xref: /freebsd/contrib/bc/src/bc_parse.c (revision 924226fba12cc9a228c73b956e1b7fa24c60b055)
1 /*
2  * *****************************************************************************
3  *
4  * SPDX-License-Identifier: BSD-2-Clause
5  *
6  * Copyright (c) 2018-2021 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  * The parser for bc.
33  *
34  */
35 
36 #if BC_ENABLED
37 
38 #include <assert.h>
39 #include <stdbool.h>
40 #include <stdlib.h>
41 #include <string.h>
42 
43 #include <setjmp.h>
44 
45 #include <bc.h>
46 #include <num.h>
47 #include <vm.h>
48 
49 // Before you embark on trying to understand this code, have you read the
50 // Development manual (manuals/development.md) and the comment in include/bc.h
51 // yet? No? Do that first. I'm serious.
52 //
53 // The reason is because this file holds the most sensitive and finicky code in
54 // the entire codebase. Even getting history to work on Windows was nothing
55 // compared to this. This is where dreams go to die, where dragons live, and
56 // from which Ken Thompson himself would flee.
57 
58 static void
59 bc_parse_else(BcParse* p);
60 
61 static void
62 bc_parse_stmt(BcParse* p);
63 
64 static BcParseStatus
65 bc_parse_expr_err(BcParse* p, uint8_t flags, BcParseNext next);
66 
67 static void
68 bc_parse_expr_status(BcParse* p, uint8_t flags, BcParseNext next);
69 
70 /**
71  * Returns true if an instruction could only have come from a "leaf" expression.
72  * For more on what leaf expressions are, read the comment for BC_PARSE_LEAF().
73  * @param t  The instruction to test.
74  */
75 static bool
76 bc_parse_inst_isLeaf(BcInst t)
77 {
78 	return (t >= BC_INST_NUM && t <= BC_INST_MAXSCALE) ||
79 #if BC_ENABLE_EXTRA_MATH
80 	       t == BC_INST_TRUNC ||
81 #endif // BC_ENABLE_EXTRA_MATH
82 	       t <= BC_INST_DEC;
83 }
84 
85 /**
86  * Returns true if the *previous* token was a delimiter. A delimiter is anything
87  * that can legally end a statement. In bc's case, it could be a newline, a
88  * semicolon, and a brace in certain cases.
89  * @param p  The parser.
90  * @return   True if the token is a legal delimiter.
91  */
92 static bool
93 bc_parse_isDelimiter(const BcParse* p)
94 {
95 	BcLexType t = p->l.t;
96 	bool good;
97 
98 	// If it's an obvious delimiter, say so.
99 	if (BC_PARSE_DELIMITER(t)) return true;
100 
101 	good = false;
102 
103 	// If the current token is a keyword, then...beware. That means that we need
104 	// to check for a "dangling" else, where there was no brace-delimited block
105 	// on the previous if.
106 	if (t == BC_LEX_KW_ELSE)
107 	{
108 		size_t i;
109 		uint16_t *fptr = NULL, flags = BC_PARSE_FLAG_ELSE;
110 
111 		// As long as going up the stack is valid for a dangling else, keep on.
112 		for (i = 0; i < p->flags.len && BC_PARSE_BLOCK_STMT(flags); ++i)
113 		{
114 			fptr = bc_vec_item_rev(&p->flags, i);
115 			flags = *fptr;
116 
117 			// If we need a brace and don't have one, then we don't have a
118 			// delimiter.
119 			if ((flags & BC_PARSE_FLAG_BRACE) && p->l.last != BC_LEX_RBRACE)
120 			{
121 				return false;
122 			}
123 		}
124 
125 		// Oh, and we had also better have an if statement somewhere.
126 		good = ((flags & BC_PARSE_FLAG_IF) != 0);
127 	}
128 	else if (t == BC_LEX_RBRACE)
129 	{
130 		size_t i;
131 
132 		// Since we have a brace, we need to just check if a brace was needed.
133 		for (i = 0; !good && i < p->flags.len; ++i)
134 		{
135 			uint16_t* fptr = bc_vec_item_rev(&p->flags, i);
136 			good = (((*fptr) & BC_PARSE_FLAG_BRACE) != 0);
137 		}
138 	}
139 
140 	return good;
141 }
142 
143 /**
144  * Returns true if we are in top level of a function body. The POSIX grammar
145  * is defined such that anything is allowed after a function body, so we must
146  * use this function to detect that case when ending a function body.
147  * @param p  The parser.
148  * @return   True if we are in the top level of parsing a function body.
149  */
150 static bool
151 bc_parse_TopFunc(const BcParse* p)
152 {
153 	bool good = p->flags.len == 2;
154 
155 	uint16_t val = BC_PARSE_FLAG_BRACE | BC_PARSE_FLAG_FUNC_INNER;
156 	val |= BC_PARSE_FLAG_FUNC;
157 
158 	return good && BC_PARSE_TOP_FLAG(p) == val;
159 }
160 
161 /**
162  * Sets a previously defined exit label. What are labels? See the bc Parsing
163  * section of the Development manual (manuals/development.md).
164  * @param p  The parser.
165  */
166 static void
167 bc_parse_setLabel(BcParse* p)
168 {
169 	BcFunc* func = p->func;
170 	BcInstPtr* ip = bc_vec_top(&p->exits);
171 	size_t* label;
172 
173 	assert(func == bc_vec_item(&p->prog->fns, p->fidx));
174 
175 	// Set the preallocated label to the correct index.
176 	label = bc_vec_item(&func->labels, ip->idx);
177 	*label = func->code.len;
178 
179 	// Now, we don't need the exit label; it is done.
180 	bc_vec_pop(&p->exits);
181 }
182 
183 /**
184  * Creates a label and sets it to idx. If this is an exit label, then idx is
185  * actually invalid, but it doesn't matter because it will be fixed by
186  * bc_parse_setLabel() later.
187  * @param p    The parser.
188  * @param idx  The index of the label.
189  */
190 static void
191 bc_parse_createLabel(BcParse* p, size_t idx)
192 {
193 	bc_vec_push(&p->func->labels, &idx);
194 }
195 
196 /**
197  * Creates a conditional label. Unlike an exit label, this label is set at
198  * creation time because it comes *before* the code that will target it.
199  * @param p    The parser.
200  * @param idx  The index of the label.
201  */
202 static void
203 bc_parse_createCondLabel(BcParse* p, size_t idx)
204 {
205 	bc_parse_createLabel(p, p->func->code.len);
206 	bc_vec_push(&p->conds, &idx);
207 }
208 
209 /*
210  * Creates an exit label to be filled in later by bc_parse_setLabel(). Also, why
211  * create a label to be filled in later? Because exit labels are meant to be
212  * targeted by code that comes *before* the label. Since we have to parse that
213  * code first, and don't know how long it will be, we need to just make sure to
214  * reserve a slot to be filled in later when we know.
215  *
216  * By the way, this uses BcInstPtr because it was convenient. The field idx
217  * holds the index, and the field func holds the loop boolean.
218  *
219  * @param p     The parser.
220  * @param idx   The index of the label's position.
221  * @param loop  True if the exit label is for a loop or not.
222  */
223 static void
224 bc_parse_createExitLabel(BcParse* p, size_t idx, bool loop)
225 {
226 	BcInstPtr ip;
227 
228 	assert(p->func == bc_vec_item(&p->prog->fns, p->fidx));
229 
230 	ip.func = loop;
231 	ip.idx = idx;
232 	ip.len = 0;
233 
234 	bc_vec_push(&p->exits, &ip);
235 	bc_parse_createLabel(p, SIZE_MAX);
236 }
237 
238 /**
239  * Pops the correct operators off of the operator stack based on the current
240  * operator. This is because of the Shunting-Yard algorithm. Lower prec means
241  * higher precedence.
242  * @param p       The parser.
243  * @param type    The operator.
244  * @param start   The previous start of the operator stack. For more
245  *                information, see the bc Parsing section of the Development
246  *                manual (manuals/development.md).
247  * @param nexprs  A pointer to the current number of expressions that have not
248  *                been consumed yet. This is an IN and OUT parameter.
249  */
250 static void
251 bc_parse_operator(BcParse* p, BcLexType type, size_t start, size_t* nexprs)
252 {
253 	BcLexType t;
254 	uchar l, r = BC_PARSE_OP_PREC(type);
255 	uchar left = BC_PARSE_OP_LEFT(type);
256 
257 	// While we haven't hit the stop point yet.
258 	while (p->ops.len > start)
259 	{
260 		// Get the top operator.
261 		t = BC_PARSE_TOP_OP(p);
262 
263 		// If it's a right paren, we have reached the end of whatever expression
264 		// this is no matter what.
265 		if (t == BC_LEX_LPAREN) break;
266 
267 		// Break for precedence. Precedence operates differently on left and
268 		// right associativity, by the way. A left associative operator that
269 		// matches the current precedence should take priority, but a right
270 		// associative operator should not.
271 		l = BC_PARSE_OP_PREC(t);
272 		if (l >= r && (l != r || !left)) break;
273 
274 		// Do the housekeeping. In particular, make sure to note that one
275 		// expression was consumed. (Two were, but another was added.)
276 		bc_parse_push(p, BC_PARSE_TOKEN_INST(t));
277 		bc_vec_pop(&p->ops);
278 		*nexprs -= !BC_PARSE_OP_PREFIX(t);
279 	}
280 
281 	bc_vec_push(&p->ops, &type);
282 }
283 
284 /**
285  * Parses a right paren. In the Shunting-Yard algorithm, it needs to be put on
286  * the operator stack. But before that, it needs to consume whatever operators
287  * there are until it hits a left paren.
288  * @param p       The parser.
289  * @param nexprs  A pointer to the current number of expressions that have not
290  *                been consumed yet. This is an IN and OUT parameter.
291  */
292 static void
293 bc_parse_rightParen(BcParse* p, size_t* nexprs)
294 {
295 	BcLexType top;
296 
297 	// Consume operators until a left paren.
298 	while ((top = BC_PARSE_TOP_OP(p)) != BC_LEX_LPAREN)
299 	{
300 		bc_parse_push(p, BC_PARSE_TOKEN_INST(top));
301 		bc_vec_pop(&p->ops);
302 		*nexprs -= !BC_PARSE_OP_PREFIX(top);
303 	}
304 
305 	// We need to pop the left paren as well.
306 	bc_vec_pop(&p->ops);
307 
308 	// Oh, and we also want the next token.
309 	bc_lex_next(&p->l);
310 }
311 
312 /**
313  * Parses function arguments.
314  * @param p      The parser.
315  * @param flags  Flags restricting what kind of expressions the arguments can
316  *               be.
317  */
318 static void
319 bc_parse_args(BcParse* p, uint8_t flags)
320 {
321 	bool comma = false;
322 	size_t nargs;
323 
324 	bc_lex_next(&p->l);
325 
326 	// Print and comparison operators not allowed. Well, comparison operators
327 	// only for POSIX. But we do allow arrays, and we *must* get a value.
328 	flags &= ~(BC_PARSE_PRINT | BC_PARSE_REL);
329 	flags |= (BC_PARSE_ARRAY | BC_PARSE_NEEDVAL);
330 
331 	// Count the arguments and parse them.
332 	for (nargs = 0; p->l.t != BC_LEX_RPAREN; ++nargs)
333 	{
334 		bc_parse_expr_status(p, flags, bc_parse_next_arg);
335 
336 		comma = (p->l.t == BC_LEX_COMMA);
337 		if (comma) bc_lex_next(&p->l);
338 	}
339 
340 	// An ending comma is FAIL.
341 	if (BC_ERR(comma)) bc_parse_err(p, BC_ERR_PARSE_TOKEN);
342 
343 	// Now do the call with the number of arguments.
344 	bc_parse_push(p, BC_INST_CALL);
345 	bc_parse_pushIndex(p, nargs);
346 }
347 
348 /**
349  * Parses a function call.
350  * @param p      The parser.
351  * @param flags  Flags restricting what kind of expressions the arguments can
352  *               be.
353  */
354 static void
355 bc_parse_call(BcParse* p, const char* name, uint8_t flags)
356 {
357 	size_t idx;
358 
359 	bc_parse_args(p, flags);
360 
361 	// We just assert this because bc_parse_args() should
362 	// ensure that the next token is what it should be.
363 	assert(p->l.t == BC_LEX_RPAREN);
364 
365 	// We cannot use bc_program_insertFunc() here
366 	// because it will overwrite an existing function.
367 	idx = bc_map_index(&p->prog->fn_map, name);
368 
369 	// The function does not exist yet. Create a space for it. If the user does
370 	// not define it, it's a *runtime* error, not a parse error.
371 	if (idx == BC_VEC_INVALID_IDX)
372 	{
373 		idx = bc_program_insertFunc(p->prog, name);
374 
375 		assert(idx != BC_VEC_INVALID_IDX);
376 
377 		// Make sure that this pointer was not invalidated.
378 		p->func = bc_vec_item(&p->prog->fns, p->fidx);
379 	}
380 	// The function exists, so set the right function index.
381 	else idx = ((BcId*) bc_vec_item(&p->prog->fn_map, idx))->idx;
382 
383 	bc_parse_pushIndex(p, idx);
384 
385 	// Make sure to get the next token.
386 	bc_lex_next(&p->l);
387 }
388 
389 /**
390  * Parses a name/identifier-based expression. It could be a variable, an array
391  * element, an array itself (for function arguments), a function call, etc.
392  *
393  */
394 static void
395 bc_parse_name(BcParse* p, BcInst* type, bool* can_assign, uint8_t flags)
396 {
397 	char* name;
398 
399 	BC_SIG_ASSERT_LOCKED;
400 
401 	// We want a copy of the name since the lexer might overwrite its copy.
402 	name = bc_vm_strdup(p->l.str.v);
403 
404 	BC_SETJMP_LOCKED(err);
405 
406 	// We need the next token to see if it's just a variable or something more.
407 	bc_lex_next(&p->l);
408 
409 	// Array element or array.
410 	if (p->l.t == BC_LEX_LBRACKET)
411 	{
412 		bc_lex_next(&p->l);
413 
414 		// Array only. This has to be a function parameter.
415 		if (p->l.t == BC_LEX_RBRACKET)
416 		{
417 			// Error if arrays are not allowed.
418 			if (BC_ERR(!(flags & BC_PARSE_ARRAY)))
419 			{
420 				bc_parse_err(p, BC_ERR_PARSE_EXPR);
421 			}
422 
423 			*type = BC_INST_ARRAY;
424 			*can_assign = false;
425 		}
426 		else
427 		{
428 			// If we are here, we have an array element. We need to set the
429 			// expression parsing flags.
430 			uint8_t flags2 = (flags & ~(BC_PARSE_PRINT | BC_PARSE_REL)) |
431 			                 BC_PARSE_NEEDVAL;
432 
433 			bc_parse_expr_status(p, flags2, bc_parse_next_elem);
434 
435 			// The next token *must* be a right bracket.
436 			if (BC_ERR(p->l.t != BC_LEX_RBRACKET))
437 			{
438 				bc_parse_err(p, BC_ERR_PARSE_TOKEN);
439 			}
440 
441 			*type = BC_INST_ARRAY_ELEM;
442 			*can_assign = true;
443 		}
444 
445 		// Make sure to get the next token.
446 		bc_lex_next(&p->l);
447 
448 		// Push the instruction and the name of the identifier.
449 		bc_parse_push(p, *type);
450 		bc_parse_pushName(p, name, false);
451 	}
452 	else if (p->l.t == BC_LEX_LPAREN)
453 	{
454 		// We are parsing a function call; error if not allowed.
455 		if (BC_ERR(flags & BC_PARSE_NOCALL))
456 		{
457 			bc_parse_err(p, BC_ERR_PARSE_TOKEN);
458 		}
459 
460 		*type = BC_INST_CALL;
461 		*can_assign = false;
462 
463 		bc_parse_call(p, name, flags);
464 	}
465 	else
466 	{
467 		// Just a variable.
468 		*type = BC_INST_VAR;
469 		*can_assign = true;
470 		bc_parse_push(p, BC_INST_VAR);
471 		bc_parse_pushName(p, name, true);
472 	}
473 
474 err:
475 	// Need to make sure to unallocate the name.
476 	free(name);
477 	BC_LONGJMP_CONT;
478 	BC_SIG_MAYLOCK;
479 }
480 
481 /**
482  * Parses a builtin function that takes no arguments. This includes read(),
483  * rand(), maxibase(), maxobase(), maxscale(), and maxrand().
484  * @param p     The parser.
485  * @param inst  The instruction corresponding to the builtin.
486  */
487 static void
488 bc_parse_noArgBuiltin(BcParse* p, BcInst inst)
489 {
490 	// Must have a left paren.
491 	bc_lex_next(&p->l);
492 	if (BC_ERR(p->l.t != BC_LEX_LPAREN)) bc_parse_err(p, BC_ERR_PARSE_TOKEN);
493 
494 	// Must have a right paren.
495 	bc_lex_next(&p->l);
496 	if ((p->l.t != BC_LEX_RPAREN)) bc_parse_err(p, BC_ERR_PARSE_TOKEN);
497 
498 	bc_parse_push(p, inst);
499 
500 	bc_lex_next(&p->l);
501 }
502 
503 /**
504  * Parses a builtin function that takes 1 argument. This includes length(),
505  * sqrt(), abs(), scale(), and irand().
506  * @param p      The parser.
507  * @param type   The lex token.
508  * @param flags  The expression parsing flags for parsing the argument.
509  * @param prev   An out parameter; the previous instruction pointer.
510  */
511 static void
512 bc_parse_builtin(BcParse* p, BcLexType type, uint8_t flags, BcInst* prev)
513 {
514 	// Must have a left paren.
515 	bc_lex_next(&p->l);
516 	if (BC_ERR(p->l.t != BC_LEX_LPAREN)) bc_parse_err(p, BC_ERR_PARSE_TOKEN);
517 
518 	bc_lex_next(&p->l);
519 
520 	// Change the flags as needed for parsing the argument.
521 	flags &= ~(BC_PARSE_PRINT | BC_PARSE_REL);
522 	flags |= BC_PARSE_NEEDVAL;
523 
524 	// Since length can take arrays, we need to specially add that flag.
525 	if (type == BC_LEX_KW_LENGTH) flags |= BC_PARSE_ARRAY;
526 
527 	bc_parse_expr_status(p, flags, bc_parse_next_rel);
528 
529 	// Must have a right paren.
530 	if (BC_ERR(p->l.t != BC_LEX_RPAREN)) bc_parse_err(p, BC_ERR_PARSE_TOKEN);
531 
532 	// Adjust previous based on the token and push it.
533 	*prev = type - BC_LEX_KW_LENGTH + BC_INST_LENGTH;
534 	bc_parse_push(p, *prev);
535 
536 	bc_lex_next(&p->l);
537 }
538 
539 /**
540  * Parses a builtin function that takes 3 arguments. This includes modexp() and
541  * divmod().
542  */
543 static void
544 bc_parse_builtin3(BcParse* p, BcLexType type, uint8_t flags, BcInst* prev)
545 {
546 	assert(type == BC_LEX_KW_MODEXP || type == BC_LEX_KW_DIVMOD);
547 
548 	// Must have a left paren.
549 	bc_lex_next(&p->l);
550 	if (BC_ERR(p->l.t != BC_LEX_LPAREN)) bc_parse_err(p, BC_ERR_PARSE_TOKEN);
551 
552 	bc_lex_next(&p->l);
553 
554 	// Change the flags as needed for parsing the argument.
555 	flags &= ~(BC_PARSE_PRINT | BC_PARSE_REL);
556 	flags |= BC_PARSE_NEEDVAL;
557 
558 	bc_parse_expr_status(p, flags, bc_parse_next_builtin);
559 
560 	// Must have a comma.
561 	if (BC_ERR(p->l.t != BC_LEX_COMMA)) bc_parse_err(p, BC_ERR_PARSE_TOKEN);
562 
563 	bc_lex_next(&p->l);
564 
565 	bc_parse_expr_status(p, flags, bc_parse_next_builtin);
566 
567 	// Must have a comma.
568 	if (BC_ERR(p->l.t != BC_LEX_COMMA)) bc_parse_err(p, BC_ERR_PARSE_TOKEN);
569 
570 	bc_lex_next(&p->l);
571 
572 	// If it is a divmod, parse an array name. Otherwise, just parse another
573 	// expression.
574 	if (type == BC_LEX_KW_DIVMOD)
575 	{
576 		// Must have a name.
577 		if (BC_ERR(p->l.t != BC_LEX_NAME)) bc_parse_err(p, BC_ERR_PARSE_TOKEN);
578 
579 		// This is safe because the next token should not overwrite the name.
580 		bc_lex_next(&p->l);
581 
582 		// Must have a left bracket.
583 		if (BC_ERR(p->l.t != BC_LEX_LBRACKET))
584 		{
585 			bc_parse_err(p, BC_ERR_PARSE_TOKEN);
586 		}
587 
588 		// This is safe because the next token should not overwrite the name.
589 		bc_lex_next(&p->l);
590 
591 		// Must have a right bracket.
592 		if (BC_ERR(p->l.t != BC_LEX_RBRACKET))
593 		{
594 			bc_parse_err(p, BC_ERR_PARSE_TOKEN);
595 		}
596 
597 		// This is safe because the next token should not overwrite the name.
598 		bc_lex_next(&p->l);
599 	}
600 	else bc_parse_expr_status(p, flags, bc_parse_next_rel);
601 
602 	// Must have a right paren.
603 	if (BC_ERR(p->l.t != BC_LEX_RPAREN)) bc_parse_err(p, BC_ERR_PARSE_TOKEN);
604 
605 	// Adjust previous based on the token and push it.
606 	*prev = type - BC_LEX_KW_MODEXP + BC_INST_MODEXP;
607 	bc_parse_push(p, *prev);
608 
609 	// If we have divmod, we need to assign the modulus to the array element, so
610 	// we need to push the instructions for doing so.
611 	if (type == BC_LEX_KW_DIVMOD)
612 	{
613 		// The zeroth element.
614 		bc_parse_push(p, BC_INST_ZERO);
615 		bc_parse_push(p, BC_INST_ARRAY_ELEM);
616 
617 		// Push the array.
618 		bc_parse_pushName(p, p->l.str.v, false);
619 
620 		// Swap them and assign. After this, the top item on the stack should
621 		// be the quotient.
622 		bc_parse_push(p, BC_INST_SWAP);
623 		bc_parse_push(p, BC_INST_ASSIGN_NO_VAL);
624 	}
625 
626 	bc_lex_next(&p->l);
627 }
628 
629 /**
630  * Parses the scale keyword. This is special because scale can be a value or a
631  * builtin function.
632  * @param p           The parser.
633  * @param type        An out parameter; the instruction for the parse.
634  * @param can_assign  An out parameter; whether the expression can be assigned
635  *                    to.
636  * @param flags       The expression parsing flags for parsing a scale() arg.
637  */
638 static void
639 bc_parse_scale(BcParse* p, BcInst* type, bool* can_assign, uint8_t flags)
640 {
641 	bc_lex_next(&p->l);
642 
643 	// Without the left paren, it's just the keyword.
644 	if (p->l.t != BC_LEX_LPAREN)
645 	{
646 		// Set, push, and return.
647 		*type = BC_INST_SCALE;
648 		*can_assign = true;
649 		bc_parse_push(p, BC_INST_SCALE);
650 		return;
651 	}
652 
653 	// Handle the scale function.
654 	*type = BC_INST_SCALE_FUNC;
655 	*can_assign = false;
656 
657 	// Once again, adjust the flags.
658 	flags &= ~(BC_PARSE_PRINT | BC_PARSE_REL);
659 	flags |= BC_PARSE_NEEDVAL;
660 
661 	bc_lex_next(&p->l);
662 
663 	bc_parse_expr_status(p, flags, bc_parse_next_rel);
664 
665 	// Must have a right paren.
666 	if (BC_ERR(p->l.t != BC_LEX_RPAREN)) bc_parse_err(p, BC_ERR_PARSE_TOKEN);
667 
668 	bc_parse_push(p, BC_INST_SCALE_FUNC);
669 
670 	bc_lex_next(&p->l);
671 }
672 
673 /**
674  * Parses and increment or decrement operator. This is a bit complex.
675  * @param p           The parser.
676  * @param prev        An out parameter; the previous instruction pointer.
677  * @param can_assign  An out parameter; whether the expression can be assigned
678  *                    to.
679  * @param nexs        An in/out parameter; the number of expressions in the
680  *                    parse tree that are not used.
681  * @param flags       The expression parsing flags for parsing a scale() arg.
682  */
683 static void
684 bc_parse_incdec(BcParse* p, BcInst* prev, bool* can_assign, size_t* nexs,
685                 uint8_t flags)
686 {
687 	BcLexType type;
688 	uchar inst;
689 	BcInst etype = *prev;
690 	BcLexType last = p->l.last;
691 
692 	assert(prev != NULL && can_assign != NULL);
693 
694 	// If we can't assign to the previous token, then we have an error.
695 	if (BC_ERR(last == BC_LEX_OP_INC || last == BC_LEX_OP_DEC ||
696 	           last == BC_LEX_RPAREN))
697 	{
698 		bc_parse_err(p, BC_ERR_PARSE_ASSIGN);
699 	}
700 
701 	// Is the previous instruction for a variable?
702 	if (BC_PARSE_INST_VAR(etype))
703 	{
704 		// If so, this is a postfix operator.
705 		if (!*can_assign) bc_parse_err(p, BC_ERR_PARSE_ASSIGN);
706 
707 		// Only postfix uses BC_INST_INC and BC_INST_DEC.
708 		*prev = inst = BC_INST_INC + (p->l.t != BC_LEX_OP_INC);
709 		bc_parse_push(p, inst);
710 		bc_lex_next(&p->l);
711 		*can_assign = false;
712 	}
713 	else
714 	{
715 		// This is a prefix operator. In that case, we just convert it to
716 		// an assignment instruction.
717 		*prev = inst = BC_INST_ASSIGN_PLUS + (p->l.t != BC_LEX_OP_INC);
718 
719 		bc_lex_next(&p->l);
720 		type = p->l.t;
721 
722 		// Because we parse the next part of the expression
723 		// right here, we need to increment this.
724 		*nexs = *nexs + 1;
725 
726 		// Is the next token a normal identifier?
727 		if (type == BC_LEX_NAME)
728 		{
729 			// Parse the name.
730 			uint8_t flags2 = flags & ~BC_PARSE_ARRAY;
731 			bc_parse_name(p, prev, can_assign, flags2 | BC_PARSE_NOCALL);
732 		}
733 		// Is the next token a global?
734 		else if (type >= BC_LEX_KW_LAST && type <= BC_LEX_KW_OBASE)
735 		{
736 			bc_parse_push(p, type - BC_LEX_KW_LAST + BC_INST_LAST);
737 			bc_lex_next(&p->l);
738 		}
739 		// Is the next token specifically scale, which needs special treatment?
740 		else if (BC_NO_ERR(type == BC_LEX_KW_SCALE))
741 		{
742 			bc_lex_next(&p->l);
743 
744 			// Check that scale() was not used.
745 			if (BC_ERR(p->l.t == BC_LEX_LPAREN))
746 			{
747 				bc_parse_err(p, BC_ERR_PARSE_TOKEN);
748 			}
749 			else bc_parse_push(p, BC_INST_SCALE);
750 		}
751 		// Now we know we have an error.
752 		else bc_parse_err(p, BC_ERR_PARSE_TOKEN);
753 
754 		*can_assign = false;
755 
756 		bc_parse_push(p, BC_INST_ONE);
757 		bc_parse_push(p, inst);
758 	}
759 }
760 
761 /**
762  * Parses the minus operator. This needs special treatment because it is either
763  * subtract or negation.
764  * @param p        The parser.
765  * @param prev     An in/out parameter; the previous instruction.
766  * @param ops_bgn  The size of the operator stack.
767  * @param rparen   True if the last token was a right paren.
768  * @param binlast  True if the last token was a binary operator.
769  * @param nexprs   An in/out parameter; the number of unused expressions.
770  */
771 static void
772 bc_parse_minus(BcParse* p, BcInst* prev, size_t ops_bgn, bool rparen,
773                bool binlast, size_t* nexprs)
774 {
775 	BcLexType type;
776 
777 	bc_lex_next(&p->l);
778 
779 	// Figure out if it's a minus or a negation.
780 	type = BC_PARSE_LEAF(*prev, binlast, rparen) ? BC_LEX_OP_MINUS : BC_LEX_NEG;
781 	*prev = BC_PARSE_TOKEN_INST(type);
782 
783 	// We can just push onto the op stack because this is the largest
784 	// precedence operator that gets pushed. Inc/dec does not.
785 	if (type != BC_LEX_OP_MINUS) bc_vec_push(&p->ops, &type);
786 	else bc_parse_operator(p, type, ops_bgn, nexprs);
787 }
788 
789 /**
790  * Parses a string.
791  * @param p     The parser.
792  * @param inst  The instruction corresponding to how the string was found and
793  *              how it should be printed.
794  */
795 static void
796 bc_parse_str(BcParse* p, BcInst inst)
797 {
798 	bc_parse_addString(p);
799 	bc_parse_push(p, inst);
800 	bc_lex_next(&p->l);
801 }
802 
803 /**
804  * Parses a print statement.
805  * @param p  The parser.
806  */
807 static void
808 bc_parse_print(BcParse* p, BcLexType type)
809 {
810 	BcLexType t;
811 	bool comma = false;
812 	BcInst inst = type == BC_LEX_KW_STREAM ? BC_INST_PRINT_STREAM :
813 	                                         BC_INST_PRINT_POP;
814 
815 	bc_lex_next(&p->l);
816 
817 	t = p->l.t;
818 
819 	// A print or stream statement has to have *something*.
820 	if (bc_parse_isDelimiter(p)) bc_parse_err(p, BC_ERR_PARSE_PRINT);
821 
822 	do
823 	{
824 		// If the token is a string, then print it with escapes.
825 		// BC_INST_PRINT_POP plays that role for bc.
826 		if (t == BC_LEX_STR) bc_parse_str(p, inst);
827 		else
828 		{
829 			// We have an actual number; parse and add a print instruction.
830 			bc_parse_expr_status(p, BC_PARSE_NEEDVAL, bc_parse_next_print);
831 			bc_parse_push(p, inst);
832 		}
833 
834 		// Is the next token a comma?
835 		comma = (p->l.t == BC_LEX_COMMA);
836 
837 		// Get the next token if we have a comma.
838 		if (comma) bc_lex_next(&p->l);
839 		else
840 		{
841 			// If we don't have a comma, the statement needs to end.
842 			if (!bc_parse_isDelimiter(p)) bc_parse_err(p, BC_ERR_PARSE_TOKEN);
843 			else break;
844 		}
845 
846 		t = p->l.t;
847 	}
848 	while (true);
849 
850 	// If we have a comma but no token, that's bad.
851 	if (BC_ERR(comma)) bc_parse_err(p, BC_ERR_PARSE_TOKEN);
852 }
853 
854 /**
855  * Parses a return statement.
856  * @param p  The parser.
857  */
858 static void
859 bc_parse_return(BcParse* p)
860 {
861 	BcLexType t;
862 	bool paren;
863 	uchar inst = BC_INST_RET0;
864 
865 	// If we are not in a function, that's an error.
866 	if (BC_ERR(!BC_PARSE_FUNC(p))) bc_parse_err(p, BC_ERR_PARSE_TOKEN);
867 
868 	// If we are in a void function, make sure to return void.
869 	if (p->func->voidfn) inst = BC_INST_RET_VOID;
870 
871 	bc_lex_next(&p->l);
872 
873 	t = p->l.t;
874 	paren = (t == BC_LEX_LPAREN);
875 
876 	// An empty return statement just needs to push the selected instruction.
877 	if (bc_parse_isDelimiter(p)) bc_parse_push(p, inst);
878 	else
879 	{
880 		BcParseStatus s;
881 
882 		// Need to parse the expression whose value will be returned.
883 		s = bc_parse_expr_err(p, BC_PARSE_NEEDVAL, bc_parse_next_expr);
884 
885 		// If the expression was empty, just push the selected instruction.
886 		if (s == BC_PARSE_STATUS_EMPTY_EXPR)
887 		{
888 			bc_parse_push(p, inst);
889 			bc_lex_next(&p->l);
890 		}
891 
892 		// POSIX requires parentheses.
893 		if (!paren || p->l.last != BC_LEX_RPAREN)
894 		{
895 			bc_parse_err(p, BC_ERR_POSIX_RET);
896 		}
897 
898 		// Void functions require an empty expression.
899 		if (BC_ERR(p->func->voidfn))
900 		{
901 			if (s != BC_PARSE_STATUS_EMPTY_EXPR)
902 			{
903 				bc_parse_verr(p, BC_ERR_PARSE_RET_VOID, p->func->name);
904 			}
905 		}
906 		// If we got here, we want to be sure to end the function with a real
907 		// return instruction, just in case.
908 		else bc_parse_push(p, BC_INST_RET);
909 	}
910 }
911 
912 /**
913  * Clears flags that indicate the end of an if statement and its block and sets
914  * the jump location.
915  * @param p  The parser.
916  */
917 static void
918 bc_parse_noElse(BcParse* p)
919 {
920 	uint16_t* flag_ptr = BC_PARSE_TOP_FLAG_PTR(p);
921 	*flag_ptr = (*flag_ptr & ~(BC_PARSE_FLAG_IF_END));
922 	bc_parse_setLabel(p);
923 }
924 
925 /**
926  * Ends (finishes parsing) the body of a control statement or a function.
927  * @param p      The parser.
928  * @param brace  True if the body was ended by a brace, false otherwise.
929  */
930 static void
931 bc_parse_endBody(BcParse* p, bool brace)
932 {
933 	bool has_brace, new_else = false;
934 
935 	// We cannot be ending a body if there are no bodies to end.
936 	if (BC_ERR(p->flags.len <= 1)) bc_parse_err(p, BC_ERR_PARSE_TOKEN);
937 
938 	if (brace)
939 	{
940 		// The brace was already gotten; make sure that the caller did not lie.
941 		// We check for the requirement of braces later.
942 		assert(p->l.t == BC_LEX_RBRACE);
943 
944 		bc_lex_next(&p->l);
945 
946 		// If the next token is not a delimiter, that is a problem.
947 		if (BC_ERR(!bc_parse_isDelimiter(p) && !bc_parse_TopFunc(p)))
948 		{
949 			bc_parse_err(p, BC_ERR_PARSE_TOKEN);
950 		}
951 	}
952 
953 	// Do we have a brace flag?
954 	has_brace = (BC_PARSE_BRACE(p) != 0);
955 
956 	do
957 	{
958 		size_t len = p->flags.len;
959 		bool loop;
960 
961 		// If we have a brace flag but not a brace, that's a problem.
962 		if (has_brace && !brace) bc_parse_err(p, BC_ERR_PARSE_TOKEN);
963 
964 		// Are we inside a loop?
965 		loop = (BC_PARSE_LOOP_INNER(p) != 0);
966 
967 		// If we are ending a loop or an else...
968 		if (loop || BC_PARSE_ELSE(p))
969 		{
970 			// Loops have condition labels that we have to take care of as well.
971 			if (loop)
972 			{
973 				size_t* label = bc_vec_top(&p->conds);
974 
975 				bc_parse_push(p, BC_INST_JUMP);
976 				bc_parse_pushIndex(p, *label);
977 
978 				bc_vec_pop(&p->conds);
979 			}
980 
981 			bc_parse_setLabel(p);
982 			bc_vec_pop(&p->flags);
983 		}
984 		// If we are ending a function...
985 		else if (BC_PARSE_FUNC_INNER(p))
986 		{
987 			BcInst inst = (p->func->voidfn ? BC_INST_RET_VOID : BC_INST_RET0);
988 			bc_parse_push(p, inst);
989 			bc_parse_updateFunc(p, BC_PROG_MAIN);
990 			bc_vec_pop(&p->flags);
991 		}
992 		// If we have a brace flag and not an if statement, we can pop the top
993 		// of the flags stack because they have been taken care of above.
994 		else if (has_brace && !BC_PARSE_IF(p)) bc_vec_pop(&p->flags);
995 
996 		// This needs to be last to parse nested if's properly.
997 		if (BC_PARSE_IF(p) && (len == p->flags.len || !BC_PARSE_BRACE(p)))
998 		{
999 			// Eat newlines.
1000 			while (p->l.t == BC_LEX_NLINE)
1001 			{
1002 				bc_lex_next(&p->l);
1003 			}
1004 
1005 			// *Now* we can pop the flags.
1006 			bc_vec_pop(&p->flags);
1007 
1008 			// If we are allowed non-POSIX stuff...
1009 			if (!BC_S)
1010 			{
1011 				// Have we found yet another dangling else?
1012 				*(BC_PARSE_TOP_FLAG_PTR(p)) |= BC_PARSE_FLAG_IF_END;
1013 				new_else = (p->l.t == BC_LEX_KW_ELSE);
1014 
1015 				// Parse the else or end the if statement body.
1016 				if (new_else) bc_parse_else(p);
1017 				else if (!has_brace && (!BC_PARSE_IF_END(p) || brace))
1018 				{
1019 					bc_parse_noElse(p);
1020 				}
1021 			}
1022 			// POSIX requires us to do the bare minimum only.
1023 			else bc_parse_noElse(p);
1024 		}
1025 
1026 		// If these are both true, we have "used" the braces that we found.
1027 		if (brace && has_brace) brace = false;
1028 	}
1029 	// This condition was perhaps the hardest single part of the parser. If
1030 	// the flags stack does not have enough, we should stop. If we have a
1031 	// new else statement, we should stop. If we do have the end of an if
1032 	// statement and we have eaten the brace, we should stop. If we do have
1033 	// a brace flag, we should stop.
1034 	while (p->flags.len > 1 && !new_else && (!BC_PARSE_IF_END(p) || brace) &&
1035 	       !(has_brace = (BC_PARSE_BRACE(p) != 0)));
1036 
1037 	// If we have a brace, yet no body for it, that's a problem.
1038 	if (BC_ERR(p->flags.len == 1 && brace)) bc_parse_err(p, BC_ERR_PARSE_TOKEN);
1039 	else if (brace && BC_PARSE_BRACE(p))
1040 	{
1041 		// If we make it here, we have a brace and a flag for it.
1042 		uint16_t flags = BC_PARSE_TOP_FLAG(p);
1043 
1044 		// This condition ensure that the *last* body is correctly finished by
1045 		// popping its flags.
1046 		if (!(flags & (BC_PARSE_FLAG_FUNC_INNER | BC_PARSE_FLAG_LOOP_INNER)) &&
1047 		    !(flags & (BC_PARSE_FLAG_IF | BC_PARSE_FLAG_ELSE)) &&
1048 		    !(flags & (BC_PARSE_FLAG_IF_END)))
1049 		{
1050 			bc_vec_pop(&p->flags);
1051 		}
1052 	}
1053 }
1054 
1055 /**
1056  * Starts the body of a control statement or function.
1057  * @param p      The parser.
1058  * @param flags  The current flags (will be edited).
1059  */
1060 static void
1061 bc_parse_startBody(BcParse* p, uint16_t flags)
1062 {
1063 	assert(flags);
1064 	flags |= (BC_PARSE_TOP_FLAG(p) & (BC_PARSE_FLAG_FUNC | BC_PARSE_FLAG_LOOP));
1065 	flags |= BC_PARSE_FLAG_BODY;
1066 	bc_vec_push(&p->flags, &flags);
1067 }
1068 
1069 void
1070 bc_parse_endif(BcParse* p)
1071 {
1072 	size_t i;
1073 	bool good;
1074 
1075 	// Not a problem if this is true.
1076 	if (BC_NO_ERR(!BC_PARSE_NO_EXEC(p))) return;
1077 
1078 	good = true;
1079 
1080 	// Find an instance of a body that needs closing, i.e., a statement that did
1081 	// not have a right brace when it should have.
1082 	for (i = 0; good && i < p->flags.len; ++i)
1083 	{
1084 		uint16_t flag = *((uint16_t*) bc_vec_item(&p->flags, i));
1085 		good = ((flag & BC_PARSE_FLAG_BRACE) != BC_PARSE_FLAG_BRACE);
1086 	}
1087 
1088 	// If we did not find such an instance...
1089 	if (good)
1090 	{
1091 		// We set this to restore it later. We don't want the parser thinking
1092 		// that we are on stdin for this one because it will want more.
1093 		bool is_stdin = vm.is_stdin;
1094 
1095 		vm.is_stdin = false;
1096 
1097 		// End all of the if statements and loops.
1098 		while (p->flags.len > 1 || BC_PARSE_IF_END(p))
1099 		{
1100 			if (BC_PARSE_IF_END(p)) bc_parse_noElse(p);
1101 			if (p->flags.len > 1) bc_parse_endBody(p, false);
1102 		}
1103 
1104 		vm.is_stdin = is_stdin;
1105 	}
1106 	// If we reach here, a block was not properly closed, and we should error.
1107 	else bc_parse_err(&vm.prs, BC_ERR_PARSE_BLOCK);
1108 }
1109 
1110 /**
1111  * Parses an if statement.
1112  * @param p  The parser.
1113  */
1114 static void
1115 bc_parse_if(BcParse* p)
1116 {
1117 	// We are allowed relational operators, and we must have a value.
1118 	size_t idx;
1119 	uint8_t flags = (BC_PARSE_REL | BC_PARSE_NEEDVAL);
1120 
1121 	// Get the left paren and barf if necessary.
1122 	bc_lex_next(&p->l);
1123 	if (BC_ERR(p->l.t != BC_LEX_LPAREN)) bc_parse_err(p, BC_ERR_PARSE_TOKEN);
1124 
1125 	// Parse the condition.
1126 	bc_lex_next(&p->l);
1127 	bc_parse_expr_status(p, flags, bc_parse_next_rel);
1128 
1129 	// Must have a right paren.
1130 	if (BC_ERR(p->l.t != BC_LEX_RPAREN)) bc_parse_err(p, BC_ERR_PARSE_TOKEN);
1131 
1132 	bc_lex_next(&p->l);
1133 
1134 	// Insert the conditional jump instruction.
1135 	bc_parse_push(p, BC_INST_JUMP_ZERO);
1136 
1137 	idx = p->func->labels.len;
1138 
1139 	// Push the index for the instruction and create an exit label for an else
1140 	// statement.
1141 	bc_parse_pushIndex(p, idx);
1142 	bc_parse_createExitLabel(p, idx, false);
1143 
1144 	bc_parse_startBody(p, BC_PARSE_FLAG_IF);
1145 }
1146 
1147 /**
1148  * Parses an else statement.
1149  * @param p  The parser.
1150  */
1151 static void
1152 bc_parse_else(BcParse* p)
1153 {
1154 	size_t idx = p->func->labels.len;
1155 
1156 	// We must be at the end of an if statement.
1157 	if (BC_ERR(!BC_PARSE_IF_END(p))) bc_parse_err(p, BC_ERR_PARSE_TOKEN);
1158 
1159 	// Push an unconditional jump to make bc jump over the else statement if it
1160 	// executed the original if statement.
1161 	bc_parse_push(p, BC_INST_JUMP);
1162 	bc_parse_pushIndex(p, idx);
1163 
1164 	// Clear the else stuff. Yes, that function is misnamed for its use here,
1165 	// but deal with it.
1166 	bc_parse_noElse(p);
1167 
1168 	// Create the exit label and parse the body.
1169 	bc_parse_createExitLabel(p, idx, false);
1170 	bc_parse_startBody(p, BC_PARSE_FLAG_ELSE);
1171 
1172 	bc_lex_next(&p->l);
1173 }
1174 
1175 /**
1176  * Parse a while loop.
1177  * @param p  The parser.
1178  */
1179 static void
1180 bc_parse_while(BcParse* p)
1181 {
1182 	// We are allowed relational operators, and we must have a value.
1183 	size_t idx;
1184 	uint8_t flags = (BC_PARSE_REL | BC_PARSE_NEEDVAL);
1185 
1186 	// Get the left paren and barf if necessary.
1187 	bc_lex_next(&p->l);
1188 	if (BC_ERR(p->l.t != BC_LEX_LPAREN)) bc_parse_err(p, BC_ERR_PARSE_TOKEN);
1189 	bc_lex_next(&p->l);
1190 
1191 	// Create the labels. Loops need both.
1192 	bc_parse_createCondLabel(p, p->func->labels.len);
1193 	idx = p->func->labels.len;
1194 	bc_parse_createExitLabel(p, idx, true);
1195 
1196 	// Parse the actual condition and barf on non-right paren.
1197 	bc_parse_expr_status(p, flags, bc_parse_next_rel);
1198 	if (BC_ERR(p->l.t != BC_LEX_RPAREN)) bc_parse_err(p, BC_ERR_PARSE_TOKEN);
1199 	bc_lex_next(&p->l);
1200 
1201 	// Now we can push the conditional jump and start the body.
1202 	bc_parse_push(p, BC_INST_JUMP_ZERO);
1203 	bc_parse_pushIndex(p, idx);
1204 	bc_parse_startBody(p, BC_PARSE_FLAG_LOOP | BC_PARSE_FLAG_LOOP_INNER);
1205 }
1206 
1207 /**
1208  * Parse a for loop.
1209  * @param p  The parser.
1210  */
1211 static void
1212 bc_parse_for(BcParse* p)
1213 {
1214 	size_t cond_idx, exit_idx, body_idx, update_idx;
1215 
1216 	// Barf on the missing left paren.
1217 	bc_lex_next(&p->l);
1218 	if (BC_ERR(p->l.t != BC_LEX_LPAREN)) bc_parse_err(p, BC_ERR_PARSE_TOKEN);
1219 	bc_lex_next(&p->l);
1220 
1221 	// The first statement can be empty, but if it is, check for error in POSIX
1222 	// mode. Otherwise, parse it.
1223 	if (p->l.t != BC_LEX_SCOLON) bc_parse_expr_status(p, 0, bc_parse_next_for);
1224 	else bc_parse_err(p, BC_ERR_POSIX_FOR);
1225 
1226 	// Must have a semicolon.
1227 	if (BC_ERR(p->l.t != BC_LEX_SCOLON)) bc_parse_err(p, BC_ERR_PARSE_TOKEN);
1228 	bc_lex_next(&p->l);
1229 
1230 	// These are indices for labels. There are so many of them because the end
1231 	// of the loop must unconditionally jump to the update code. Then the update
1232 	// code must unconditionally jump to the condition code. Then the condition
1233 	// code must *conditionally* jump to the exit.
1234 	cond_idx = p->func->labels.len;
1235 	update_idx = cond_idx + 1;
1236 	body_idx = update_idx + 1;
1237 	exit_idx = body_idx + 1;
1238 
1239 	// This creates the condition label.
1240 	bc_parse_createLabel(p, p->func->code.len);
1241 
1242 	// Parse an expression if it exists.
1243 	if (p->l.t != BC_LEX_SCOLON)
1244 	{
1245 		uint8_t flags = (BC_PARSE_REL | BC_PARSE_NEEDVAL);
1246 		bc_parse_expr_status(p, flags, bc_parse_next_for);
1247 	}
1248 	else
1249 	{
1250 		// Set this for the next call to bc_parse_number because an empty
1251 		// condition means that it is an infinite loop, so the condition must be
1252 		// non-zero. This is safe to set because the current token is a
1253 		// semicolon, which has no string requirement.
1254 		bc_vec_string(&p->l.str, sizeof(bc_parse_one) - 1, bc_parse_one);
1255 		bc_parse_number(p);
1256 
1257 		// An empty condition makes POSIX mad.
1258 		bc_parse_err(p, BC_ERR_POSIX_FOR);
1259 	}
1260 
1261 	// Must have a semicolon.
1262 	if (BC_ERR(p->l.t != BC_LEX_SCOLON)) bc_parse_err(p, BC_ERR_PARSE_TOKEN);
1263 	bc_lex_next(&p->l);
1264 
1265 	// Now we can set up the conditional jump to the exit and an unconditional
1266 	// jump to the body right after. The unconditional jump to the body is
1267 	// because there is update code coming right after the condition, so we need
1268 	// to skip it to get to the body.
1269 	bc_parse_push(p, BC_INST_JUMP_ZERO);
1270 	bc_parse_pushIndex(p, exit_idx);
1271 	bc_parse_push(p, BC_INST_JUMP);
1272 	bc_parse_pushIndex(p, body_idx);
1273 
1274 	// Now create the label for the update code.
1275 	bc_parse_createCondLabel(p, update_idx);
1276 
1277 	// Parse if not empty, and if it is, let POSIX yell if necessary.
1278 	if (p->l.t != BC_LEX_RPAREN) bc_parse_expr_status(p, 0, bc_parse_next_rel);
1279 	else bc_parse_err(p, BC_ERR_POSIX_FOR);
1280 
1281 	// Must have a right paren.
1282 	if (BC_ERR(p->l.t != BC_LEX_RPAREN)) bc_parse_err(p, BC_ERR_PARSE_TOKEN);
1283 
1284 	// Set up a jump to the condition right after the update code.
1285 	bc_parse_push(p, BC_INST_JUMP);
1286 	bc_parse_pushIndex(p, cond_idx);
1287 	bc_parse_createLabel(p, p->func->code.len);
1288 
1289 	// Create an exit label for the body and start the body.
1290 	bc_parse_createExitLabel(p, exit_idx, true);
1291 	bc_lex_next(&p->l);
1292 	bc_parse_startBody(p, BC_PARSE_FLAG_LOOP | BC_PARSE_FLAG_LOOP_INNER);
1293 }
1294 
1295 /**
1296  * Parse a statement or token that indicates a loop exit. This includes an
1297  * actual loop exit, the break keyword, or the continue keyword.
1298  * @param p     The parser.
1299  * @param type  The type of exit.
1300  */
1301 static void
1302 bc_parse_loopExit(BcParse* p, BcLexType type)
1303 {
1304 	size_t i;
1305 	BcInstPtr* ip;
1306 
1307 	// Must have a loop. If we don't, that's an error.
1308 	if (BC_ERR(!BC_PARSE_LOOP(p))) bc_parse_err(p, BC_ERR_PARSE_TOKEN);
1309 
1310 	// If we have a break statement...
1311 	if (type == BC_LEX_KW_BREAK)
1312 	{
1313 		// If there are no exits, something went wrong somewhere.
1314 		if (BC_ERR(!p->exits.len)) bc_parse_err(p, BC_ERR_PARSE_TOKEN);
1315 
1316 		// Get the exit.
1317 		i = p->exits.len - 1;
1318 		ip = bc_vec_item(&p->exits, i);
1319 
1320 		// The condition !ip->func is true if the exit is not for a loop, so we
1321 		// need to find the first actual loop exit.
1322 		while (!ip->func && i < p->exits.len)
1323 		{
1324 			ip = bc_vec_item(&p->exits, i);
1325 			i -= 1;
1326 		}
1327 
1328 		// Make sure everything is hunky dory.
1329 		assert(ip != NULL && (i < p->exits.len || ip->func));
1330 
1331 		// Set the index for the exit.
1332 		i = ip->idx;
1333 	}
1334 	// If we have a continue statement or just the loop end, jump to the
1335 	// condition (or update for a foor loop).
1336 	else i = *((size_t*) bc_vec_top(&p->conds));
1337 
1338 	// Add the unconditional jump.
1339 	bc_parse_push(p, BC_INST_JUMP);
1340 	bc_parse_pushIndex(p, i);
1341 
1342 	bc_lex_next(&p->l);
1343 }
1344 
1345 /**
1346  * Parse a function (header).
1347  * @param p  The parser.
1348  */
1349 static void
1350 bc_parse_func(BcParse* p)
1351 {
1352 	bool comma = false, voidfn;
1353 	uint16_t flags;
1354 	size_t idx;
1355 
1356 	bc_lex_next(&p->l);
1357 
1358 	// Must have a name.
1359 	if (BC_ERR(p->l.t != BC_LEX_NAME)) bc_parse_err(p, BC_ERR_PARSE_FUNC);
1360 
1361 	// If the name is "void", and POSIX is not on, mark as void.
1362 	voidfn = (!BC_IS_POSIX && p->l.t == BC_LEX_NAME &&
1363 	          !strcmp(p->l.str.v, "void"));
1364 
1365 	// We can safely do this because the expected token should not overwrite the
1366 	// function name.
1367 	bc_lex_next(&p->l);
1368 
1369 	// If we *don't* have another name, then void is the name of the function.
1370 	voidfn = (voidfn && p->l.t == BC_LEX_NAME);
1371 
1372 	// With a void function, allow POSIX to complain and get a new token.
1373 	if (voidfn)
1374 	{
1375 		bc_parse_err(p, BC_ERR_POSIX_VOID);
1376 
1377 		// We can safely do this because the expected token should not overwrite
1378 		// the function name.
1379 		bc_lex_next(&p->l);
1380 	}
1381 
1382 	// Must have a left paren.
1383 	if (BC_ERR(p->l.t != BC_LEX_LPAREN)) bc_parse_err(p, BC_ERR_PARSE_FUNC);
1384 
1385 	// Make sure the functions map and vector are synchronized.
1386 	assert(p->prog->fns.len == p->prog->fn_map.len);
1387 
1388 	// Insert the function by name into the map and vector.
1389 	idx = bc_program_insertFunc(p->prog, p->l.str.v);
1390 
1391 	// Make sure the insert worked.
1392 	assert(idx);
1393 
1394 	// Update the function pointer and stuff in the parser and set its void.
1395 	bc_parse_updateFunc(p, idx);
1396 	p->func->voidfn = voidfn;
1397 
1398 	bc_lex_next(&p->l);
1399 
1400 	// While we do not have a right paren, we are still parsing arguments.
1401 	while (p->l.t != BC_LEX_RPAREN)
1402 	{
1403 		BcType t = BC_TYPE_VAR;
1404 
1405 		// If we have an asterisk, we are parsing a reference argument.
1406 		if (p->l.t == BC_LEX_OP_MULTIPLY)
1407 		{
1408 			t = BC_TYPE_REF;
1409 			bc_lex_next(&p->l);
1410 
1411 			// Let POSIX complain if necessary.
1412 			bc_parse_err(p, BC_ERR_POSIX_REF);
1413 		}
1414 
1415 		// If we don't have a name, the argument will not have a name. Barf.
1416 		if (BC_ERR(p->l.t != BC_LEX_NAME)) bc_parse_err(p, BC_ERR_PARSE_FUNC);
1417 
1418 		// Increment the number of parameters.
1419 		p->func->nparams += 1;
1420 
1421 		// Copy the string in the lexer so that we can use the lexer again.
1422 		bc_vec_string(&p->buf, p->l.str.len, p->l.str.v);
1423 
1424 		bc_lex_next(&p->l);
1425 
1426 		// We are parsing an array parameter if this is true.
1427 		if (p->l.t == BC_LEX_LBRACKET)
1428 		{
1429 			// Set the array type, unless we are already parsing a reference.
1430 			if (t == BC_TYPE_VAR) t = BC_TYPE_ARRAY;
1431 
1432 			bc_lex_next(&p->l);
1433 
1434 			// The brackets *must* be empty.
1435 			if (BC_ERR(p->l.t != BC_LEX_RBRACKET))
1436 			{
1437 				bc_parse_err(p, BC_ERR_PARSE_FUNC);
1438 			}
1439 
1440 			bc_lex_next(&p->l);
1441 		}
1442 		// If we did *not* get a bracket, but we are expecting a reference, we
1443 		// have a problem.
1444 		else if (BC_ERR(t == BC_TYPE_REF))
1445 		{
1446 			bc_parse_verr(p, BC_ERR_PARSE_REF_VAR, p->buf.v);
1447 		}
1448 
1449 		// Test for comma and get the next token if it exists.
1450 		comma = (p->l.t == BC_LEX_COMMA);
1451 		if (comma) bc_lex_next(&p->l);
1452 
1453 		// Insert the parameter into the function.
1454 		bc_func_insert(p->func, p->prog, p->buf.v, t, p->l.line);
1455 	}
1456 
1457 	// If we have a comma, but no parameter, barf.
1458 	if (BC_ERR(comma)) bc_parse_err(p, BC_ERR_PARSE_FUNC);
1459 
1460 	// Start the body.
1461 	flags = BC_PARSE_FLAG_FUNC | BC_PARSE_FLAG_FUNC_INNER;
1462 	bc_parse_startBody(p, flags);
1463 
1464 	bc_lex_next(&p->l);
1465 
1466 	// POSIX requires that a brace be on the same line as the function header.
1467 	// If we don't have a brace, let POSIX throw an error.
1468 	if (p->l.t != BC_LEX_LBRACE) bc_parse_err(p, BC_ERR_POSIX_BRACE);
1469 }
1470 
1471 /**
1472  * Parse an auto list.
1473  * @param p  The parser.
1474  */
1475 static void
1476 bc_parse_auto(BcParse* p)
1477 {
1478 	bool comma, one;
1479 
1480 	// Error if the auto keyword appeared in the wrong place.
1481 	if (BC_ERR(!p->auto_part)) bc_parse_err(p, BC_ERR_PARSE_TOKEN);
1482 	bc_lex_next(&p->l);
1483 
1484 	p->auto_part = comma = false;
1485 
1486 	// We need at least one variable or array.
1487 	one = (p->l.t == BC_LEX_NAME);
1488 
1489 	// While we have a variable or array.
1490 	while (p->l.t == BC_LEX_NAME)
1491 	{
1492 		BcType t;
1493 
1494 		// Copy the name from the lexer, so we can use it again.
1495 		bc_vec_string(&p->buf, p->l.str.len - 1, p->l.str.v);
1496 
1497 		bc_lex_next(&p->l);
1498 
1499 		// If we are parsing an array...
1500 		if (p->l.t == BC_LEX_LBRACKET)
1501 		{
1502 			t = BC_TYPE_ARRAY;
1503 
1504 			bc_lex_next(&p->l);
1505 
1506 			// The brackets *must* be empty.
1507 			if (BC_ERR(p->l.t != BC_LEX_RBRACKET))
1508 			{
1509 				bc_parse_err(p, BC_ERR_PARSE_FUNC);
1510 			}
1511 
1512 			bc_lex_next(&p->l);
1513 		}
1514 		else t = BC_TYPE_VAR;
1515 
1516 		// Test for comma and get the next token if it exists.
1517 		comma = (p->l.t == BC_LEX_COMMA);
1518 		if (comma) bc_lex_next(&p->l);
1519 
1520 		// Insert the auto into the function.
1521 		bc_func_insert(p->func, p->prog, p->buf.v, t, p->l.line);
1522 	}
1523 
1524 	// If we have a comma, but no auto, barf.
1525 	if (BC_ERR(comma)) bc_parse_err(p, BC_ERR_PARSE_FUNC);
1526 
1527 	// If we don't have any variables or arrays, barf.
1528 	if (BC_ERR(!one)) bc_parse_err(p, BC_ERR_PARSE_NO_AUTO);
1529 
1530 	// The auto statement should be all that's in the statement.
1531 	if (BC_ERR(!bc_parse_isDelimiter(p))) bc_parse_err(p, BC_ERR_PARSE_TOKEN);
1532 }
1533 
1534 /**
1535  * Parses a body.
1536  * @param p      The parser.
1537  * @param brace  True if a brace was encountered, false otherwise.
1538  */
1539 static void
1540 bc_parse_body(BcParse* p, bool brace)
1541 {
1542 	uint16_t* flag_ptr = BC_PARSE_TOP_FLAG_PTR(p);
1543 
1544 	assert(flag_ptr != NULL);
1545 	assert(p->flags.len >= 2);
1546 
1547 	// The body flag is for when we expect a body. We got a body, so clear the
1548 	// flag.
1549 	*flag_ptr &= ~(BC_PARSE_FLAG_BODY);
1550 
1551 	// If we are inside a function, that means we just barely entered it, and
1552 	// we can expect an auto list.
1553 	if (*flag_ptr & BC_PARSE_FLAG_FUNC_INNER)
1554 	{
1555 		// We *must* have a brace in this case.
1556 		if (BC_ERR(!brace)) bc_parse_err(p, BC_ERR_PARSE_TOKEN);
1557 
1558 		p->auto_part = (p->l.t != BC_LEX_KW_AUTO);
1559 
1560 		if (!p->auto_part)
1561 		{
1562 			// Make sure this is true to not get a parse error.
1563 			p->auto_part = true;
1564 
1565 			// Since we already have the auto keyword, parse.
1566 			bc_parse_auto(p);
1567 		}
1568 
1569 		// Eat a newline.
1570 		if (p->l.t == BC_LEX_NLINE) bc_lex_next(&p->l);
1571 	}
1572 	else
1573 	{
1574 		// This is the easy part.
1575 		size_t len = p->flags.len;
1576 
1577 		assert(*flag_ptr);
1578 
1579 		// Parse a statement.
1580 		bc_parse_stmt(p);
1581 
1582 		// This is a very important condition to get right. If there is no
1583 		// brace, and no body flag, and the flags len hasn't shrunk, then we
1584 		// have a body that was not delimited by braces, so we need to end it
1585 		// now, after just one statement.
1586 		if (!brace && !BC_PARSE_BODY(p) && len <= p->flags.len)
1587 		{
1588 			bc_parse_endBody(p, false);
1589 		}
1590 	}
1591 }
1592 
1593 /**
1594  * Parses a statement. This is the entry point for just about everything, except
1595  * function definitions.
1596  * @param p  The parser.
1597  */
1598 static void
1599 bc_parse_stmt(BcParse* p)
1600 {
1601 	size_t len;
1602 	uint16_t flags;
1603 	BcLexType type = p->l.t;
1604 
1605 	// Eat newline.
1606 	if (type == BC_LEX_NLINE)
1607 	{
1608 		bc_lex_next(&p->l);
1609 		return;
1610 	}
1611 
1612 	// Eat auto list.
1613 	if (type == BC_LEX_KW_AUTO)
1614 	{
1615 		bc_parse_auto(p);
1616 		return;
1617 	}
1618 
1619 	// If we reach this point, no auto list is allowed.
1620 	p->auto_part = false;
1621 
1622 	// Everything but an else needs to be taken care of here, but else is
1623 	// special.
1624 	if (type != BC_LEX_KW_ELSE)
1625 	{
1626 		// After an if, no else found.
1627 		if (BC_PARSE_IF_END(p))
1628 		{
1629 			// Clear the expectation for else, end body, and return. Returning
1630 			// gives us a clean slate for parsing again.
1631 			bc_parse_noElse(p);
1632 			if (p->flags.len > 1 && !BC_PARSE_BRACE(p))
1633 			{
1634 				bc_parse_endBody(p, false);
1635 			}
1636 
1637 			return;
1638 		}
1639 		// With a left brace, we are parsing a body.
1640 		else if (type == BC_LEX_LBRACE)
1641 		{
1642 			// We need to start a body if we are not expecting one yet.
1643 			if (!BC_PARSE_BODY(p))
1644 			{
1645 				bc_parse_startBody(p, BC_PARSE_FLAG_BRACE);
1646 				bc_lex_next(&p->l);
1647 			}
1648 			// If we *are* expecting a body, that body should get a brace. This
1649 			// takes care of braces being on a different line than if and loop
1650 			// headers.
1651 			else
1652 			{
1653 				*(BC_PARSE_TOP_FLAG_PTR(p)) |= BC_PARSE_FLAG_BRACE;
1654 				bc_lex_next(&p->l);
1655 				bc_parse_body(p, true);
1656 			}
1657 
1658 			// If we have reached this point, we need to return for a clean
1659 			// slate.
1660 			return;
1661 		}
1662 		// This happens when we are expecting a body and get a single statement,
1663 		// i.e., a body with no braces surrounding it. Returns after for a clean
1664 		// slate.
1665 		else if (BC_PARSE_BODY(p) && !BC_PARSE_BRACE(p))
1666 		{
1667 			bc_parse_body(p, false);
1668 			return;
1669 		}
1670 	}
1671 
1672 	len = p->flags.len;
1673 	flags = BC_PARSE_TOP_FLAG(p);
1674 
1675 	switch (type)
1676 	{
1677 		// All of these are valid for expressions.
1678 		case BC_LEX_OP_INC:
1679 		case BC_LEX_OP_DEC:
1680 		case BC_LEX_OP_MINUS:
1681 		case BC_LEX_OP_BOOL_NOT:
1682 		case BC_LEX_LPAREN:
1683 		case BC_LEX_NAME:
1684 		case BC_LEX_NUMBER:
1685 		case BC_LEX_KW_IBASE:
1686 		case BC_LEX_KW_LAST:
1687 		case BC_LEX_KW_LENGTH:
1688 		case BC_LEX_KW_OBASE:
1689 		case BC_LEX_KW_SCALE:
1690 #if BC_ENABLE_EXTRA_MATH
1691 		case BC_LEX_KW_SEED:
1692 #endif // BC_ENABLE_EXTRA_MATH
1693 		case BC_LEX_KW_SQRT:
1694 		case BC_LEX_KW_ABS:
1695 #if BC_ENABLE_EXTRA_MATH
1696 		case BC_LEX_KW_IRAND:
1697 #endif // BC_ENABLE_EXTRA_MATH
1698 		case BC_LEX_KW_ASCIIFY:
1699 		case BC_LEX_KW_MODEXP:
1700 		case BC_LEX_KW_DIVMOD:
1701 		case BC_LEX_KW_READ:
1702 #if BC_ENABLE_EXTRA_MATH
1703 		case BC_LEX_KW_RAND:
1704 #endif // BC_ENABLE_EXTRA_MATH
1705 		case BC_LEX_KW_MAXIBASE:
1706 		case BC_LEX_KW_MAXOBASE:
1707 		case BC_LEX_KW_MAXSCALE:
1708 #if BC_ENABLE_EXTRA_MATH
1709 		case BC_LEX_KW_MAXRAND:
1710 #endif // BC_ENABLE_EXTRA_MATH
1711 		case BC_LEX_KW_LINE_LENGTH:
1712 		case BC_LEX_KW_GLOBAL_STACKS:
1713 		case BC_LEX_KW_LEADING_ZERO:
1714 		{
1715 			bc_parse_expr_status(p, BC_PARSE_PRINT, bc_parse_next_expr);
1716 			break;
1717 		}
1718 
1719 		case BC_LEX_KW_ELSE:
1720 		{
1721 			bc_parse_else(p);
1722 			break;
1723 		}
1724 
1725 		// Just eat.
1726 		case BC_LEX_SCOLON:
1727 		{
1728 			// Do nothing.
1729 			break;
1730 		}
1731 
1732 		case BC_LEX_RBRACE:
1733 		{
1734 			bc_parse_endBody(p, true);
1735 			break;
1736 		}
1737 
1738 		case BC_LEX_STR:
1739 		{
1740 			bc_parse_str(p, BC_INST_PRINT_STR);
1741 			break;
1742 		}
1743 
1744 		case BC_LEX_KW_BREAK:
1745 		case BC_LEX_KW_CONTINUE:
1746 		{
1747 			bc_parse_loopExit(p, p->l.t);
1748 			break;
1749 		}
1750 
1751 		case BC_LEX_KW_FOR:
1752 		{
1753 			bc_parse_for(p);
1754 			break;
1755 		}
1756 
1757 		case BC_LEX_KW_HALT:
1758 		{
1759 			bc_parse_push(p, BC_INST_HALT);
1760 			bc_lex_next(&p->l);
1761 			break;
1762 		}
1763 
1764 		case BC_LEX_KW_IF:
1765 		{
1766 			bc_parse_if(p);
1767 			break;
1768 		}
1769 
1770 		case BC_LEX_KW_LIMITS:
1771 		{
1772 			// `limits` is a compile-time command, so execute it right away.
1773 			bc_vm_printf("BC_LONG_BIT      = %lu\n", (ulong) BC_LONG_BIT);
1774 			bc_vm_printf("BC_BASE_DIGS     = %lu\n", (ulong) BC_BASE_DIGS);
1775 			bc_vm_printf("BC_BASE_POW      = %lu\n", (ulong) BC_BASE_POW);
1776 			bc_vm_printf("BC_OVERFLOW_MAX  = %lu\n", (ulong) BC_NUM_BIGDIG_MAX);
1777 			bc_vm_printf("\n");
1778 			bc_vm_printf("BC_BASE_MAX      = %lu\n", BC_MAX_OBASE);
1779 			bc_vm_printf("BC_DIM_MAX       = %lu\n", BC_MAX_DIM);
1780 			bc_vm_printf("BC_SCALE_MAX     = %lu\n", BC_MAX_SCALE);
1781 			bc_vm_printf("BC_STRING_MAX    = %lu\n", BC_MAX_STRING);
1782 			bc_vm_printf("BC_NAME_MAX      = %lu\n", BC_MAX_NAME);
1783 			bc_vm_printf("BC_NUM_MAX       = %lu\n", BC_MAX_NUM);
1784 #if BC_ENABLE_EXTRA_MATH
1785 			bc_vm_printf("BC_RAND_MAX      = %lu\n", BC_MAX_RAND);
1786 #endif // BC_ENABLE_EXTRA_MATH
1787 			bc_vm_printf("MAX Exponent     = %lu\n", BC_MAX_EXP);
1788 			bc_vm_printf("Number of vars   = %lu\n", BC_MAX_VARS);
1789 
1790 			bc_lex_next(&p->l);
1791 
1792 			break;
1793 		}
1794 
1795 		case BC_LEX_KW_STREAM:
1796 		case BC_LEX_KW_PRINT:
1797 		{
1798 			bc_parse_print(p, type);
1799 			break;
1800 		}
1801 
1802 		case BC_LEX_KW_QUIT:
1803 		{
1804 			// Quit is a compile-time command. We don't exit directly, so the vm
1805 			// can clean up.
1806 			vm.status = BC_STATUS_QUIT;
1807 			BC_JMP;
1808 			break;
1809 		}
1810 
1811 		case BC_LEX_KW_RETURN:
1812 		{
1813 			bc_parse_return(p);
1814 			break;
1815 		}
1816 
1817 		case BC_LEX_KW_WHILE:
1818 		{
1819 			bc_parse_while(p);
1820 			break;
1821 		}
1822 
1823 		default:
1824 		{
1825 			bc_parse_err(p, BC_ERR_PARSE_TOKEN);
1826 		}
1827 	}
1828 
1829 	// If the flags did not change, we expect a delimiter.
1830 	if (len == p->flags.len && flags == BC_PARSE_TOP_FLAG(p))
1831 	{
1832 		if (BC_ERR(!bc_parse_isDelimiter(p)))
1833 		{
1834 			bc_parse_err(p, BC_ERR_PARSE_TOKEN);
1835 		}
1836 	}
1837 
1838 	// Make sure semicolons are eaten.
1839 	while (p->l.t == BC_LEX_SCOLON)
1840 	{
1841 		bc_lex_next(&p->l);
1842 	}
1843 
1844 	// POSIX's grammar does not allow a function definition after a semicolon
1845 	// without a newline, so check specifically for that case and error if
1846 	// the POSIX standard flag is set.
1847 	if (p->l.last == BC_LEX_SCOLON && p->l.t == BC_LEX_KW_DEFINE && BC_IS_POSIX)
1848 	{
1849 		bc_parse_err(p, BC_ERR_POSIX_FUNC_AFTER_SEMICOLON);
1850 	}
1851 }
1852 
1853 void
1854 bc_parse_parse(BcParse* p)
1855 {
1856 	assert(p);
1857 
1858 	BC_SETJMP_LOCKED(exit);
1859 
1860 	// We should not let an EOF get here unless some partial parse was not
1861 	// completed, in which case, it's the user's fault.
1862 	if (BC_ERR(p->l.t == BC_LEX_EOF)) bc_parse_err(p, BC_ERR_PARSE_EOF);
1863 
1864 	// Functions need special parsing.
1865 	else if (p->l.t == BC_LEX_KW_DEFINE)
1866 	{
1867 		if (BC_ERR(BC_PARSE_NO_EXEC(p)))
1868 		{
1869 			bc_parse_endif(p);
1870 			if (BC_ERR(BC_PARSE_NO_EXEC(p)))
1871 			{
1872 				bc_parse_err(p, BC_ERR_PARSE_TOKEN);
1873 			}
1874 		}
1875 		bc_parse_func(p);
1876 	}
1877 
1878 	// Otherwise, parse a normal statement.
1879 	else bc_parse_stmt(p);
1880 
1881 exit:
1882 
1883 	// We need to reset on error.
1884 	if (BC_ERR(((vm.status && vm.status != BC_STATUS_QUIT) || vm.sig)))
1885 	{
1886 		bc_parse_reset(p);
1887 	}
1888 
1889 	BC_LONGJMP_CONT;
1890 	BC_SIG_MAYLOCK;
1891 }
1892 
1893 /**
1894  * Parse an expression. This is the actual implementation of the Shunting-Yard
1895  * Algorithm.
1896  * @param p      The parser.
1897  * @param flags  The flags for what is valid in the expression.
1898  * @param next   A set of tokens for what is valid *after* the expression.
1899  * @return       A parse status. In some places, an empty expression is an
1900  *               error, and sometimes, it is required. This allows this function
1901  *               to tell the caller if the expression was empty and let the
1902  *               caller handle it.
1903  */
1904 static BcParseStatus
1905 bc_parse_expr_err(BcParse* p, uint8_t flags, BcParseNext next)
1906 {
1907 	BcInst prev = BC_INST_PRINT;
1908 	uchar inst = BC_INST_INVALID;
1909 	BcLexType top, t;
1910 	size_t nexprs, ops_bgn;
1911 	uint32_t i, nparens, nrelops;
1912 	bool pfirst, rprn, done, get_token, assign, bin_last, incdec, can_assign;
1913 
1914 	// One of these *must* be true.
1915 	assert(!(flags & BC_PARSE_PRINT) || !(flags & BC_PARSE_NEEDVAL));
1916 
1917 	// These are set very carefully. In fact, controlling the values of these
1918 	// locals is the biggest part of making this work. ops_bgn especially is
1919 	// important because it marks where the operator stack begins for *this*
1920 	// invocation of this function. That's because bc_parse_expr_err() is
1921 	// recursive (the Shunting-Yard Algorithm is most easily expressed
1922 	// recursively when parsing subexpressions), and each invocation needs to
1923 	// know where to stop.
1924 	//
1925 	// - nparens is the number of left parens without matches.
1926 	// - nrelops is the number of relational operators that appear in the expr.
1927 	// - nexprs is the number of unused expressions.
1928 	// - rprn is a right paren encountered last.
1929 	// - done means the expression has been fully parsed.
1930 	// - get_token is true when a token is needed at the end of an iteration.
1931 	// - assign is true when an assignment statement was parsed last.
1932 	// - incdec is true when the previous operator was an inc or dec operator.
1933 	// - can_assign is true when an assignemnt is valid.
1934 	// - bin_last is true when the previous instruction was a binary operator.
1935 	t = p->l.t;
1936 	pfirst = (p->l.t == BC_LEX_LPAREN);
1937 	nparens = nrelops = 0;
1938 	nexprs = 0;
1939 	ops_bgn = p->ops.len;
1940 	rprn = done = get_token = assign = incdec = can_assign = false;
1941 	bin_last = true;
1942 
1943 	// We want to eat newlines if newlines are not a valid ending token.
1944 	// This is for spacing in things like for loop headers.
1945 	if (!(flags & BC_PARSE_NOREAD))
1946 	{
1947 		while ((t = p->l.t) == BC_LEX_NLINE)
1948 		{
1949 			bc_lex_next(&p->l);
1950 		}
1951 	}
1952 
1953 	// This is the Shunting-Yard algorithm loop.
1954 	for (; !done && BC_PARSE_EXPR(t); t = p->l.t)
1955 	{
1956 		switch (t)
1957 		{
1958 			case BC_LEX_OP_INC:
1959 			case BC_LEX_OP_DEC:
1960 			{
1961 				// These operators can only be used with items that can be
1962 				// assigned to.
1963 				if (BC_ERR(incdec)) bc_parse_err(p, BC_ERR_PARSE_ASSIGN);
1964 
1965 				bc_parse_incdec(p, &prev, &can_assign, &nexprs, flags);
1966 
1967 				rprn = get_token = bin_last = false;
1968 				incdec = true;
1969 				flags &= ~(BC_PARSE_ARRAY);
1970 
1971 				break;
1972 			}
1973 
1974 #if BC_ENABLE_EXTRA_MATH
1975 			case BC_LEX_OP_TRUNC:
1976 			{
1977 				// The previous token must have been a leaf expression, or the
1978 				// operator is in the wrong place.
1979 				if (BC_ERR(!BC_PARSE_LEAF(prev, bin_last, rprn)))
1980 				{
1981 					bc_parse_err(p, BC_ERR_PARSE_TOKEN);
1982 				}
1983 
1984 				// I can just add the instruction because
1985 				// negative will already be taken care of.
1986 				bc_parse_push(p, BC_INST_TRUNC);
1987 
1988 				rprn = can_assign = incdec = false;
1989 				get_token = true;
1990 				flags &= ~(BC_PARSE_ARRAY);
1991 
1992 				break;
1993 			}
1994 #endif // BC_ENABLE_EXTRA_MATH
1995 
1996 			case BC_LEX_OP_MINUS:
1997 			{
1998 				bc_parse_minus(p, &prev, ops_bgn, rprn, bin_last, &nexprs);
1999 
2000 				rprn = get_token = can_assign = false;
2001 
2002 				// This is true if it was a binary operator last.
2003 				bin_last = (prev == BC_INST_MINUS);
2004 				if (bin_last) incdec = false;
2005 
2006 				flags &= ~(BC_PARSE_ARRAY);
2007 
2008 				break;
2009 			}
2010 
2011 			// All of this group, including the fallthrough, is to parse binary
2012 			// operators.
2013 			case BC_LEX_OP_ASSIGN_POWER:
2014 			case BC_LEX_OP_ASSIGN_MULTIPLY:
2015 			case BC_LEX_OP_ASSIGN_DIVIDE:
2016 			case BC_LEX_OP_ASSIGN_MODULUS:
2017 			case BC_LEX_OP_ASSIGN_PLUS:
2018 			case BC_LEX_OP_ASSIGN_MINUS:
2019 #if BC_ENABLE_EXTRA_MATH
2020 			case BC_LEX_OP_ASSIGN_PLACES:
2021 			case BC_LEX_OP_ASSIGN_LSHIFT:
2022 			case BC_LEX_OP_ASSIGN_RSHIFT:
2023 #endif // BC_ENABLE_EXTRA_MATH
2024 			case BC_LEX_OP_ASSIGN:
2025 			{
2026 				// We need to make sure the assignment is valid.
2027 				if (!BC_PARSE_INST_VAR(prev))
2028 				{
2029 					bc_parse_err(p, BC_ERR_PARSE_ASSIGN);
2030 				}
2031 
2032 				// Fallthrough.
2033 				BC_FALLTHROUGH
2034 			}
2035 
2036 			case BC_LEX_OP_POWER:
2037 			case BC_LEX_OP_MULTIPLY:
2038 			case BC_LEX_OP_DIVIDE:
2039 			case BC_LEX_OP_MODULUS:
2040 			case BC_LEX_OP_PLUS:
2041 #if BC_ENABLE_EXTRA_MATH
2042 			case BC_LEX_OP_PLACES:
2043 			case BC_LEX_OP_LSHIFT:
2044 			case BC_LEX_OP_RSHIFT:
2045 #endif // BC_ENABLE_EXTRA_MATH
2046 			case BC_LEX_OP_REL_EQ:
2047 			case BC_LEX_OP_REL_LE:
2048 			case BC_LEX_OP_REL_GE:
2049 			case BC_LEX_OP_REL_NE:
2050 			case BC_LEX_OP_REL_LT:
2051 			case BC_LEX_OP_REL_GT:
2052 			case BC_LEX_OP_BOOL_NOT:
2053 			case BC_LEX_OP_BOOL_OR:
2054 			case BC_LEX_OP_BOOL_AND:
2055 			{
2056 				// This is true if the operator if the token is a prefix
2057 				// operator. This is only for boolean not.
2058 				if (BC_PARSE_OP_PREFIX(t))
2059 				{
2060 					// Prefix operators are only allowed after binary operators
2061 					// or prefix operators.
2062 					if (BC_ERR(!bin_last && !BC_PARSE_OP_PREFIX(p->l.last)))
2063 					{
2064 						bc_parse_err(p, BC_ERR_PARSE_EXPR);
2065 					}
2066 				}
2067 				// If we execute the else, that means we have a binary operator.
2068 				// If the previous operator was a prefix or a binary operator,
2069 				// then a binary operator is not allowed.
2070 				else if (BC_ERR(BC_PARSE_PREV_PREFIX(prev) || bin_last))
2071 				{
2072 					bc_parse_err(p, BC_ERR_PARSE_EXPR);
2073 				}
2074 
2075 				nrelops += (t >= BC_LEX_OP_REL_EQ && t <= BC_LEX_OP_REL_GT);
2076 				prev = BC_PARSE_TOKEN_INST(t);
2077 
2078 				bc_parse_operator(p, t, ops_bgn, &nexprs);
2079 
2080 				rprn = incdec = can_assign = false;
2081 				get_token = true;
2082 				bin_last = !BC_PARSE_OP_PREFIX(t);
2083 				flags &= ~(BC_PARSE_ARRAY);
2084 
2085 				break;
2086 			}
2087 
2088 			case BC_LEX_LPAREN:
2089 			{
2090 				// A left paren is *not* allowed right after a leaf expr.
2091 				if (BC_ERR(BC_PARSE_LEAF(prev, bin_last, rprn)))
2092 				{
2093 					bc_parse_err(p, BC_ERR_PARSE_EXPR);
2094 				}
2095 
2096 				nparens += 1;
2097 				rprn = incdec = can_assign = false;
2098 				get_token = true;
2099 
2100 				// Push the paren onto the operator stack.
2101 				bc_vec_push(&p->ops, &t);
2102 
2103 				break;
2104 			}
2105 
2106 			case BC_LEX_RPAREN:
2107 			{
2108 				// This needs to be a status. The error is handled in
2109 				// bc_parse_expr_status().
2110 				if (BC_ERR(p->l.last == BC_LEX_LPAREN))
2111 				{
2112 					return BC_PARSE_STATUS_EMPTY_EXPR;
2113 				}
2114 
2115 				// The right paren must not come after a prefix or binary
2116 				// operator.
2117 				if (BC_ERR(bin_last || BC_PARSE_PREV_PREFIX(prev)))
2118 				{
2119 					bc_parse_err(p, BC_ERR_PARSE_EXPR);
2120 				}
2121 
2122 				// If there are no parens left, we are done, but we need another
2123 				// token.
2124 				if (!nparens)
2125 				{
2126 					done = true;
2127 					get_token = false;
2128 					break;
2129 				}
2130 
2131 				nparens -= 1;
2132 				rprn = true;
2133 				get_token = bin_last = incdec = false;
2134 
2135 				bc_parse_rightParen(p, &nexprs);
2136 
2137 				break;
2138 			}
2139 
2140 			case BC_LEX_STR:
2141 			{
2142 				// POSIX only allows strings alone.
2143 				if (BC_IS_POSIX) bc_parse_err(p, BC_ERR_POSIX_EXPR_STRING);
2144 
2145 				// A string is a leaf and cannot come right after a leaf.
2146 				if (BC_ERR(BC_PARSE_LEAF(prev, bin_last, rprn)))
2147 				{
2148 					bc_parse_err(p, BC_ERR_PARSE_EXPR);
2149 				}
2150 
2151 				bc_parse_addString(p);
2152 
2153 				get_token = true;
2154 				bin_last = rprn = false;
2155 				nexprs += 1;
2156 
2157 				break;
2158 			}
2159 
2160 			case BC_LEX_NAME:
2161 			{
2162 				// A name is a leaf and cannot come right after a leaf.
2163 				if (BC_ERR(BC_PARSE_LEAF(prev, bin_last, rprn)))
2164 				{
2165 					bc_parse_err(p, BC_ERR_PARSE_EXPR);
2166 				}
2167 
2168 				get_token = bin_last = false;
2169 
2170 				bc_parse_name(p, &prev, &can_assign, flags & ~BC_PARSE_NOCALL);
2171 
2172 				rprn = (prev == BC_INST_CALL);
2173 				nexprs += 1;
2174 				flags &= ~(BC_PARSE_ARRAY);
2175 
2176 				break;
2177 			}
2178 
2179 			case BC_LEX_NUMBER:
2180 			{
2181 				// A number is a leaf and cannot come right after a leaf.
2182 				if (BC_ERR(BC_PARSE_LEAF(prev, bin_last, rprn)))
2183 				{
2184 					bc_parse_err(p, BC_ERR_PARSE_EXPR);
2185 				}
2186 
2187 				// The number instruction is pushed in here.
2188 				bc_parse_number(p);
2189 
2190 				nexprs += 1;
2191 				prev = BC_INST_NUM;
2192 				get_token = true;
2193 				rprn = bin_last = can_assign = false;
2194 				flags &= ~(BC_PARSE_ARRAY);
2195 
2196 				break;
2197 			}
2198 
2199 			case BC_LEX_KW_IBASE:
2200 			case BC_LEX_KW_LAST:
2201 			case BC_LEX_KW_OBASE:
2202 #if BC_ENABLE_EXTRA_MATH
2203 			case BC_LEX_KW_SEED:
2204 #endif // BC_ENABLE_EXTRA_MATH
2205 			{
2206 				// All of these are leaves and cannot come right after a leaf.
2207 				if (BC_ERR(BC_PARSE_LEAF(prev, bin_last, rprn)))
2208 				{
2209 					bc_parse_err(p, BC_ERR_PARSE_EXPR);
2210 				}
2211 
2212 				prev = t - BC_LEX_KW_LAST + BC_INST_LAST;
2213 				bc_parse_push(p, prev);
2214 
2215 				get_token = can_assign = true;
2216 				rprn = bin_last = false;
2217 				nexprs += 1;
2218 				flags &= ~(BC_PARSE_ARRAY);
2219 
2220 				break;
2221 			}
2222 
2223 			case BC_LEX_KW_LENGTH:
2224 			case BC_LEX_KW_SQRT:
2225 			case BC_LEX_KW_ABS:
2226 #if BC_ENABLE_EXTRA_MATH
2227 			case BC_LEX_KW_IRAND:
2228 #endif // BC_ENABLE_EXTRA_MATH
2229 			case BC_LEX_KW_ASCIIFY:
2230 			{
2231 				// All of these are leaves and cannot come right after a leaf.
2232 				if (BC_ERR(BC_PARSE_LEAF(prev, bin_last, rprn)))
2233 				{
2234 					bc_parse_err(p, BC_ERR_PARSE_EXPR);
2235 				}
2236 
2237 				bc_parse_builtin(p, t, flags, &prev);
2238 
2239 				rprn = get_token = bin_last = incdec = can_assign = false;
2240 				nexprs += 1;
2241 				flags &= ~(BC_PARSE_ARRAY);
2242 
2243 				break;
2244 			}
2245 
2246 			case BC_LEX_KW_READ:
2247 #if BC_ENABLE_EXTRA_MATH
2248 			case BC_LEX_KW_RAND:
2249 #endif // BC_ENABLE_EXTRA_MATH
2250 			case BC_LEX_KW_MAXIBASE:
2251 			case BC_LEX_KW_MAXOBASE:
2252 			case BC_LEX_KW_MAXSCALE:
2253 #if BC_ENABLE_EXTRA_MATH
2254 			case BC_LEX_KW_MAXRAND:
2255 #endif // BC_ENABLE_EXTRA_MATH
2256 			case BC_LEX_KW_LINE_LENGTH:
2257 			case BC_LEX_KW_GLOBAL_STACKS:
2258 			case BC_LEX_KW_LEADING_ZERO:
2259 			{
2260 				// All of these are leaves and cannot come right after a leaf.
2261 				if (BC_ERR(BC_PARSE_LEAF(prev, bin_last, rprn)))
2262 				{
2263 					bc_parse_err(p, BC_ERR_PARSE_EXPR);
2264 				}
2265 
2266 				// Error if we have read and it's not allowed.
2267 				else if (t == BC_LEX_KW_READ && BC_ERR(flags & BC_PARSE_NOREAD))
2268 				{
2269 					bc_parse_err(p, BC_ERR_EXEC_REC_READ);
2270 				}
2271 
2272 				prev = t - BC_LEX_KW_READ + BC_INST_READ;
2273 				bc_parse_noArgBuiltin(p, prev);
2274 
2275 				rprn = get_token = bin_last = incdec = can_assign = false;
2276 				nexprs += 1;
2277 				flags &= ~(BC_PARSE_ARRAY);
2278 
2279 				break;
2280 			}
2281 
2282 			case BC_LEX_KW_SCALE:
2283 			{
2284 				// This is a leaf and cannot come right after a leaf.
2285 				if (BC_ERR(BC_PARSE_LEAF(prev, bin_last, rprn)))
2286 				{
2287 					bc_parse_err(p, BC_ERR_PARSE_EXPR);
2288 				}
2289 
2290 				// Scale needs special work because it can be a variable *or* a
2291 				// function.
2292 				bc_parse_scale(p, &prev, &can_assign, flags);
2293 
2294 				rprn = get_token = bin_last = false;
2295 				nexprs += 1;
2296 				flags &= ~(BC_PARSE_ARRAY);
2297 
2298 				break;
2299 			}
2300 
2301 			case BC_LEX_KW_MODEXP:
2302 			case BC_LEX_KW_DIVMOD:
2303 			{
2304 				// This is a leaf and cannot come right after a leaf.
2305 				if (BC_ERR(BC_PARSE_LEAF(prev, bin_last, rprn)))
2306 				{
2307 					bc_parse_err(p, BC_ERR_PARSE_EXPR);
2308 				}
2309 
2310 				bc_parse_builtin3(p, t, flags, &prev);
2311 
2312 				rprn = get_token = bin_last = incdec = can_assign = false;
2313 				nexprs += 1;
2314 				flags &= ~(BC_PARSE_ARRAY);
2315 
2316 				break;
2317 			}
2318 
2319 			default:
2320 			{
2321 #ifndef NDEBUG
2322 				// We should never get here, even in debug builds.
2323 				bc_parse_err(p, BC_ERR_PARSE_TOKEN);
2324 				break;
2325 #endif // NDEBUG
2326 			}
2327 		}
2328 
2329 		if (get_token) bc_lex_next(&p->l);
2330 	}
2331 
2332 	// Now that we have parsed the expression, we need to empty the operator
2333 	// stack.
2334 	while (p->ops.len > ops_bgn)
2335 	{
2336 		top = BC_PARSE_TOP_OP(p);
2337 		assign = top >= BC_LEX_OP_ASSIGN_POWER && top <= BC_LEX_OP_ASSIGN;
2338 
2339 		// There should not be *any* parens on the stack anymore.
2340 		if (BC_ERR(top == BC_LEX_LPAREN || top == BC_LEX_RPAREN))
2341 		{
2342 			bc_parse_err(p, BC_ERR_PARSE_EXPR);
2343 		}
2344 
2345 		bc_parse_push(p, BC_PARSE_TOKEN_INST(top));
2346 
2347 		// Adjust the number of unused expressions.
2348 		nexprs -= !BC_PARSE_OP_PREFIX(top);
2349 		bc_vec_pop(&p->ops);
2350 
2351 		incdec = false;
2352 	}
2353 
2354 	// There must be only one expression at the top.
2355 	if (BC_ERR(nexprs != 1)) bc_parse_err(p, BC_ERR_PARSE_EXPR);
2356 
2357 	// Check that the next token is correct.
2358 	for (i = 0; i < next.len && t != next.tokens[i]; ++i)
2359 	{
2360 		continue;
2361 	}
2362 	if (BC_ERR(i == next.len && !bc_parse_isDelimiter(p)))
2363 	{
2364 		bc_parse_err(p, BC_ERR_PARSE_EXPR);
2365 	}
2366 
2367 	// Check that POSIX would be happy with the number of relational operators.
2368 	if (!(flags & BC_PARSE_REL) && nrelops)
2369 	{
2370 		bc_parse_err(p, BC_ERR_POSIX_REL_POS);
2371 	}
2372 	else if ((flags & BC_PARSE_REL) && nrelops > 1)
2373 	{
2374 		bc_parse_err(p, BC_ERR_POSIX_MULTIREL);
2375 	}
2376 
2377 	// If this is true, then we might be in a situation where we don't print.
2378 	// We would want to have the increment/decrement operator not make an extra
2379 	// copy if it's not necessary.
2380 	if (!(flags & BC_PARSE_NEEDVAL) && !pfirst)
2381 	{
2382 		// We have the easy case if the last operator was an assignment
2383 		// operator.
2384 		if (assign)
2385 		{
2386 			inst = *((uchar*) bc_vec_top(&p->func->code));
2387 			inst += (BC_INST_ASSIGN_POWER_NO_VAL - BC_INST_ASSIGN_POWER);
2388 			incdec = false;
2389 		}
2390 		// If we have an inc/dec operator and we are *not* printing, implement
2391 		// the optimization to get rid of the extra copy.
2392 		else if (incdec && !(flags & BC_PARSE_PRINT))
2393 		{
2394 			inst = *((uchar*) bc_vec_top(&p->func->code));
2395 			incdec = (inst <= BC_INST_DEC);
2396 			inst = BC_INST_ASSIGN_PLUS_NO_VAL +
2397 			       (inst != BC_INST_INC && inst != BC_INST_ASSIGN_PLUS);
2398 		}
2399 
2400 		// This condition allows us to change the previous assignment
2401 		// instruction (which does a copy) for a NO_VAL version, which does not.
2402 		// This condition is set if either of the above if statements ends up
2403 		// being true.
2404 		if (inst >= BC_INST_ASSIGN_POWER_NO_VAL &&
2405 		    inst <= BC_INST_ASSIGN_NO_VAL)
2406 		{
2407 			// Pop the previous assignment instruction and push a new one.
2408 			// Inc/dec needs the extra instruction because it is now a binary
2409 			// operator and needs a second operand.
2410 			bc_vec_pop(&p->func->code);
2411 			if (incdec) bc_parse_push(p, BC_INST_ONE);
2412 			bc_parse_push(p, inst);
2413 		}
2414 	}
2415 
2416 	// If we might have to print...
2417 	if ((flags & BC_PARSE_PRINT))
2418 	{
2419 		// With a paren first or the last operator not being an assignment, we
2420 		// *do* want to print.
2421 		if (pfirst || !assign) bc_parse_push(p, BC_INST_PRINT);
2422 	}
2423 	// We need to make sure to push a pop instruction for assignment statements
2424 	// that will not print. The print will pop, but without it, we need to pop.
2425 	else if (!(flags & BC_PARSE_NEEDVAL) &&
2426 	         (inst < BC_INST_ASSIGN_POWER_NO_VAL ||
2427 	          inst > BC_INST_ASSIGN_NO_VAL))
2428 	{
2429 		bc_parse_push(p, BC_INST_POP);
2430 	}
2431 
2432 	// We want to eat newlines if newlines are not a valid ending token.
2433 	// This is for spacing in things like for loop headers.
2434 	//
2435 	// Yes, this is one case where I reuse a variable for a different purpose;
2436 	// in this case, incdec being true now means that newlines are not valid.
2437 	for (incdec = true, i = 0; i < next.len && incdec; ++i)
2438 	{
2439 		incdec = (next.tokens[i] != BC_LEX_NLINE);
2440 	}
2441 	if (incdec)
2442 	{
2443 		while (p->l.t == BC_LEX_NLINE)
2444 		{
2445 			bc_lex_next(&p->l);
2446 		}
2447 	}
2448 
2449 	return BC_PARSE_STATUS_SUCCESS;
2450 }
2451 
2452 /**
2453  * Parses an expression with bc_parse_expr_err(), but throws an error if it gets
2454  * an empty expression.
2455  * @param p      The parser.
2456  * @param flags  The flags for what is valid in the expression.
2457  * @param next   A set of tokens for what is valid *after* the expression.
2458  */
2459 static void
2460 bc_parse_expr_status(BcParse* p, uint8_t flags, BcParseNext next)
2461 {
2462 	BcParseStatus s = bc_parse_expr_err(p, flags, next);
2463 
2464 	if (BC_ERR(s == BC_PARSE_STATUS_EMPTY_EXPR))
2465 	{
2466 		bc_parse_err(p, BC_ERR_PARSE_EMPTY_EXPR);
2467 	}
2468 }
2469 
2470 void
2471 bc_parse_expr(BcParse* p, uint8_t flags)
2472 {
2473 	assert(p);
2474 	bc_parse_expr_status(p, flags, bc_parse_next_read);
2475 }
2476 #endif // BC_ENABLED
2477