm, err);
2640 
2641 	BC_SIG_UNLOCK;
2642 
2643 	// Pointers for easy switching.
2644 	m1 = &mult1;
2645 	m2 = &mult2;
2646 
2647 	// Parse the fractional part. This is the hard part.
2648 	for (i += 1, digs = 0; i < len && (c = val[i]); ++i, ++digs)
2649 	{
2650 		size_t rdx;
2651 
2652 		// Convert the character to a digit.
2653 		v = bc_num_parseChar(c, base);
2654 
2655 		// We keep growing result2 according to the base because the more digits
2656 		// after the radix, the more significant the digits close to the radix
2657 		// should be.
2658 		bc_num_mulArray(&result1, base, &result2);
2659 
2660 		// Convert the digit to a number.
2661 		bc_num_bigdig2num(&temp, v);
2662 
2663 		// Add the digit into the fraction part.
2664 		bc_num_add(&result2, &temp, &result1, 0);
2665 
2666 		// Keep growing m1 and m2 for use after the loop.
2667 		bc_num_mulArray(m1, base, m2);
2668 
2669 		rdx = BC_NUM_RDX_VAL(m2);
2670 
2671 		if (m2->len < rdx) m2->len = rdx;
2672 
2673 		// Switch.
2674 		ptr = m1;
2675 		m1 = m2;
2676 		m2 = ptr;
2677 	}
2678 
2679 	// This one cannot be a divide by 0 because mult starts out at 1, then is
2680 	// multiplied by base, and base cannot be 0, so mult cannot be 0. And this
2681 	// is the reason we keep growing m1 and m2; this division is what converts
2682 	// the parsed fractional part from an integer to a fractional part.
2683 	bc_num_div(&result1, m1, &result2, digs * 2);
2684 
2685 	// Pretruncate.
2686 	bc_num_truncate(&result2, digs);
2687 
2688 	// The final add of the integer part to the fractional part.
2689 	bc_num_add(n, &result2, n, digs);
2690 
2691 	// Basic cleanup.
2692 	if (BC_NUM_NONZERO(n))
2693 	{
2694 		if (n->scale < digs) bc_num_extend(n, digs - n->scale);
2695 	}
2696 	else bc_num_zero(n);
2697 
2698 err:
2699 	BC_SIG_MAYLOCK;
2700 	bc_num_free(&result2);
2701 	bc_num_free(&result1);
2702 	bc_num_free(&mult2);
2703 int_err:
2704 	BC_SIG_MAYLOCK;
2705 	bc_num_free(&mult1);
2706 	bc_num_free(&temp);
2707 	BC_LONGJMP_CONT(vm);
2708 }
2709 
2710 /**
2711  * Prints a backslash+newline combo if the number of characters needs it. This
2712  * is really a convenience function.
2713  */
2714 static inline void
2715 bc_num_printNewline(void)
2716 {
2717 #if !BC_ENABLE_LIBRARY
2718 	if (vm->nchars >= vm->line_len - 1 && vm->line_len)
2719 	{
2720 		bc_vm_putchar('\\', bc_flush_none);
2721 		bc_vm_putchar('\n', bc_flush_err);
2722 	}
2723 #endif // !BC_ENABLE_LIBRARY
2724 }
2725 
2726 /**
2727  * Prints a character after a backslash+newline, if needed.
2728  * @param c       The character to print.
2729  * @param bslash  Whether to print a backslash+newline.
2730  */
2731 static void
2732 bc_num_putchar(int c, bool bslash)
2733 {
2734 	if (c != '\n' && bslash) bc_num_printNewline();
2735 	bc_vm_putchar(c, bc_flush_save);
2736 }
2737 
2738 #if !BC_ENABLE_LIBRARY
2739 
2740 /**
2741  * Prints a character for a number's digit. This is for printing for dc's P
2742  * command. This function does not need to worry about radix points. This is a
2743  * BcNumDigitOp.
2744  * @param n       The "digit" to print.
2745  * @param len     The "length" of the digit, or number of characters that will
2746  *                need to be printed for the digit.
2747  * @param rdx     True if a decimal (radix) point should be printed.
2748  * @param bslash  True if a backslash+newline should be printed if the character
2749  *                limit for the line is reached, false otherwise.
2750  */
2751 static void
2752 bc_num_printChar(size_t n, size_t len, bool rdx, bool bslash)
2753 {
2754 	BC_UNUSED(rdx);
2755 	BC_UNUSED(len);
2756 	BC_UNUSED(bslash);
2757 	assert(len == 1);
2758 	bc_vm_putchar((uchar) n, bc_flush_save);
2759 }
2760 
2761 #endif // !BC_ENABLE_LIBRARY
2762 
2763 /**
2764  * Prints a series of characters for large bases. This is for printing in bases
2765  * above hexadecimal. This is a BcNumDigitOp.
2766  * @param n       The "digit" to print.
2767  * @param len     The "length" of the digit, or number of characters that will
2768  *                need to be printed for the digit.
2769  * @param rdx     True if a decimal (radix) point should be printed.
2770  * @param bslash  True if a backslash+newline should be printed if the character
2771  *                limit for the line is reached, false otherwise.
2772  */
2773 static void
2774 bc_num_printDigits(size_t n, size_t len, bool rdx, bool bslash)
2775 {
2776 	size_t exp, pow;
2777 
2778 	// If needed, print the radix; otherwise, print a space to separate digits.
2779 	bc_num_putchar(rdx ? '.' : ' ', true);
2780 
2781 	// Calculate the exponent and power.
2782 	for (exp = 0, pow = 1; exp < len - 1; ++exp, pow *= BC_BASE)
2783 	{
2784 		continue;
2785 	}
2786 
2787 	// Print each character individually.
2788 	for (exp = 0; exp < len; pow /= BC_BASE, ++exp)
2789 	{
2790 		// The individual subdigit.
2791 		size_t dig = n / pow;
2792 
2793 		// Take the subdigit away.
2794 		n -= dig * pow;
2795 
2796 		// Print the subdigit.
2797 		bc_num_putchar(((uchar) dig) + '0', bslash || exp != len - 1);
2798 	}
2799 }
2800 
2801 /**
2802  * Prints a character for a number's digit. This is for printing in bases for
2803  * hexadecimal and below because they always print only one character at a time.
2804  * This is a BcNumDigitOp.
2805  * @param n       The "digit" to print.
2806  * @param len     The "length" of the digit, or number of characters that will
2807  *                need to be printed for the digit.
2808  * @param rdx     True if a decimal (radix) point should be printed.
2809  * @param bslash  True if a backslash+newline should be printed if the character
2810  *                limit for the line is reached, false otherwise.
2811  */
2812 static void
2813 bc_num_printHex(size_t n, size_t len, bool rdx, bool bslash)
2814 {
2815 	BC_UNUSED(len);
2816 	BC_UNUSED(bslash);
2817 
2818 	assert(len == 1);
2819 
2820 	if (rdx) bc_num_putchar('.', true);
2821 
2822 	bc_num_putchar(bc_num_hex_digits[n], bslash);
2823 }
2824 
2825 /**
2826  * Prints a decimal number. This is specially written for optimization since
2827  * this will be used the most and because bc's numbers are already in decimal.
2828  * @param n        The number to print.
2829  * @param newline  Whether to print backslash+newlines on long enough lines.
2830  */
2831 static void
2832 bc_num_printDecimal(const BcNum* restrict n, bool newline)
2833 {
2834 	size_t i, j, rdx = BC_NUM_RDX_VAL(n);
2835 	bool zero = true;
2836 	size_t buffer[BC_BASE_DIGS];
2837 
2838 	// Print loop.
2839 	for (i = n->len - 1; i < n->len; --i)
2840 	{
2841 		BcDig n9 = n->num[i];
2842 		size_t temp;
2843 		bool irdx = (i == rdx - 1);
2844 
2845 		// Calculate the number of digits in the limb.
2846 		zero = (zero & !irdx);
2847 		temp = n->scale % BC_BASE_DIGS;
2848 		temp = i || !temp ? 0 : BC_BASE_DIGS - temp;
2849 
2850 		// NOLINTNEXTLINE
2851 		memset(buffer, 0, BC_BASE_DIGS * sizeof(size_t));
2852 
2853 		// Fill the buffer with individual digits.
2854 		for (j = 0; n9 && j < BC_BASE_DIGS; ++j)
2855 		{
2856 			buffer[j] = ((size_t) n9) % BC_BASE;
2857 			n9 /= BC_BASE;
2858 		}
2859 
2860 		// Print the digits in the buffer.
2861 		for (j = BC_BASE_DIGS - 1; j < BC_BASE_DIGS && j >= temp; --j)
2862 		{
2863 			// Figure out whether to print the decimal point.
2864 			bool print_rdx = (irdx & (j == BC_BASE_DIGS - 1));
2865 
2866 			// The zero variable helps us skip leading zero digits in the limb.
2867 			zero = (zero && buffer[j] == 0);
2868 
2869 			if (!zero)
2870 			{
2871 				// While the first three arguments should be self-explanatory,
2872 				// the last needs explaining. I don't want to print a newline
2873 				// when the last digit to be printed could take the place of the
2874 				// backslash rather than being pushed, as a single character, to
2875 				// the next line. That's what that last argument does for bc.
2876 				bc_num_printHex(buffer[j], 1, print_rdx,
2877 				                !newline || (j > temp || i != 0));
2878 			}
2879 		}
2880 	}
2881 }
2882 
2883 #if BC_ENABLE_EXTRA_MATH
2884 
2885 /**
2886  * Prints a number in scientific or engineering format. When doing this, we are
2887  * always printing in decimal.
2888  * @param n        The number to print.
2889  * @param eng      True if we are in engineering mode.
2890  * @param newline  Whether to print backslash+newlines on long enough lines.
2891  */
2892 static void
2893 bc_num_printExponent(const BcNum* restrict n, bool eng, bool newline)
2894 {
2895 	size_t places, mod, nrdx = BC_NUM_RDX_VAL(n);
2896 	bool neg = (n->len <= nrdx);
2897 	BcNum temp, exp;
2898 	BcDig digs[BC_NUM_BIGDIG_LOG10];
2899 #if BC_ENABLE_LIBRARY
2900 	BcVm* vm = bcl_getspecific();
2901 #endif // BC_ENABLE_LIBRARY
2902 
2903 	BC_SIG_LOCK;
2904 
2905 	bc_num_createCopy(&temp, n);
2906 
2907 	BC_SETJMP_LOCKED(vm, exit);
2908 
2909 	BC_SIG_UNLOCK;
2910 
2911 	// We need to calculate the exponents, and they change based on whether the
2912 	// number is all fractional or not, obviously.
2913 	if (neg)
2914 	{
2915 		// Figure out the negative power of 10.
2916 		places = bc_num_negPow10(n);
2917 
2918 		// Figure out how many digits mod 3 there are (important for
2919 		// engineering mode).
2920 		mod = places % 3;
2921 
2922 		// Calculate places if we are in engineering mode.
2923 		if (eng && mod != 0) places += 3 - mod;
2924 
2925 		// Shift the temp to the right place.
2926 		bc_num_shiftLeft(&temp, places);
2927 	}
2928 	else
2929 	{
2930 		// This is the number of digits that we are supposed to put behind the
2931 		// decimal point.
2932 		places = bc_num_intDigits(n) - 1;
2933 
2934 		// Calculate the true number based on whether engineering mode is
2935 		// activated.
2936 		mod = places % 3;
2937 		if (eng && mod != 0) places -= 3 - (3 - mod);
2938 
2939 		// Shift the temp to the right place.
2940 		bc_num_shiftRight(&temp, places);
2941 	}
2942 
2943 	// Print the shifted number.
2944 	bc_num_printDecimal(&temp, newline);
2945 
2946 	// Print the e.
2947 	bc_num_putchar('e', !newline);
2948 
2949 	// Need to explicitly print a zero exponent.
2950 	if (!places)
2951 	{
2952 		bc_num_printHex(0, 1, false, !newline);
2953 		goto exit;
2954 	}
2955 
2956 	// Need to print sign for the exponent.
2957 	if (neg) bc_num_putchar('-', true);
2958 
2959 	// Create a temporary for the exponent...
2960 	bc_num_setup(&exp, digs, BC_NUM_BIGDIG_LOG10);
2961 	bc_num_bigdig2num(&exp, (BcBigDig) places);
2962 
2963 	/// ..and print it.
2964 	bc_num_printDecimal(&exp, newline);
2965 
2966 exit:
2967 	BC_SIG_MAYLOCK;
2968 	bc_num_free(&temp);
2969 	BC_LONGJMP_CONT(vm);
2970 }
2971 #endif // BC_ENABLE_EXTRA_MATH
2972 
2973 /**
2974  * Takes a number with limbs with base BC_BASE_POW and converts the limb at the
2975  * given index to base @a pow, where @a pow is obase^N.
2976  * @param n    The number to convert.
2977  * @param rem  BC_BASE_POW - @a pow.
2978  * @param pow  The power of obase we will convert the number to.
2979  * @param idx  The index of the number to start converting at. Doing the
2980  *             conversion is O(n^2); we have to sweep through starting at the
2981  *             least significant limb.
2982  */
2983 static void
2984 bc_num_printFixup(BcNum* restrict n, BcBigDig rem, BcBigDig pow, size_t idx)
2985 {
2986 	size_t i, len = n->len - idx;
2987 	BcBigDig acc;
2988 	BcDig* a = n->num + idx;
2989 
2990 	// Ignore if there's just one limb left. This is the part that requires the
2991 	// extra loop after the one calling this function in bc_num_printPrepare().
2992 	if (len < 2) return;
2993 
2994 	// Loop through the remaining limbs and convert. We start at the second limb
2995 	// because we pull the value from the previous one as well.
2996 	for (i = len - 1; i > 0; --i)
2997 	{
2998 		// Get the limb and add it to the previous, along with multiplying by
2999 		// the remainder because that's the proper overflow. "acc" means
3000 		// "accumulator," by the way.
3001 		acc = ((BcBigDig) a[i]) * rem + ((BcBigDig) a[i - 1]);
3002 
3003 		// Store a value in base pow in the previous limb.
3004 		a[i - 1] = (BcDig) (acc % pow);
3005 
3006 		// Divide by the base and accumulate the remaining value in the limb.
3007 		acc /= pow;
3008 		acc += (BcBigDig) a[i];
3009 
3010 		// If the accumulator is greater than the base...
3011 		if (acc >= BC_BASE_POW)
3012 		{
3013 			// Do we need to grow?
3014 			if (i == len - 1)
3015 			{
3016 				// Grow.
3017 				len = bc_vm_growSize(len, 1);
3018 				bc_num_expand(n, bc_vm_growSize(len, idx));
3019 
3020 				// Update the pointer because it may have moved.
3021 				a = n->num + idx;
3022 
3023 				// Zero out the last limb.
3024 				a[len - 1] = 0;
3025 			}
3026 
3027 			// Overflow into the next limb since we are over the base.
3028 			a[i + 1] += acc / BC_BASE_POW;
3029 			acc %= BC_BASE_POW;
3030 		}
3031 
3032 		assert(acc < BC_BASE_POW);
3033 
3034 		// Set the limb.
3035 		a[i] = (BcDig) acc;
3036 	}
3037 
3038 	// We may have grown the number, so adjust the length.
3039 	n->len = len + idx;
3040 }
3041 
3042 /**
3043  * Prepares a number for printing in a base that does not have BC_BASE_POW as a
3044  * power. This basically converts the number from having limbs of base
3045  * BC_BASE_POW to limbs of pow, where pow is obase^N.
3046  * @param n    The number to prepare for printing.
3047  * @param rem  The remainder of BC_BASE_POW when divided by a power of the base.
3048  * @param pow  The power of the base.
3049  */
3050 static void
3051 bc_num_printPrepare(BcNum* restrict n, BcBigDig rem, BcBigDig pow)
3052 {
3053 	size_t i;
3054 
3055 	// Loop from the least significant limb to the most significant limb and
3056 	// convert limbs in each pass.
3057 	for (i = 0; i < n->len; ++i)
3058 	{
3059 		bc_num_printFixup(n, rem, pow, i);
3060 	}
3061 
3062 	// bc_num_printFixup() does not do everything it is supposed to, so we do
3063 	// the last bit of cleanup here. That cleanup is to ensure that each limb
3064 	// is less than pow and to expand the number to fit new limbs as necessary.
3065 	for (i = 0; i < n->len; ++i)
3066 	{
3067 		assert(pow == ((BcBigDig) ((BcDig) pow)));
3068 
3069 		// If the limb needs fixing...
3070 		if (n->num[i] >= (BcDig) pow)
3071 		{
3072 			// Do we need to grow?
3073 			if (i + 1 == n->len)
3074 			{
3075 				// Grow the number.
3076 				n->len = bc_vm_growSize(n->len, 1);
3077 				bc_num_expand(n, n->len);
3078 
3079 				// Without this, we might use uninitialized data.
3080 				n->num[i + 1] = 0;
3081 			}
3082 
3083 			assert(pow < BC_BASE_POW);
3084 
3085 			// Overflow into the next limb.
3086 			n->num[i + 1] += n->num[i] / ((BcDig) pow);
3087 			n->num[i] %= (BcDig) pow;
3088 		}
3089 	}
3090 }
3091 
3092 static void
3093 bc_num_printNum(BcNum* restrict n, BcBigDig base, size_t len,
3094                 BcNumDigitOp print, bool newline)
3095 {
3096 	BcVec stack;
3097 	BcNum intp, fracp1, fracp2, digit, flen1, flen2;
3098 	BcNum* n1;
3099 	BcNum* n2;
3100 	BcNum* temp;
3101 	BcBigDig dig = 0, acc, exp;
3102 	BcBigDig* ptr;
3103 	size_t i, j, nrdx, idigits;
3104 	bool radix;
3105 	BcDig digit_digs[BC_NUM_BIGDIG_LOG10 + 1];
3106 #if BC_ENABLE_LIBRARY
3107 	BcVm* vm = bcl_getspecific();
3108 #endif // BC_ENABLE_LIBRARY
3109 
3110 	assert(base > 1);
3111 
3112 	// Easy case. Even with scale, we just print this.
3113 	if (BC_NUM_ZERO(n))
3114 	{
3115 		print(0, len, false, !newline);
3116 		return;
3117 	}
3118 
3119 	// This function uses an algorithm that Stefan Esser <se@freebsd.org> came
3120 	// up with to print the integer part of a number. What it does is convert
3121 	// intp into a number of the specified base, but it does it directly,
3122 	// instead of just doing a series of divisions and printing the remainders
3123 	// in reverse order.
3124 	//
3125 	// Let me explain in a bit more detail:
3126 	//
3127 	// The algorithm takes the current least significant limb (after intp has
3128 	// been converted to an integer) and the next to least significant limb, and
3129 	// it converts the least significant limb into one of the specified base,
3130 	// putting any overflow into the next to least significant limb. It iterates
3131 	// through the whole number, from least significant to most significant,
3132 	// doing this conversion. At the end of that iteration, the least
3133 	// significant limb is converted, but the others are not, so it iterates
3134 	// again, starting at the next to least significant limb. It keeps doing
3135 	// that conversion, skipping one more limb than the last time, until all
3136 	// limbs have been converted. Then it prints them in reverse order.
3137 	//
3138 	// That is the gist of the algorithm. It leaves out several things, such as
3139 	// the fact that limbs are not always converted into the specified base, but
3140 	// into something close, basically a power of the specified base. In
3141 	// Stefan's words, "You could consider BcDigs to be of base 10^BC_BASE_DIGS
3142 	// in the normal case and obase^N for the largest value of N that satisfies
3143 	// obase^N <= 10^BC_BASE_DIGS. [This means that] the result is not in base
3144 	// "obase", but in base "obase^N", which happens to be printable as a number
3145 	// of base "obase" without consideration for neighbouring BcDigs." This fact
3146 	// is what necessitates the existence of the loop later in this function.
3147 	//
3148 	// The conversion happens in bc_num_printPrepare() where the outer loop
3149 	// happens and bc_num_printFixup() where the inner loop, or actual
3150 	// conversion, happens. In other words, bc_num_printPrepare() is where the
3151 	// loop that starts at the least significant limb and goes to the most
3152 	// significant limb. Then, on every iteration of its loop, it calls
3153 	// bc_num_printFixup(), which has the inner loop of actually converting
3154 	// the limbs it passes into limbs of base obase^N rather than base
3155 	// BC_BASE_POW.
3156 
3157 	nrdx = BC_NUM_RDX_VAL(n);
3158 
3159 	BC_SIG_LOCK;
3160 
3161 	// The stack is what allows us to reverse the digits for printing.
3162 	bc_vec_init(&stack, sizeof(BcBigDig), BC_DTOR_NONE);
3163 	bc_num_init(&fracp1, nrdx);
3164 
3165 	// intp will be the "integer part" of the number, so copy it.
3166 	bc_num_createCopy(&intp, n);
3167 
3168 	BC_SETJMP_LOCKED(vm, err);
3169 
3170 	BC_SIG_UNLOCK;
3171 
3172 	// Make intp an integer.
3173 	bc_num_truncate(&intp, intp.scale);
3174 
3175 	// Get the fractional part out.
3176 	bc_num_sub(n, &intp, &fracp1, 0);
3177 
3178 	// If the base is not the same as the last base used for printing, we need
3179 	// to update the cached exponent and power. Yes, we cache the values of the
3180 	// exponent and power. That is to prevent us from calculating them every
3181 	// time because printing will probably happen multiple times on the same
3182 	// base.
3183 	if (base != vm->last_base)
3184 	{
3185 		vm->last_pow = 1;
3186 		vm->last_exp = 0;
3187 
3188 		// Calculate the exponent and power.
3189 		while (vm->last_pow * base <= BC_BASE_POW)
3190 		{
3191 			vm->last_pow *= base;
3192 			vm->last_exp += 1;
3193 		}
3194 
3195 		// Also, the remainder and base itself.
3196 		vm->last_rem = BC_BASE_POW - vm->last_pow;
3197 		vm->last_base = base;
3198 	}
3199 
3200 	exp = vm->last_exp;
3201 
3202 	// If vm->last_rem is 0, then the base we are printing in is a divisor of
3203 	// BC_BASE_POW, which is the easy case because it means that BC_BASE_POW is
3204 	// a power of obase, and no conversion is needed. If it *is* 0, then we have
3205 	// the hard case, and we have to prepare the number for the base.
3206 	if (vm->last_rem != 0)
3207 	{
3208 		bc_num_printPrepare(&intp, vm->last_rem, vm->last_pow);
3209 	}
3210 
3211 	// After the conversion comes the surprisingly easy part. From here on out,
3212 	// this is basically naive code that I wrote, adjusted for the larger bases.
3213 
3214 	// Fill the stack of digits for the integer part.
3215 	for (i = 0; i < intp.len; ++i)
3216 	{
3217 		// Get the limb.
3218 		acc = (BcBigDig) intp.num[i];
3219 
3220 		// Turn the limb into digits of base obase.
3221 		for (j = 0; j < exp && (i < intp.len - 1 || acc != 0); ++j)
3222 		{
3223 			// This condition is true if we are not at the last digit.
3224 			if (j != exp - 1)
3225 			{
3226 				dig = acc % base;
3227 				acc /= base;
3228 			}
3229 			else
3230 			{
3231 				dig = acc;
3232 				acc = 0;
3233 			}
3234 
3235 			assert(dig < base);
3236 
3237 			// Push the digit onto the stack.
3238 			bc_vec_push(&stack, &dig);
3239 		}
3240 
3241 		assert(acc == 0);
3242 	}
3243 
3244 	// Go through the stack backwards and print each digit.
3245 	for (i = 0; i < stack.len; ++i)
3246 	{
3247 		ptr = bc_vec_item_rev(&stack, i);
3248 
3249 		assert(ptr != NULL);
3250 
3251 		// While the first three arguments should be self-explanatory, the last
3252 		// needs explaining. I don't want to print a backslash+newline when the
3253 		// last digit to be printed could take the place of the backslash rather
3254 		// than being pushed, as a single character, to the next line. That's
3255 		// what that last argument does for bc.
3256 		//
3257 		// First, it needs to check if newlines are completely disabled. If they
3258 		// are not disabled, it needs to check the next part.
3259 		//
3260 		// If the number has a scale, then because we are printing just the
3261 		// integer part, there will be at least two more characters (a radix
3262 		// point plus at least one digit). So if there is a scale, a backslash
3263 		// is necessary.
3264 		//
3265 		// Finally, the last condition checks to see if we are at the end of the
3266 		// stack. If we are *not* (i.e., the index is not one less than the
3267 		// stack length), then a backslash is necessary because there is at
3268 		// least one more character for at least one more digit). Otherwise, if
3269 		// the index is equal to one less than the stack length, we want to
3270 		// disable backslash printing.
3271 		//
3272 		// The function that prints bases 17 and above will take care of not
3273 		// printing a backslash in the right case.
3274 		print(*ptr, len, false,
3275 		      !newline || (n->scale != 0 || i < stack.len - 1));
3276 	}
3277 
3278 	// We are done if there is no fractional part.
3279 	if (!n->scale) goto err;
3280 
3281 	BC_SIG_LOCK;
3282 
3283 	// Reset the jump because some locals are changing.
3284 	BC_UNSETJMP(vm);
3285 
3286 	bc_num_init(&fracp2, nrdx);
3287 	bc_num_setup(&digit, digit_digs, sizeof(digit_digs) / sizeof(BcDig));
3288 	bc_num_init(&flen1, BC_NUM_BIGDIG_LOG10);
3289 	bc_num_init(&flen2, BC_NUM_BIGDIG_LOG10);
3290 
3291 	BC_SETJMP_LOCKED(vm, frac_err);
3292 
3293 	BC_SIG_UNLOCK;
3294 
3295 	bc_num_one(&flen1);
3296 
3297 	radix = true;
3298 
3299 	// Pointers for easy switching.
3300 	n1 = &flen1;
3301 	n2 = &flen2;
3302 
3303 	fracp2.scale = n->scale;
3304 	BC_NUM_RDX_SET_NP(fracp2, BC_NUM_RDX(fracp2.scale));
3305 
3306 	// As long as we have not reached the scale of the number, keep printing.
3307 	while ((idigits = bc_num_intDigits(n1)) <= n->scale)
3308 	{
3309 		// These numbers will keep growing.
3310 		bc_num_expand(&fracp2, fracp1.len + 1);
3311 		bc_num_mulArray(&fracp1, base, &fracp2);
3312 
3313 		nrdx = BC_NUM_RDX_VAL_NP(fracp2);
3314 
3315 		// Ensure an invariant.
3316 		if (fracp2.len < nrdx) fracp2.len = nrdx;
3317 
3318 		// fracp is guaranteed to be non-negative and small enough.
3319 		dig = bc_num_bigdig2(&fracp2);
3320 
3321 		// Convert the digit to a number and subtract it from the number.
3322 		bc_num_bigdig2num(&digit, dig);
3323 		bc_num_sub(&fracp2, &digit, &fracp1, 0);
3324 
3325 		// While the first three arguments should be self-explanatory, the last
3326 		// needs explaining. I don't want to print a newline when the last digit
3327 		// to be printed could take the place of the backslash rather than being
3328 		// pushed, as a single character, to the next line. That's what that
3329 		// last argument does for bc.
3330 		print(dig, len, radix, !newline || idigits != n->scale);
3331 
3332 		// Update the multipliers.
3333 		bc_num_mulArray(n1, base, n2);
3334 
3335 		radix = false;
3336 
3337 		// Switch.
3338 		temp = n1;
3339 		n1 = n2;
3340 		n2 = temp;
3341 	}
3342 
3343 frac_err:
3344 	BC_SIG_MAYLOCK;
3345 	bc_num_free(&flen2);
3346 	bc_num_free(&flen1);
3347 	bc_num_free(&fracp2);
3348 err:
3349 	BC_SIG_MAYLOCK;
3350 	bc_num_free(&fracp1);
3351 	bc_num_free(&intp);
3352 	bc_vec_free(&stack);
3353 	BC_LONGJMP_CONT(vm);
3354 }
3355 
3356 /**
3357  * Prints a number in the specified base, or rather, figures out which function
3358  * to call to print the number in the specified base and calls it.
3359  * @param n        The number to print.
3360  * @param base     The base to print in.
3361  * @param newline  Whether to print backslash+newlines on long enough lines.
3362  */
3363 static void
3364 bc_num_printBase(BcNum* restrict n, BcBigDig base, bool newline)
3365 {
3366 	size_t width;
3367 	BcNumDigitOp print;
3368 	bool neg = BC_NUM_NEG(n);
3369 
3370 	// Clear the sign because it makes the actual printing easier when we have
3371 	// to do math.
3372 	BC_NUM_NEG_CLR(n);
3373 
3374 	// Bases at hexadecimal and below are printed as one character, larger bases
3375 	// are printed as a series of digits separated by spaces.
3376 	if (base <= BC_NUM_MAX_POSIX_IBASE)
3377 	{
3378 		width = 1;
3379 		print = bc_num_printHex;
3380 	}
3381 	else
3382 	{
3383 		assert(base <= BC_BASE_POW);
3384 		width = bc_num_log10(base - 1);
3385 		print = bc_num_printDigits;
3386 	}
3387 
3388 	// Print.
3389 	bc_num_printNum(n, base, width, print, newline);
3390 
3391 	// Reset the sign.
3392 	n->rdx = BC_NUM_NEG_VAL(n, neg);
3393 }
3394 
3395 #if !BC_ENABLE_LIBRARY
3396 
3397 void
3398 bc_num_stream(BcNum* restrict n)
3399 {
3400 	bc_num_printNum(n, BC_NUM_STREAM_BASE, 1, bc_num_printChar, false);
3401 }
3402 
3403 #endif // !BC_ENABLE_LIBRARY
3404 
3405 void
3406 bc_num_setup(BcNum* restrict n, BcDig* restrict num, size_t cap)
3407 {
3408 	assert(n != NULL);
3409 	n->num = num;
3410 	n->cap = cap;
3411 	bc_num_zero(n);
3412 }
3413 
3414 void
3415 bc_num_init(BcNum* restrict n, size_t req)
3416 {
3417 	BcDig* num;
3418 
3419 	BC_SIG_ASSERT_LOCKED;
3420 
3421 	assert(n != NULL);
3422 
3423 	// BC_NUM_DEF_SIZE is set to be about the smallest allocation size that
3424 	// malloc() returns in practice, so just use it.
3425 	req = req >= BC_NUM_DEF_SIZE ? req : BC_NUM_DEF_SIZE;
3426 
3427 	// If we can't use a temp, allocate.
3428 	if (req != BC_NUM_DEF_SIZE) num = bc_vm_malloc(BC_NUM_SIZE(req));
3429 	else
3430 	{
3431 		num = bc_vm_getTemp() == NULL ? bc_vm_malloc(BC_NUM_SIZE(req)) :
3432 		                                bc_vm_takeTemp();
3433 	}
3434 
3435 	bc_num_setup(n, num, req);
3436 }
3437 
3438 void
3439 bc_num_clear(BcNum* restrict n)
3440 {
3441 	n->num = NULL;
3442 	n->cap = 0;
3443 }
3444 
3445 void
3446 bc_num_free(void* num)
3447 {
3448 	BcNum* n = (BcNum*) num;
3449 
3450 	BC_SIG_ASSERT_LOCKED;
3451 
3452 	assert(n != NULL);
3453 
3454 	if (n->cap == BC_NUM_DEF_SIZE) bc_vm_addTemp(n->num);
3455 	else free(n->num);
3456 }
3457 
3458 void
3459 bc_num_copy(BcNum* d, const BcNum* s)
3460 {
3461 	assert(d != NULL && s != NULL);
3462 
3463 	if (d == s) return;
3464 
3465 	bc_num_expand(d, s->len);
3466 	d->len = s->len;
3467 
3468 	// I can just copy directly here because the sign *and* rdx will be
3469 	// properly preserved.
3470 	d->rdx = s->rdx;
3471 	d->scale = s->scale;
3472 	// NOLINTNEXTLINE
3473 	memcpy(d->num, s->num, BC_NUM_SIZE(d->len));
3474 }
3475 
3476 void
3477 bc_num_createCopy(BcNum* d, const BcNum* s)
3478 {
3479 	BC_SIG_ASSERT_LOCKED;
3480 	bc_num_init(d, s->len);
3481 	bc_num_copy(d, s);
3482 }
3483 
3484 void
3485 bc_num_createFromBigdig(BcNum* restrict n, BcBigDig val)
3486 {
3487 	BC_SIG_ASSERT_LOCKED;
3488 	bc_num_init(n, BC_NUM_BIGDIG_LOG10);
3489 	bc_num_bigdig2num(n, val);
3490 }
3491 
3492 size_t
3493 bc_num_scale(const BcNum* restrict n)
3494 {
3495 	return n->scale;
3496 }
3497 
3498 size_t
3499 bc_num_len(const BcNum* restrict n)
3500 {
3501 	size_t len = n->len;
3502 
3503 	// Always return at least 1.
3504 	if (BC_NUM_ZERO(n)) return n->scale ? n->scale : 1;
3505 
3506 	// If this is true, there is no integer portion of the number.
3507 	if (BC_NUM_RDX_VAL(n) == len)
3508 	{
3509 		// We have to take into account the fact that some of the digits right
3510 		// after the decimal could be zero. If that is the case, we need to
3511 		// ignore them until we hit the first non-zero digit.
3512 
3513 		size_t zero, scale;
3514 
3515 		// The number of limbs with non-zero digits.
3516 		len = bc_num_nonZeroLen(n);
3517 
3518 		// Get the number of digits in the last limb.
3519 		scale = n->scale % BC_BASE_DIGS;
3520 		scale = scale ? scale : BC_BASE_DIGS;
3521 
3522 		// Get the number of zero digits.
3523 		zero = bc_num_zeroDigits(n->num + len - 1);
3524 
3525 		// Calculate the true length.
3526 		len = len * BC_BASE_DIGS - zero - (BC_BASE_DIGS - scale);
3527 	}
3528 	// Otherwise, count the number of int digits and return that plus the scale.
3529 	else len = bc_num_intDigits(n) + n->scale;
3530 
3531 	return len;
3532 }
3533 
3534 void
3535 bc_num_parse(BcNum* restrict n, const char* restrict val, BcBigDig base)
3536 {
3537 #if BC_DEBUG
3538 #if BC_ENABLE_LIBRARY
3539 	BcVm* vm = bcl_getspecific();
3540 #endif // BC_ENABLE_LIBRARY
3541 #endif // BC_DEBUG
3542 
3543 	assert(n != NULL && val != NULL && base);
3544 	assert(base >= BC_NUM_MIN_BASE && base <= vm->maxes[BC_PROG_GLOBALS_IBASE]);
3545 	assert(bc_num_strValid(val));
3546 
3547 	// A one character number is *always* parsed as though the base was the
3548 	// maximum allowed ibase, per the bc spec.
3549 	if (!val[1])
3550 	{
3551 		BcBigDig dig = bc_num_parseChar(val[0], BC_NUM_MAX_LBASE);
3552 		bc_num_bigdig2num(n, dig);
3553 	}
3554 	else if (base == BC_BASE) bc_num_parseDecimal(n, val);
3555 	else bc_num_parseBase(n, val, base);
3556 
3557 	assert(BC_NUM_RDX_VALID(n));
3558 }
3559 
3560 void
3561 bc_num_print(BcNum* restrict n, BcBigDig base, bool newline)
3562 {
3563 	assert(n != NULL);
3564 	assert(BC_ENABLE_EXTRA_MATH || base >= BC_NUM_MIN_BASE);
3565 
3566 	// We may need a newline, just to start.
3567 	bc_num_printNewline();
3568 
3569 	if (BC_NUM_NONZERO(n))
3570 	{
3571 #if BC_ENABLE_LIBRARY
3572 		BcVm* vm = bcl_getspecific();
3573 #endif // BC_ENABLE_LIBRARY
3574 
3575 		// Print the sign.
3576 		if (BC_NUM_NEG(n)) bc_num_putchar('-', true);
3577 
3578 		// Print the leading zero if necessary. We don't print when using
3579 		// scientific or engineering modes.
3580 		if (BC_Z && BC_NUM_RDX_VAL(n) == n->len && base != 0 && base != 1)
3581 		{
3582 			bc_num_printHex(0, 1, false, !newline);
3583 		}
3584 	}
3585 
3586 	// Short-circuit 0.
3587 	if (BC_NUM_ZERO(n)) bc_num_printHex(0, 1, false, !newline);
3588 	else if (base == BC_BASE) bc_num_printDecimal(n, newline);
3589 #if BC_ENABLE_EXTRA_MATH
3590 	else if (base == 0 || base == 1)
3591 	{
3592 		bc_num_printExponent(n, base != 0, newline);
3593 	}
3594 #endif // BC_ENABLE_EXTRA_MATH
3595 	else bc_num_printBase(n, base, newline);
3596 
3597 	if (newline) bc_num_putchar('\n', false);
3598 }
3599 
3600 BcBigDig
3601 bc_num_bigdig2(const BcNum* restrict n)
3602 {
3603 #if BC_DEBUG
3604 #if BC_ENABLE_LIBRARY
3605 	BcVm* vm = bcl_getspecific();
3606 #endif // BC_ENABLE_LIBRARY
3607 #endif // BC_DEBUG
3608 
3609 	// This function returns no errors because it's guaranteed to succeed if
3610 	// its preconditions are met. Those preconditions include both n needs to
3611 	// be non-NULL, n being non-negative, and n being less than vm->max. If all
3612 	// of that is true, then we can just convert without worrying about negative
3613 	// errors or overflow.
3614 
3615 	BcBigDig r = 0;
3616 	size_t nrdx = BC_NUM_RDX_VAL(n);
3617 
3618 	assert(n != NULL);
3619 	assert(!BC_NUM_NEG(n));
3620 	assert(bc_num_cmp(n, &vm->max) < 0);
3621 	assert(n->len - nrdx <= 3);
3622 
3623 	// There is a small speed win from unrolling the loop here, and since it
3624 	// only adds 53 bytes, I decided that it was worth it.
3625 	switch (n->len - nrdx)
3626 	{
3627 		case 3:
3628 		{
3629 			r = (BcBigDig) n->num[nrdx + 2];
3630 
3631 			// Fallthrough.
3632 			BC_FALLTHROUGH
3633 		}
3634 
3635 		case 2:
3636 		{
3637 			r = r * BC_BASE_POW + (BcBigDig) n->num[nrdx + 1];
3638 
3639 			// Fallthrough.
3640 			BC_FALLTHROUGH
3641 		}
3642 
3643 		case 1:
3644 		{
3645 			r = r * BC_BASE_POW + (BcBigDig) n->num[nrdx];
3646 		}
3647 	}
3648 
3649 	return r;
3650 }
3651 
3652 BcBigDig
3653 bc_num_bigdig(const BcNum* restrict n)
3654 {
3655 #if BC_ENABLE_LIBRARY
3656 	BcVm* vm = bcl_getspecific();
3657 #endif // BC_ENABLE_LIBRARY
3658 
3659 	assert(n != NULL);
3660 
3661 	// This error checking is extremely important, and if you do not have a
3662 	// guarantee that converting a number will always succeed in a particular
3663 	// case, you *must* call this function to get these error checks. This
3664 	// includes all instances of numbers inputted by the user or calculated by
3665 	// the user. Otherwise, you can call the faster bc_num_bigdig2().
3666 	if (BC_ERR(BC_NUM_NEG(n))) bc_err(BC_ERR_MATH_NEGATIVE);
3667 	if (BC_ERR(bc_num_cmp(n, &vm->max) >= 0)) bc_err(BC_ERR_MATH_OVERFLOW);
3668 
3669 	return bc_num_bigdig2(n);
3670 }
3671 
3672 void
3673 bc_num_bigdig2num(BcNum* restrict n, BcBigDig val)
3674 {
3675 	BcDig* ptr;
3676 	size_t i;
3677 
3678 	assert(n != NULL);
3679 
3680 	bc_num_zero(n);
3681 
3682 	// Already 0.
3683 	if (!val) return;
3684 
3685 	// Expand first. This is the only way this function can fail, and it's a
3686 	// fatal error.
3687 	bc_num_expand(n, BC_NUM_BIGDIG_LOG10);
3688 
3689 	// The conversion is easy because numbers are laid out in little-endian
3690 	// order.
3691 	for (ptr = n->num, i = 0; val; ++i, val /= BC_BASE_POW)
3692 	{
3693 		ptr[i] = val % BC_BASE_POW;
3694 	}
3695 
3696 	n->len = i;
3697 }
3698 
3699 #if BC_ENABLE_EXTRA_MATH
3700 
3701 void
3702 bc_num_rng(const BcNum* restrict n, BcRNG* rng)
3703 {
3704 	BcNum temp, temp2, intn, frac;
3705 	BcRand state1, state2, inc1, inc2;
3706 	size_t nrdx = BC_NUM_RDX_VAL(n);
3707 #if BC_ENABLE_LIBRARY
3708 	BcVm* vm = bcl_getspecific();
3709 #endif // BC_ENABLE_LIBRARY
3710 
3711 	// This function holds the secret of how I interpret a seed number for the
3712 	// PRNG. Well, it's actually in the development manual
3713 	// (manuals/development.md#pseudo-random-number-generator), so look there
3714 	// before you try to understand this.
3715 
3716 	BC_SIG_LOCK;
3717 
3718 	bc_num_init(&temp, n->len);
3719 	bc_num_init(&temp2, n->len);
3720 	bc_num_init(&frac, nrdx);
3721 	bc_num_init(&intn, bc_num_int(n));
3722 
3723 	BC_SETJMP_LOCKED(vm, err);
3724 
3725 	BC_SIG_UNLOCK;
3726 
3727 	assert(BC_NUM_RDX_VALID_NP(vm->max));
3728 
3729 	// NOLINTNEXTLINE
3730 	memcpy(frac.num, n->num, BC_NUM_SIZE(nrdx));
3731 	frac.len = nrdx;
3732 	BC_NUM_RDX_SET_NP(frac, nrdx);
3733 	frac.scale = n->scale;
3734 
3735 	assert(BC_NUM_RDX_VALID_NP(frac));
3736 	assert(BC_NUM_RDX_VALID_NP(vm->max2));
3737 
3738 	// Multiply the fraction and truncate so that it's an integer. The
3739 	// truncation is what clamps it, by the way.
3740 	bc_num_mul(&frac, &vm->max2, &temp, 0);
3741 	bc_num_truncate(&temp, temp.scale);
3742 	bc_num_copy(&frac, &temp);
3743 
3744 	// Get the integer.
3745 	// NOLINTNEXTLINE
3746 	memcpy(intn.num, n->num + nrdx, BC_NUM_SIZE(bc_num_int(n)));
3747 	intn.len = bc_num_int(n);
3748 
3749 	// This assert is here because it has to be true. It is also here to justify
3750 	// some optimizations.
3751 	assert(BC_NUM_NONZERO(&vm->max));
3752 
3753 	// If there *was* a fractional part...
3754 	if (BC_NUM_NONZERO(&frac))
3755 	{
3756 		// This divmod splits frac into the two state parts.
3757 		bc_num_divmod(&frac, &vm->max, &temp, &temp2, 0);
3758 
3759 		// frac is guaranteed to be smaller than vm->max * vm->max (pow).
3760 		// This means that when dividing frac by vm->max, as above, the
3761 		// quotient and remainder are both guaranteed to be less than vm->max,
3762 		// which means we can use bc_num_bigdig2() here and not worry about
3763 		// overflow.
3764 		state1 = (BcRand) bc_num_bigdig2(&temp2);
3765 		state2 = (BcRand) bc_num_bigdig2(&temp);
3766 	}
3767 	else state1 = state2 = 0;
3768 
3769 	// If there *was* an integer part...
3770 	if (BC_NUM_NONZERO(&intn))
3771 	{
3772 		// This divmod splits intn into the two inc parts.
3773 		bc_num_divmod(&intn, &vm->max, &temp, &temp2, 0);
3774 
3775 		// Because temp2 is the mod of vm->max, from above, it is guaranteed
3776 		// to be small enough to use bc_num_bigdig2().
3777 		inc1 = (BcRand) bc_num_bigdig2(&temp2);
3778 
3779 		// Clamp the second inc part.
3780 		if (bc_num_cmp(&temp, &vm->max) >= 0)
3781 		{
3782 			bc_num_copy(&temp2, &temp);
3783 			bc_num_mod(&temp2, &vm->max, &temp, 0);
3784 		}
3785 
3786 		// The if statement above ensures that temp is less than vm->max, which
3787 		// means that we can use bc_num_bigdig2() here.
3788 		inc2 = (BcRand) bc_num_bigdig2(&temp);
3789 	}
3790 	else inc1 = inc2 = 0;
3791 
3792 	bc_rand_seed(rng, state1, state2, inc1, inc2);
3793 
3794 err:
3795 	BC_SIG_MAYLOCK;
3796 	bc_num_free(&intn);
3797 	bc_num_free(&frac);
3798 	bc_num_free(&temp2);
3799 	bc_num_free(&temp);
3800 	BC_LONGJMP_CONT(vm);
3801 }
3802 
3803 void
3804 bc_num_createFromRNG(BcNum* restrict n, BcRNG* rng)
3805 {
3806 	BcRand s1, s2, i1, i2;
3807 	BcNum conv, temp1, temp2, temp3;
3808 	BcDig temp1_num[BC_RAND_NUM_SIZE], temp2_num[BC_RAND_NUM_SIZE];
3809 	BcDig conv_num[BC_NUM_BIGDIG_LOG10];
3810 #if BC_ENABLE_LIBRARY
3811 	BcVm* vm = bcl_getspecific();
3812 #endif // BC_ENABLE_LIBRARY
3813 
3814 	BC_SIG_LOCK;
3815 
3816 	bc_num_init(&temp3, 2 * BC_RAND_NUM_SIZE);
3817 
3818 	BC_SETJMP_LOCKED(vm, err);
3819 
3820 	BC_SIG_UNLOCK;
3821 
3822 	bc_num_setup(&temp1, temp1_num, sizeof(temp1_num) / sizeof(BcDig));
3823 	bc_num_setup(&temp2, temp2_num, sizeof(temp2_num) / sizeof(BcDig));
3824 	bc_num_setup(&conv, conv_num, sizeof(conv_num) / sizeof(BcDig));
3825 
3826 	// This assert is here because it has to be true. It is also here to justify
3827 	// the assumption that vm->max is not zero.
3828 	assert(BC_NUM_NONZERO(&vm->max));
3829 
3830 	// Because this is true, we can just ignore math errors that would happen
3831 	// otherwise.
3832 	assert(BC_NUM_NONZERO(&vm->max2));
3833 
3834 	bc_rand_getRands(rng, &s1, &s2, &i1, &i2);
3835 
3836 	// Put the second piece of state into a number.
3837 	bc_num_bigdig2num(&conv, (BcBigDig) s2);
3838 
3839 	assert(BC_NUM_RDX_VALID_NP(conv));
3840 
3841 	// Multiply by max to make room for the first piece of state.
3842 	bc_num_mul(&conv, &vm->max, &temp1, 0);
3843 
3844 	// Add in the first piece of state.
3845 	bc_num_bigdig2num(&conv, (BcBigDig) s1);
3846 	bc_num_add(&conv, &temp1, &temp2, 0);
3847 
3848 	// Divide to make it an entirely fractional part.
3849 	bc_num_div(&temp2, &vm->max2, &temp3, BC_RAND_STATE_BITS);
3850 
3851 	// Now start on the increment parts. It's the same process without the
3852 	// divide, so put the second piece of increment into a number.
3853 	bc_num_bigdig2num(&conv, (BcBigDig) i2);
3854 
3855 	assert(BC_NUM_RDX_VALID_NP(conv));
3856 
3857 	// Multiply by max to make room for the first piece of increment.
3858 	bc_num_mul(&conv, &vm->max, &temp1, 0);
3859 
3860 	// Add in the first piece of increment.
3861 	bc_num_bigdig2num(&conv, (BcBigDig) i1);
3862 	bc_num_add(&conv, &temp1, &temp2, 0);
3863 
3864 	// Now add the two together.
3865 	bc_num_add(&temp2, &temp3, n, 0);
3866 
3867 	assert(BC_NUM_RDX_VALID(n));
3868 
3869 err:
3870 	BC_SIG_MAYLOCK;
3871 	bc_num_free(&temp3);
3872 	BC_LONGJMP_CONT(vm);
3873 }
3874 
3875 void
3876 bc_num_irand(BcNum* restrict a, BcNum* restrict b, BcRNG* restrict rng)
3877 {
3878 	BcNum atemp;
3879 	size_t i;
3880 
3881 	assert(a != b);
3882 
3883 	if (BC_ERR(BC_NUM_NEG(a))) bc_err(BC_ERR_MATH_NEGATIVE);
3884 
3885 	// If either of these are true, then the numbers are integers.
3886 	if (BC_NUM_ZERO(a) || BC_NUM_ONE(a)) return;
3887 
3888 #if BC_GCC
3889 	// This is here in GCC to quiet the "maybe-uninitialized" warning.
3890 	atemp.num = NULL;
3891 	atemp.len = 0;
3892 #endif // BC_GCC
3893 
3894 	if (BC_ERR(bc_num_nonInt(a, &atemp))) bc_err(BC_ERR_MATH_NON_INTEGER);
3895 
3896 	assert(atemp.num != NULL);
3897 	assert(atemp.len);
3898 
3899 	if (atemp.len > 2)
3900 	{
3901 		size_t len;
3902 
3903 		len = atemp.len - 2;
3904 
3905 		// Just generate a random number for each limb.
3906 		for (i = 0; i < len; i += 2)
3907 		{
3908 			BcRand dig;
3909 
3910 			dig = bc_rand_bounded(rng, BC_BASE_RAND_POW);
3911 
3912 			b->num[i] = (BcDig) (dig % BC_BASE_POW);
3913 			b->num[i + 1] = (BcDig) (dig / BC_BASE_POW);
3914 		}
3915 	}
3916 	else
3917 	{
3918 		// We need this set.
3919 		i = 0;
3920 	}
3921 
3922 	// This will be true if there's one full limb after the two limb groups.
3923 	if (i == atemp.len - 2)
3924 	{
3925 		// Increment this for easy use.
3926 		i += 1;
3927 
3928 		// If the last digit is not one, we need to set a bound for it
3929 		// explicitly. Since there's still an empty limb, we need to fill that.
3930 		if (atemp.num[i] != 1)
3931 		{
3932 			BcRand dig;
3933 			BcRand bound;
3934 
3935 			// Set the bound to the bound of the last limb times the amount
3936 			// needed to fill the second-to-last limb as well.
3937 			bound = ((BcRand) atemp.num[i]) * BC_BASE_POW;
3938 
3939 			dig = bc_rand_bounded(rng, bound);
3940 
3941 			// Fill the last two.
3942 			b->num[i - 1] = (BcDig) (dig % BC_BASE_POW);
3943 			b->num[i] = (BcDig) (dig / BC_BASE_POW);
3944 
3945 			// Ensure that the length will be correct. If the last limb is zero,
3946 			// then the length needs to be one less than the bound.
3947 			b->len = atemp.len - (b->num[i] == 0);
3948 		}
3949 		// Here the last limb *is* one, which means the last limb does *not*
3950 		// need to be filled. Also, the length needs to be one less because the
3951 		// last limb is 0.
3952 		else
3953 		{
3954 			b->num[i - 1] = (BcDig) bc_rand_bounded(rng, BC_BASE_POW);
3955 			b->len = atemp.len - 1;
3956 		}
3957 	}
3958 	// Here, there is only one limb to fill.
3959 	else
3960 	{
3961 		// See above for how this works.
3962 		if (atemp.num[i] != 1)
3963 		{
3964 			b->num[i] = (BcDig) bc_rand_bounded(rng, (BcRand) atemp.num[i]);
3965 			b->len = atemp.len - (b->num[i] == 0);
3966 		}
3967 		else b->len = atemp.len - 1;
3968 	}
3969 
3970 	bc_num_clean(b);
3971 
3972 	assert(BC_NUM_RDX_VALID(b));
3973 }
3974 #endif // BC_ENABLE_EXTRA_MATH
3975 
3976 size_t
3977 bc_num_addReq(const BcNum* a, const BcNum* b, size_t scale)
3978 {
3979 	size_t aint, bint, ardx, brdx;
3980 
3981 	// Addition and subtraction require the max of the length of the two numbers
3982 	// plus 1.
3983 
3984 	BC_UNUSED(scale);
3985 
3986 	ardx = BC_NUM_RDX_VAL(a);
3987 	aint = bc_num_int(a);
3988 	assert(aint <= a->len && ardx <= a->len);
3989 
3990 	brdx = BC_NUM_RDX_VAL(b);
3991 	bint = bc_num_int(b);
3992 	assert(bint <= b->len && brdx <= b->len);
3993 
3994 	ardx = BC_MAX(ardx, brdx);
3995 	aint = BC_MAX(aint, bint);
3996 
3997 	return bc_vm_growSize(bc_vm_growSize(ardx, aint), 1);
3998 }
3999 
4000 size_t
4001 bc_num_mulReq(const BcNum* a, const BcNum* b, size_t scale)
4002 {
4003 	size_t max, rdx;
4004 
4005 	// Multiplication requires the sum of the lengths of the numbers.
4006 
4007 	rdx = bc_vm_growSize(BC_NUM_RDX_VAL(a), BC_NUM_RDX_VAL(b));
4008 
4009 	max = BC_NUM_RDX(scale);
4010 
4011 	max = bc_vm_growSize(BC_MAX(max, rdx), 1);
4012 	rdx = bc_vm_growSize(bc_vm_growSize(bc_num_int(a), bc_num_int(b)), max);
4013 
4014 	return rdx;
4015 }
4016 
4017 size_t
4018 bc_num_divReq(const BcNum* a, const BcNum* b, size_t scale)
4019 {
4020 	size_t max, rdx;
4021 
4022 	// Division requires the length of the dividend plus the scale.
4023 
4024 	rdx = bc_vm_growSize(BC_NUM_RDX_VAL(a), BC_NUM_RDX_VAL(b));
4025 
4026 	max = BC_NUM_RDX(scale);
4027 
4028 	max = bc_vm_growSize(BC_MAX(max, rdx), 1);
4029 	rdx = bc_vm_growSize(bc_num_int(a), max);
4030 
4031 	return rdx;
4032 }
4033 
4034 size_t
4035 bc_num_powReq(const BcNum* a, const BcNum* b, size_t scale)
4036 {
4037 	BC_UNUSED(scale);
4038 	return bc_vm_growSize(bc_vm_growSize(a->len, b->len), 1);
4039 }
4040 
4041 #if BC_ENABLE_EXTRA_MATH
4042 size_t
4043 bc_num_placesReq(const BcNum* a, const BcNum* b, size_t scale)
4044 {
4045 	BC_UNUSED(scale);
4046 	return a->len + b->len - BC_NUM_RDX_VAL(a) - BC_NUM_RDX_VAL(b);
4047 }
4048 #endif // BC_ENABLE_EXTRA_MATH
4049 
4050 void
4051 bc_num_add(BcNum* a, BcNum* b, BcNum* c, size_t scale)
4052 {
4053 	assert(BC_NUM_RDX_VALID(a));
4054 	assert(BC_NUM_RDX_VALID(b));
4055 	bc_num_binary(a, b, c, false, bc_num_as, bc_num_addReq(a, b, scale));
4056 }
4057 
4058 void
4059 bc_num_sub(BcNum* a, BcNum* b, BcNum* c, size_t scale)
4060 {
4061 	assert(BC_NUM_RDX_VALID(a));
4062 	assert(BC_NUM_RDX_VALID(b));
4063 	bc_num_binary(a, b, c, true, bc_num_as, bc_num_addReq(a, b, scale));
4064 }
4065 
4066 void
4067 bc_num_mul(BcNum* a, BcNum* b, BcNum* c, size_t scale)
4068 {
4069 	assert(BC_NUM_RDX_VALID(a));
4070 	assert(BC_NUM_RDX_VALID(b));
4071 	bc_num_binary(a, b, c, scale, bc_num_m, bc_num_mulReq(a, b, scale));
4072 }
4073 
4074 void
4075 bc_num_div(BcNum* a, BcNum* b, BcNum* c, size_t scale)
4076 {
4077 	assert(BC_NUM_RDX_VALID(a));
4078 	assert(BC_NUM_RDX_VALID(b));
4079 	bc_num_binary(a, b, c, scale, bc_num_d, bc_num_divReq(a, b, scale));
4080 }
4081 
4082 void
4083 bc_num_mod(BcNum* a, BcNum* b, BcNum* c, size_t scale)
4084 {
4085 	assert(BC_NUM_RDX_VALID(a));
4086 	assert(BC_NUM_RDX_VALID(b));
4087 	bc_num_binary(a, b, c, scale, bc_num_rem, bc_num_divReq(a, b, scale));
4088 }
4089 
4090 void
4091 bc_num_pow(BcNum* a, BcNum* b, BcNum* c, size_t scale)
4092 {
4093 	assert(BC_NUM_RDX_VALID(a));
4094 	assert(BC_NUM_RDX_VALID(b));
4095 	bc_num_binary(a, b, c, scale, bc_num_p, bc_num_powReq(a, b, scale));
4096 }
4097 
4098 #if BC_ENABLE_EXTRA_MATH
4099 void
4100 bc_num_places(BcNum* a, BcNum* b, BcNum* c, size_t scale)
4101 {
4102 	assert(BC_NUM_RDX_VALID(a));
4103 	assert(BC_NUM_RDX_VALID(b));
4104 	bc_num_binary(a, b, c, scale, bc_num_place, bc_num_placesReq(a, b, scale));
4105 }
4106 
4107 void
4108 bc_num_lshift(BcNum* a, BcNum* b, BcNum* c, size_t scale)
4109 {
4110 	assert(BC_NUM_RDX_VALID(a));
4111 	assert(BC_NUM_RDX_VALID(b));
4112 	bc_num_binary(a, b, c, scale, bc_num_left, bc_num_placesReq(a, b, scale));
4113 }
4114 
4115 void
4116 bc_num_rshift(BcNum* a, BcNum* b, BcNum* c, size_t scale)
4117 {
4118 	assert(BC_NUM_RDX_VALID(a));
4119 	assert(BC_NUM_RDX_VALID(b));
4120 	bc_num_binary(a, b, c, scale, bc_num_right, bc_num_placesReq(a, b, scale));
4121 }
4122 #endif // BC_ENABLE_EXTRA_MATH
4123 
4124 void
4125 bc_num_sqrt(BcNum* restrict a, BcNum* restrict b, size_t scale)
4126 {
4127 	BcNum num1, num2, half, f, fprime;
4128 	BcNum* x0;
4129 	BcNum* x1;
4130 	BcNum* temp;
4131 	// realscale is meant to quiet a warning on GCC about longjmp() clobbering.
4132 	// This one is real.
4133 	size_t pow, len, rdx, req, resscale, realscale;
4134 	BcDig half_digs[1];
4135 #if BC_ENABLE_LIBRARY
4136 	BcVm* vm = bcl_getspecific();
4137 #endif // BC_ENABLE_LIBRARY
4138 
4139 	assert(a != NULL && b != NULL && a != b);
4140 
4141 	if (BC_ERR(BC_NUM_NEG(a))) bc_err(BC_ERR_MATH_NEGATIVE);
4142 
4143 	// We want to calculate to a's scale if it is bigger so that the result will
4144 	// truncate properly.
4145 	if (a->scale > scale) realscale = a->scale;
4146 	else realscale = scale;
4147 
4148 	// Set parameters for the result.
4149 	len = bc_vm_growSize(bc_num_intDigits(a), 1);
4150 	rdx = BC_NUM_RDX(realscale);
4151 
4152 	// Square root needs half of the length of the parameter.
4153 	req = bc_vm_growSize(BC_MAX(rdx, BC_NUM_RDX_VAL(a)), len >> 1);
4154 	req = bc_vm_growSize(req, 1);
4155 
4156 	BC_SIG_LOCK;
4157 
4158 	// Unlike the binary operators, this function is the only single parameter
4159 	// function and is expected to initialize the result. This means that it
4160 	// expects that b is *NOT* preallocated. We allocate it here.
4161 	bc_num_init(b, req);
4162 
4163 	BC_SIG_UNLOCK;
4164 
4165 	assert(a != NULL && b != NULL && a != b);
4166 	assert(a->num != NULL && b->num != NULL);
4167 
4168 	// Easy case.
4169 	if (BC_NUM_ZERO(a))
4170 	{
4171 		bc_num_setToZero(b, realscale);
4172 		return;
4173 	}
4174 
4175 	// Another easy case.
4176 	if (BC_NUM_ONE(a))
4177 	{
4178 		bc_num_one(b);
4179 		bc_num_extend(b, realscale);
4180 		return;
4181 	}
4182 
4183 	// Set the parameters again.
4184 	rdx = BC_NUM_RDX(realscale);
4185 	rdx = BC_MAX(rdx, BC_NUM_RDX_VAL(a));
4186 	len = bc_vm_growSize(a->len, rdx);
4187 
4188 	BC_SIG_LOCK;
4189 
4190 	bc_num_init(&num1, len);
4191 	bc_num_init(&num2, len);
4192 	bc_num_setup(&half, half_digs, sizeof(half_digs) / sizeof(BcDig));
4193 
4194 	// There is a division by two in the formula. We set up a number that's 1/2
4195 	// so that we can use multiplication instead of heavy division.
4196 	bc_num_setToZero(&half, 1);
4197 	half.num[0] = BC_BASE_POW / 2;
4198 	half.len = 1;
4199 	BC_NUM_RDX_SET_NP(half, 1);
4200 
4201 	bc_num_init(&f, len);
4202 	bc_num_init(&fprime, len);
4203 
4204 	BC_SETJMP_LOCKED(vm, err);
4205 
4206 	BC_SIG_UNLOCK;
4207 
4208 	// Pointers for easy switching.
4209 	x0 = &num1;
4210 	x1 = &num2;
4211 
4212 	// Start with 1.
4213 	bc_num_one(x0);
4214 
4215 	// The power of the operand is needed for the estimate.
4216 	pow = bc_num_intDigits(a);
4217 
4218 	// The code in this if statement calculates the initial estimate. First, if
4219 	// a is less than 1, then 0 is a good estimate. Otherwise, we want something
4220 	// in the same ballpark. That ballpark is half of pow because the result
4221 	// will have half the digits.
4222 	if (pow)
4223 	{
4224 		// An odd number is served by starting with 2^((pow-1)/2), and an even
4225 		// number is served by starting with 6^((pow-2)/2). Why? Because math.
4226 		if (pow & 1) x0->num[0] = 2;
4227 		else x0->num[0] = 6;
4228 
4229 		pow -= 2 - (pow & 1);
4230 		bc_num_shiftLeft(x0, pow / 2);
4231 	}
4232 
4233 	// I can set the rdx here directly because neg should be false.
4234 	x0->scale = x0->rdx = 0;
4235 	resscale = (realscale + BC_BASE_DIGS) + 2;
4236 
4237 	// This is the calculation loop. This compare goes to 0 eventually as the
4238 	// difference between the two numbers gets smaller than resscale.
4239 	while (bc_num_cmp(x1, x0))
4240 	{
4241 		assert(BC_NUM_NONZERO(x0));
4242 
4243 		// This loop directly corresponds to the iteration in Newton's method.
4244 		// If you know the formula, this loop makes sense. Go study the formula.
4245 
4246 		bc_num_div(a, x0, &f, resscale);
4247 		bc_num_add(x0, &f, &fprime, resscale);
4248 
4249 		assert(BC_NUM_RDX_VALID_NP(fprime));
4250 		assert(BC_NUM_RDX_VALID_NP(half));
4251 
4252 		bc_num_mul(&fprime, &half, x1, resscale);
4253 
4254 		// Switch.
4255 		temp = x0;
4256 		x0 = x1;
4257 		x1 = temp;
4258 	}
4259 
4260 	// Copy to the result and truncate.
4261 	bc_num_copy(b, x0);
4262 	if (b->scale > realscale) bc_num_truncate(b, b->scale - realscale);
4263 
4264 	assert(!BC_NUM_NEG(b) || BC_NUM_NONZERO(b));
4265 	assert(BC_NUM_RDX_VALID(b));
4266 	assert(BC_NUM_RDX_VAL(b) <= b->len || !b->len);
4267 	assert(!b->len || b->num[b->len - 1] || BC_NUM_RDX_VAL(b) == b->len);
4268 
4269 err:
4270 	BC_SIG_MAYLOCK;
4271 	bc_num_free(&fprime);
4272 	bc_num_free(&f);
4273 	bc_num_free(&num2);
4274 	bc_num_free(&num1);
4275 	BC_LONGJMP_CONT(vm);
4276 }
4277 
4278 void
4279 bc_num_divmod(BcNum* a, BcNum* b, BcNum* c, BcNum* d, size_t scale)
4280 {
4281 	size_t ts, len;
4282 	BcNum *ptr_a, num2;
4283 	// This is volatile to quiet a warning on GCC about clobbering with
4284 	// longjmp().
4285 	volatile bool init = false;
4286 #if BC_ENABLE_LIBRARY
4287 	BcVm* vm = bcl_getspecific();
4288 #endif // BC_ENABLE_LIBRARY
4289 
4290 	// The bulk of this function is just doing what bc_num_binary() does for the
4291 	// binary operators. However, it assumes that only c and a can be equal.
4292 
4293 	// Set up the parameters.
4294 	ts = BC_MAX(scale + b->scale, a->scale);
4295 	len = bc_num_mulReq(a, b, ts);
4296 
4297 	assert(a != NULL && b != NULL && c != NULL && d != NULL);
4298 	assert(c != d && a != d && b != d && b != c);
4299 
4300 	// Initialize or expand as necessary.
4301 	if (c == a)
4302 	{
4303 		// NOLINTNEXTLINE
4304 		memcpy(&num2, c, sizeof(BcNum));
4305 		ptr_a = &num2;
4306 
4307 		BC_SIG_LOCK;
4308 
4309 		bc_num_init(c, len);
4310 
4311 		init = true;
4312 
4313 		BC_SETJMP_LOCKED(vm, err);
4314 
4315 		BC_SIG_UNLOCK;
4316 	}
4317 	else
4318 	{
4319 		ptr_a = a;
4320 		bc_num_expand(c, len);
4321 	}
4322 
4323 	// Do the quick version if possible.
4324 	if (BC_NUM_NONZERO(a) && !BC_NUM_RDX_VAL(a) && !BC_NUM_RDX_VAL(b) &&
4325 	    b->len == 1 && !scale)
4326 	{
4327 		BcBigDig rem;
4328 
4329 		bc_num_divArray(ptr_a, (BcBigDig) b->num[0], c, &rem);
4330 
4331 		assert(rem < BC_BASE_POW);
4332 
4333 		d->num[0] = (BcDig) rem;
4334 		d->len = (rem != 0);
4335 	}
4336 	// Do the slow method.
4337 	else bc_num_r(ptr_a, b, c, d, scale, ts);
4338 
4339 	assert(!BC_NUM_NEG(c) || BC_NUM_NONZERO(c));
4340 	assert(BC_NUM_RDX_VALID(c));
4341 	assert(BC_NUM_RDX_VAL(c) <= c->len || !c->len);
4342 	assert(!c->len || c->num[c->len - 1] || BC_NUM_RDX_VAL(c) == c->len);
4343 	assert(!BC_NUM_NEG(d) || BC_NUM_NONZERO(d));
4344 	assert(BC_NUM_RDX_VALID(d));
4345 	assert(BC_NUM_RDX_VAL(d) <= d->len || !d->len);
4346 	assert(!d->len || d->num[d->len - 1] || BC_NUM_RDX_VAL(d) == d->len);
4347 
4348 err:
4349 	// Only cleanup if we initialized.
4350 	if (init)
4351 	{
4352 		BC_SIG_MAYLOCK;
4353 		bc_num_free(&num2);
4354 		BC_LONGJMP_CONT(vm);
4355 	}
4356 }
4357 
4358 void
4359 bc_num_modexp(BcNum* a, BcNum* b, BcNum* c, BcNum* restrict d)
4360 {
4361 	BcNum base, exp, two, temp, atemp, btemp, ctemp;
4362 	BcDig two_digs[2];
4363 #if BC_ENABLE_LIBRARY
4364 	BcVm* vm = bcl_getspecific();
4365 #endif // BC_ENABLE_LIBRARY
4366 
4367 	assert(a != NULL && b != NULL && c != NULL && d != NULL);
4368 	assert(a != d && b != d && c != d);
4369 
4370 	if (BC_ERR(BC_NUM_ZERO(c))) bc_err(BC_ERR_MATH_DIVIDE_BY_ZERO);
4371 	if (BC_ERR(BC_NUM_NEG(b))) bc_err(BC_ERR_MATH_NEGATIVE);
4372 
4373 #if BC_DEBUG || BC_GCC
4374 	// This is entirely for quieting a useless scan-build error.
4375 	btemp.len = 0;
4376 	ctemp.len = 0;
4377 #endif // BC_DEBUG || BC_GCC
4378 
4379 	// Eliminate fractional parts that are zero or error if they are not zero.
4380 	if (BC_ERR(bc_num_nonInt(a, &atemp) || bc_num_nonInt(b, &btemp) ||
4381 	           bc_num_nonInt(c, &ctemp)))
4382 	{
4383 		bc_err(BC_ERR_MATH_NON_INTEGER);
4384 	}
4385 
4386 	bc_num_expand(d, ctemp.len);
4387 
4388 	BC_SIG_LOCK;
4389 
4390 	bc_num_init(&base, ctemp.len);
4391 	bc_num_setup(&two, two_digs, sizeof(two_digs) / sizeof(BcDig));
4392 	bc_num_init(&temp, btemp.len + 1);
4393 	bc_num_createCopy(&exp, &btemp);
4394 
4395 	BC_SETJMP_LOCKED(vm, err);
4396 
4397 	BC_SIG_UNLOCK;
4398 
4399 	bc_num_one(&two);
4400 	two.num[0] = 2;
4401 	bc_num_one(d);
4402 
4403 	// We already checked for 0.
4404 	bc_num_rem(&atemp, &ctemp, &base, 0);
4405 
4406 	// If you know the algorithm I used, the memory-efficient method, then this
4407 	// loop should be self-explanatory because it is the calculation loop.
4408 	while (BC_NUM_NONZERO(&exp))
4409 	{
4410 		// Num two cannot be 0, so no errors.
4411 		bc_num_divmod(&exp, &two, &exp, &temp, 0);
4412 
4413 		if (BC_NUM_ONE(&temp) && !BC_NUM_NEG_NP(temp))
4414 		{
4415 			assert(BC_NUM_RDX_VALID(d));
4416 			assert(BC_NUM_RDX_VALID_NP(base));
4417 
4418 			bc_num_mul(d, &base, &temp, 0);
4419 
4420 			// We already checked for 0.
4421 			bc_num_rem(&temp, &ctemp, d, 0);
4422 		}
4423 
4424 		assert(BC_NUM_RDX_VALID_NP(base));
4425 
4426 		bc_num_mul(&base, &base, &temp, 0);
4427 
4428 		// We already checked for 0.
4429 		bc_num_rem(&temp, &ctemp, &base, 0);
4430 	}
4431 
4432 err:
4433 	BC_SIG_MAYLOCK;
4434 	bc_num_free(&exp);
4435 	bc_num_free(&temp);
4436 	bc_num_free(&base);
4437 	BC_LONGJMP_CONT(vm);
4438 	assert(!BC_NUM_NEG(d) || d->len);
4439 	assert(BC_NUM_RDX_VALID(d));
4440 	assert(!d->len || d->num[d->len - 1] || BC_NUM_RDX_VAL(d) == d->len);
4441 }
4442 
4443 #if BC_DEBUG_CODE
4444 void
4445 bc_num_printDebug(const BcNum* n, const char* name, bool emptyline)
4446 {
4447 	bc_file_puts(&vm->fout, bc_flush_none, name);
4448 	bc_file_puts(&vm->fout, bc_flush_none, ": ");
4449 	bc_num_printDecimal(n, true);
4450 	bc_file_putchar(&vm->fout, bc_flush_err, '\n');
4451 	if (emptyline) bc_file_putchar(&vm->fout, bc_flush_err, '\n');
4452 	vm->nchars = 0;
4453 }
4454 
4455 void
4456 bc_num_printDigs(const BcDig* n, size_t len, bool emptyline)
4457 {
4458 	size_t i;
4459 
4460 	for (i = len - 1; i < len; --i)
4461 	{
4462 		bc_file_printf(&vm->fout, " %lu", (unsigned long) n[i]);
4463 	}
4464 
4465 	bc_file_putchar(&vm->fout, bc_flush_err, '\n');
4466 	if (emptyline) bc_file_putchar(&vm->fout, bc_flush_err, '\n');
4467 	vm->nchars = 0;
4468 }
4469 
4470 void
4471 bc_num_printWithDigs(const BcNum* n, const char* name, bool emptyline)
4472 {
4473 	bc_file_puts(&vm->fout, bc_flush_none, name);
4474 	bc_file_printf(&vm->fout, " len: %zu, rdx: %zu, scale: %zu\n", name, n->len,
4475 	               BC_NUM_RDX_VAL(n), n->scale);
4476 	bc_num_printDigs(n->num, n->len, emptyline);
4477 }
4478 
4479 void
4480 bc_num_dump(const char* varname, const BcNum* n)
4481 {
4482 	ulong i, scale = n->scale;
4483 
4484 	bc_file_printf(&vm->ferr, "\n%s = %s", varname,
4485 	               n->len ? (BC_NUM_NEG(n) ? "-" : "+") : "0 ");
4486 
4487 	for (i = n->len - 1; i < n->len; --i)
4488 	{
4489 		if (i + 1 == BC_NUM_RDX_VAL(n))
4490 		{
4491 			bc_file_puts(&vm->ferr, bc_flush_none, ". ");
4492 		}
4493 
4494 		if (scale / BC_BASE_DIGS != BC_NUM_RDX_VAL(n) - i - 1)
4495 		{
4496 			bc_file_printf(&vm->ferr, "%lu ", (unsigned long) n->num[i]);
4497 		}
4498 		else
4499 		{
4500 			int mod = scale % BC_BASE_DIGS;
4501 			int d = BC_BASE_DIGS - mod;
4502 			BcDig div;
4503 
4504 			if (mod != 0)
4505 			{
4506 				div = n->num[i] / ((BcDig) bc_num_pow10[(ulong) d]);
4507 				bc_file_printf(&vm->ferr, "%lu", (unsigned long) div);
4508 			}
4509 
4510 			div = n->num[i] % ((BcDig) bc_num_pow10[(ulong) d]);
4511 			bc_file_printf(&vm->ferr, " ' %lu ", (unsigned long) div);
4512 		}
4513 	}
4514 
4515 	bc_file_printf(&vm->ferr, "(%zu | %zu.%zu / %zu) %lu\n", n->scale, n->len,
4516 	               BC_NUM_RDX_VAL(n), n->cap, (unsigned long) (void*) n->num);
4517 
4518 	bc_file_flush(&vm->ferr, bc_flush_err);
4519 }
4520 #endif // BC_DEBUG_CODE
4521