[sanitizer] Improve FreeBSD ASLR detection
[llvm-project.git] / llvm / lib / CodeGen / MIRParser / MILexer.cpp
blob0ca820f160aa4ce742cd1fee29e1c790a9911663
1 //===- MILexer.cpp - Machine instructions lexer implementation ------------===//
2 //
3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4 // See https://llvm.org/LICENSE.txt for license information.
5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6 //
7 //===----------------------------------------------------------------------===//
8 //
9 // This file implements the lexing of machine instructions.
11 //===----------------------------------------------------------------------===//
13 #include "MILexer.h"
14 #include "llvm/ADT/None.h"
15 #include "llvm/ADT/StringExtras.h"
16 #include "llvm/ADT/StringSwitch.h"
17 #include "llvm/ADT/Twine.h"
18 #include <algorithm>
19 #include <cassert>
20 #include <cctype>
21 #include <string>
23 using namespace llvm;
25 namespace {
27 using ErrorCallbackType =
28 function_ref<void(StringRef::iterator Loc, const Twine &)>;
30 /// This class provides a way to iterate and get characters from the source
31 /// string.
32 class Cursor {
33 const char *Ptr = nullptr;
34 const char *End = nullptr;
36 public:
37 Cursor(NoneType) {}
39 explicit Cursor(StringRef Str) {
40 Ptr = Str.data();
41 End = Ptr + Str.size();
44 bool isEOF() const { return Ptr == End; }
46 char peek(int I = 0) const { return End - Ptr <= I ? 0 : Ptr[I]; }
48 void advance(unsigned I = 1) { Ptr += I; }
50 StringRef remaining() const { return StringRef(Ptr, End - Ptr); }
52 StringRef upto(Cursor C) const {
53 assert(C.Ptr >= Ptr && C.Ptr <= End);
54 return StringRef(Ptr, C.Ptr - Ptr);
57 StringRef::iterator location() const { return Ptr; }
59 operator bool() const { return Ptr != nullptr; }
62 } // end anonymous namespace
64 MIToken &MIToken::reset(TokenKind Kind, StringRef Range) {
65 this->Kind = Kind;
66 this->Range = Range;
67 return *this;
70 MIToken &MIToken::setStringValue(StringRef StrVal) {
71 StringValue = StrVal;
72 return *this;
75 MIToken &MIToken::setOwnedStringValue(std::string StrVal) {
76 StringValueStorage = std::move(StrVal);
77 StringValue = StringValueStorage;
78 return *this;
81 MIToken &MIToken::setIntegerValue(APSInt IntVal) {
82 this->IntVal = std::move(IntVal);
83 return *this;
86 /// Skip the leading whitespace characters and return the updated cursor.
87 static Cursor skipWhitespace(Cursor C) {
88 while (isblank(C.peek()))
89 C.advance();
90 return C;
93 static bool isNewlineChar(char C) { return C == '\n' || C == '\r'; }
95 /// Skip a line comment and return the updated cursor.
96 static Cursor skipComment(Cursor C) {
97 if (C.peek() != ';')
98 return C;
99 while (!isNewlineChar(C.peek()) && !C.isEOF())
100 C.advance();
101 return C;
104 /// Machine operands can have comments, enclosed between /* and */.
105 /// This eats up all tokens, including /* and */.
106 static Cursor skipMachineOperandComment(Cursor C) {
107 if (C.peek() != '/' || C.peek(1) != '*')
108 return C;
110 while (C.peek() != '*' || C.peek(1) != '/')
111 C.advance();
113 C.advance();
114 C.advance();
115 return C;
118 /// Return true if the given character satisfies the following regular
119 /// expression: [-a-zA-Z$._0-9]
120 static bool isIdentifierChar(char C) {
121 return isalpha(C) || isdigit(C) || C == '_' || C == '-' || C == '.' ||
122 C == '$';
125 /// Unescapes the given string value.
127 /// Expects the string value to be quoted.
128 static std::string unescapeQuotedString(StringRef Value) {
129 assert(Value.front() == '"' && Value.back() == '"');
130 Cursor C = Cursor(Value.substr(1, Value.size() - 2));
132 std::string Str;
133 Str.reserve(C.remaining().size());
134 while (!C.isEOF()) {
135 char Char = C.peek();
136 if (Char == '\\') {
137 if (C.peek(1) == '\\') {
138 // Two '\' become one
139 Str += '\\';
140 C.advance(2);
141 continue;
143 if (isxdigit(C.peek(1)) && isxdigit(C.peek(2))) {
144 Str += hexDigitValue(C.peek(1)) * 16 + hexDigitValue(C.peek(2));
145 C.advance(3);
146 continue;
149 Str += Char;
150 C.advance();
152 return Str;
155 /// Lex a string constant using the following regular expression: \"[^\"]*\"
156 static Cursor lexStringConstant(Cursor C, ErrorCallbackType ErrorCallback) {
157 assert(C.peek() == '"');
158 for (C.advance(); C.peek() != '"'; C.advance()) {
159 if (C.isEOF() || isNewlineChar(C.peek())) {
160 ErrorCallback(
161 C.location(),
162 "end of machine instruction reached before the closing '\"'");
163 return None;
166 C.advance();
167 return C;
170 static Cursor lexName(Cursor C, MIToken &Token, MIToken::TokenKind Type,
171 unsigned PrefixLength, ErrorCallbackType ErrorCallback) {
172 auto Range = C;
173 C.advance(PrefixLength);
174 if (C.peek() == '"') {
175 if (Cursor R = lexStringConstant(C, ErrorCallback)) {
176 StringRef String = Range.upto(R);
177 Token.reset(Type, String)
178 .setOwnedStringValue(
179 unescapeQuotedString(String.drop_front(PrefixLength)));
180 return R;
182 Token.reset(MIToken::Error, Range.remaining());
183 return Range;
185 while (isIdentifierChar(C.peek()))
186 C.advance();
187 Token.reset(Type, Range.upto(C))
188 .setStringValue(Range.upto(C).drop_front(PrefixLength));
189 return C;
192 static MIToken::TokenKind getIdentifierKind(StringRef Identifier) {
193 return StringSwitch<MIToken::TokenKind>(Identifier)
194 .Case("_", MIToken::underscore)
195 .Case("implicit", MIToken::kw_implicit)
196 .Case("implicit-def", MIToken::kw_implicit_define)
197 .Case("def", MIToken::kw_def)
198 .Case("dead", MIToken::kw_dead)
199 .Case("killed", MIToken::kw_killed)
200 .Case("undef", MIToken::kw_undef)
201 .Case("internal", MIToken::kw_internal)
202 .Case("early-clobber", MIToken::kw_early_clobber)
203 .Case("debug-use", MIToken::kw_debug_use)
204 .Case("renamable", MIToken::kw_renamable)
205 .Case("tied-def", MIToken::kw_tied_def)
206 .Case("frame-setup", MIToken::kw_frame_setup)
207 .Case("frame-destroy", MIToken::kw_frame_destroy)
208 .Case("nnan", MIToken::kw_nnan)
209 .Case("ninf", MIToken::kw_ninf)
210 .Case("nsz", MIToken::kw_nsz)
211 .Case("arcp", MIToken::kw_arcp)
212 .Case("contract", MIToken::kw_contract)
213 .Case("afn", MIToken::kw_afn)
214 .Case("reassoc", MIToken::kw_reassoc)
215 .Case("nuw", MIToken::kw_nuw)
216 .Case("nsw", MIToken::kw_nsw)
217 .Case("exact", MIToken::kw_exact)
218 .Case("nofpexcept", MIToken::kw_nofpexcept)
219 .Case("debug-location", MIToken::kw_debug_location)
220 .Case("debug-instr-number", MIToken::kw_debug_instr_number)
221 .Case("same_value", MIToken::kw_cfi_same_value)
222 .Case("offset", MIToken::kw_cfi_offset)
223 .Case("rel_offset", MIToken::kw_cfi_rel_offset)
224 .Case("def_cfa_register", MIToken::kw_cfi_def_cfa_register)
225 .Case("def_cfa_offset", MIToken::kw_cfi_def_cfa_offset)
226 .Case("adjust_cfa_offset", MIToken::kw_cfi_adjust_cfa_offset)
227 .Case("escape", MIToken::kw_cfi_escape)
228 .Case("def_cfa", MIToken::kw_cfi_def_cfa)
229 .Case("llvm_def_aspace_cfa", MIToken::kw_cfi_llvm_def_aspace_cfa)
230 .Case("remember_state", MIToken::kw_cfi_remember_state)
231 .Case("restore", MIToken::kw_cfi_restore)
232 .Case("restore_state", MIToken::kw_cfi_restore_state)
233 .Case("undefined", MIToken::kw_cfi_undefined)
234 .Case("register", MIToken::kw_cfi_register)
235 .Case("window_save", MIToken::kw_cfi_window_save)
236 .Case("negate_ra_sign_state",
237 MIToken::kw_cfi_aarch64_negate_ra_sign_state)
238 .Case("blockaddress", MIToken::kw_blockaddress)
239 .Case("intrinsic", MIToken::kw_intrinsic)
240 .Case("target-index", MIToken::kw_target_index)
241 .Case("half", MIToken::kw_half)
242 .Case("float", MIToken::kw_float)
243 .Case("double", MIToken::kw_double)
244 .Case("x86_fp80", MIToken::kw_x86_fp80)
245 .Case("fp128", MIToken::kw_fp128)
246 .Case("ppc_fp128", MIToken::kw_ppc_fp128)
247 .Case("target-flags", MIToken::kw_target_flags)
248 .Case("volatile", MIToken::kw_volatile)
249 .Case("non-temporal", MIToken::kw_non_temporal)
250 .Case("dereferenceable", MIToken::kw_dereferenceable)
251 .Case("invariant", MIToken::kw_invariant)
252 .Case("align", MIToken::kw_align)
253 .Case("basealign", MIToken::kw_align)
254 .Case("addrspace", MIToken::kw_addrspace)
255 .Case("stack", MIToken::kw_stack)
256 .Case("got", MIToken::kw_got)
257 .Case("jump-table", MIToken::kw_jump_table)
258 .Case("constant-pool", MIToken::kw_constant_pool)
259 .Case("call-entry", MIToken::kw_call_entry)
260 .Case("custom", MIToken::kw_custom)
261 .Case("liveout", MIToken::kw_liveout)
262 .Case("address-taken", MIToken::kw_address_taken)
263 .Case("landing-pad", MIToken::kw_landing_pad)
264 .Case("inlineasm-br-indirect-target",
265 MIToken::kw_inlineasm_br_indirect_target)
266 .Case("ehfunclet-entry", MIToken::kw_ehfunclet_entry)
267 .Case("liveins", MIToken::kw_liveins)
268 .Case("successors", MIToken::kw_successors)
269 .Case("floatpred", MIToken::kw_floatpred)
270 .Case("intpred", MIToken::kw_intpred)
271 .Case("shufflemask", MIToken::kw_shufflemask)
272 .Case("pre-instr-symbol", MIToken::kw_pre_instr_symbol)
273 .Case("post-instr-symbol", MIToken::kw_post_instr_symbol)
274 .Case("heap-alloc-marker", MIToken::kw_heap_alloc_marker)
275 .Case("bbsections", MIToken::kw_bbsections)
276 .Case("unknown-size", MIToken::kw_unknown_size)
277 .Case("unknown-address", MIToken::kw_unknown_address)
278 .Case("distinct", MIToken::kw_distinct)
279 .Default(MIToken::Identifier);
282 static Cursor maybeLexIdentifier(Cursor C, MIToken &Token) {
283 if (!isalpha(C.peek()) && C.peek() != '_')
284 return None;
285 auto Range = C;
286 while (isIdentifierChar(C.peek()))
287 C.advance();
288 auto Identifier = Range.upto(C);
289 Token.reset(getIdentifierKind(Identifier), Identifier)
290 .setStringValue(Identifier);
291 return C;
294 static Cursor maybeLexMachineBasicBlock(Cursor C, MIToken &Token,
295 ErrorCallbackType ErrorCallback) {
296 bool IsReference = C.remaining().startswith("%bb.");
297 if (!IsReference && !C.remaining().startswith("bb."))
298 return None;
299 auto Range = C;
300 unsigned PrefixLength = IsReference ? 4 : 3;
301 C.advance(PrefixLength); // Skip '%bb.' or 'bb.'
302 if (!isdigit(C.peek())) {
303 Token.reset(MIToken::Error, C.remaining());
304 ErrorCallback(C.location(), "expected a number after '%bb.'");
305 return C;
307 auto NumberRange = C;
308 while (isdigit(C.peek()))
309 C.advance();
310 StringRef Number = NumberRange.upto(C);
311 unsigned StringOffset = PrefixLength + Number.size(); // Drop '%bb.<id>'
312 // TODO: The format bb.<id>.<irname> is supported only when it's not a
313 // reference. Once we deprecate the format where the irname shows up, we
314 // should only lex forward if it is a reference.
315 if (C.peek() == '.') {
316 C.advance(); // Skip '.'
317 ++StringOffset;
318 while (isIdentifierChar(C.peek()))
319 C.advance();
321 Token.reset(IsReference ? MIToken::MachineBasicBlock
322 : MIToken::MachineBasicBlockLabel,
323 Range.upto(C))
324 .setIntegerValue(APSInt(Number))
325 .setStringValue(Range.upto(C).drop_front(StringOffset));
326 return C;
329 static Cursor maybeLexIndex(Cursor C, MIToken &Token, StringRef Rule,
330 MIToken::TokenKind Kind) {
331 if (!C.remaining().startswith(Rule) || !isdigit(C.peek(Rule.size())))
332 return None;
333 auto Range = C;
334 C.advance(Rule.size());
335 auto NumberRange = C;
336 while (isdigit(C.peek()))
337 C.advance();
338 Token.reset(Kind, Range.upto(C)).setIntegerValue(APSInt(NumberRange.upto(C)));
339 return C;
342 static Cursor maybeLexIndexAndName(Cursor C, MIToken &Token, StringRef Rule,
343 MIToken::TokenKind Kind) {
344 if (!C.remaining().startswith(Rule) || !isdigit(C.peek(Rule.size())))
345 return None;
346 auto Range = C;
347 C.advance(Rule.size());
348 auto NumberRange = C;
349 while (isdigit(C.peek()))
350 C.advance();
351 StringRef Number = NumberRange.upto(C);
352 unsigned StringOffset = Rule.size() + Number.size();
353 if (C.peek() == '.') {
354 C.advance();
355 ++StringOffset;
356 while (isIdentifierChar(C.peek()))
357 C.advance();
359 Token.reset(Kind, Range.upto(C))
360 .setIntegerValue(APSInt(Number))
361 .setStringValue(Range.upto(C).drop_front(StringOffset));
362 return C;
365 static Cursor maybeLexJumpTableIndex(Cursor C, MIToken &Token) {
366 return maybeLexIndex(C, Token, "%jump-table.", MIToken::JumpTableIndex);
369 static Cursor maybeLexStackObject(Cursor C, MIToken &Token) {
370 return maybeLexIndexAndName(C, Token, "%stack.", MIToken::StackObject);
373 static Cursor maybeLexFixedStackObject(Cursor C, MIToken &Token) {
374 return maybeLexIndex(C, Token, "%fixed-stack.", MIToken::FixedStackObject);
377 static Cursor maybeLexConstantPoolItem(Cursor C, MIToken &Token) {
378 return maybeLexIndex(C, Token, "%const.", MIToken::ConstantPoolItem);
381 static Cursor maybeLexSubRegisterIndex(Cursor C, MIToken &Token,
382 ErrorCallbackType ErrorCallback) {
383 const StringRef Rule = "%subreg.";
384 if (!C.remaining().startswith(Rule))
385 return None;
386 return lexName(C, Token, MIToken::SubRegisterIndex, Rule.size(),
387 ErrorCallback);
390 static Cursor maybeLexIRBlock(Cursor C, MIToken &Token,
391 ErrorCallbackType ErrorCallback) {
392 const StringRef Rule = "%ir-block.";
393 if (!C.remaining().startswith(Rule))
394 return None;
395 if (isdigit(C.peek(Rule.size())))
396 return maybeLexIndex(C, Token, Rule, MIToken::IRBlock);
397 return lexName(C, Token, MIToken::NamedIRBlock, Rule.size(), ErrorCallback);
400 static Cursor maybeLexIRValue(Cursor C, MIToken &Token,
401 ErrorCallbackType ErrorCallback) {
402 const StringRef Rule = "%ir.";
403 if (!C.remaining().startswith(Rule))
404 return None;
405 if (isdigit(C.peek(Rule.size())))
406 return maybeLexIndex(C, Token, Rule, MIToken::IRValue);
407 return lexName(C, Token, MIToken::NamedIRValue, Rule.size(), ErrorCallback);
410 static Cursor maybeLexStringConstant(Cursor C, MIToken &Token,
411 ErrorCallbackType ErrorCallback) {
412 if (C.peek() != '"')
413 return None;
414 return lexName(C, Token, MIToken::StringConstant, /*PrefixLength=*/0,
415 ErrorCallback);
418 static Cursor lexVirtualRegister(Cursor C, MIToken &Token) {
419 auto Range = C;
420 C.advance(); // Skip '%'
421 auto NumberRange = C;
422 while (isdigit(C.peek()))
423 C.advance();
424 Token.reset(MIToken::VirtualRegister, Range.upto(C))
425 .setIntegerValue(APSInt(NumberRange.upto(C)));
426 return C;
429 /// Returns true for a character allowed in a register name.
430 static bool isRegisterChar(char C) {
431 return isIdentifierChar(C) && C != '.';
434 static Cursor lexNamedVirtualRegister(Cursor C, MIToken &Token) {
435 Cursor Range = C;
436 C.advance(); // Skip '%'
437 while (isRegisterChar(C.peek()))
438 C.advance();
439 Token.reset(MIToken::NamedVirtualRegister, Range.upto(C))
440 .setStringValue(Range.upto(C).drop_front(1)); // Drop the '%'
441 return C;
444 static Cursor maybeLexRegister(Cursor C, MIToken &Token,
445 ErrorCallbackType ErrorCallback) {
446 if (C.peek() != '%' && C.peek() != '$')
447 return None;
449 if (C.peek() == '%') {
450 if (isdigit(C.peek(1)))
451 return lexVirtualRegister(C, Token);
453 if (isRegisterChar(C.peek(1)))
454 return lexNamedVirtualRegister(C, Token);
456 return None;
459 assert(C.peek() == '$');
460 auto Range = C;
461 C.advance(); // Skip '$'
462 while (isRegisterChar(C.peek()))
463 C.advance();
464 Token.reset(MIToken::NamedRegister, Range.upto(C))
465 .setStringValue(Range.upto(C).drop_front(1)); // Drop the '$'
466 return C;
469 static Cursor maybeLexGlobalValue(Cursor C, MIToken &Token,
470 ErrorCallbackType ErrorCallback) {
471 if (C.peek() != '@')
472 return None;
473 if (!isdigit(C.peek(1)))
474 return lexName(C, Token, MIToken::NamedGlobalValue, /*PrefixLength=*/1,
475 ErrorCallback);
476 auto Range = C;
477 C.advance(1); // Skip the '@'
478 auto NumberRange = C;
479 while (isdigit(C.peek()))
480 C.advance();
481 Token.reset(MIToken::GlobalValue, Range.upto(C))
482 .setIntegerValue(APSInt(NumberRange.upto(C)));
483 return C;
486 static Cursor maybeLexExternalSymbol(Cursor C, MIToken &Token,
487 ErrorCallbackType ErrorCallback) {
488 if (C.peek() != '&')
489 return None;
490 return lexName(C, Token, MIToken::ExternalSymbol, /*PrefixLength=*/1,
491 ErrorCallback);
494 static Cursor maybeLexMCSymbol(Cursor C, MIToken &Token,
495 ErrorCallbackType ErrorCallback) {
496 const StringRef Rule = "<mcsymbol ";
497 if (!C.remaining().startswith(Rule))
498 return None;
499 auto Start = C;
500 C.advance(Rule.size());
502 // Try a simple unquoted name.
503 if (C.peek() != '"') {
504 while (isIdentifierChar(C.peek()))
505 C.advance();
506 StringRef String = Start.upto(C).drop_front(Rule.size());
507 if (C.peek() != '>') {
508 ErrorCallback(C.location(),
509 "expected the '<mcsymbol ...' to be closed by a '>'");
510 Token.reset(MIToken::Error, Start.remaining());
511 return Start;
513 C.advance();
515 Token.reset(MIToken::MCSymbol, Start.upto(C)).setStringValue(String);
516 return C;
519 // Otherwise lex out a quoted name.
520 Cursor R = lexStringConstant(C, ErrorCallback);
521 if (!R) {
522 ErrorCallback(C.location(),
523 "unable to parse quoted string from opening quote");
524 Token.reset(MIToken::Error, Start.remaining());
525 return Start;
527 StringRef String = Start.upto(R).drop_front(Rule.size());
528 if (R.peek() != '>') {
529 ErrorCallback(R.location(),
530 "expected the '<mcsymbol ...' to be closed by a '>'");
531 Token.reset(MIToken::Error, Start.remaining());
532 return Start;
534 R.advance();
536 Token.reset(MIToken::MCSymbol, Start.upto(R))
537 .setOwnedStringValue(unescapeQuotedString(String));
538 return R;
541 static bool isValidHexFloatingPointPrefix(char C) {
542 return C == 'H' || C == 'K' || C == 'L' || C == 'M' || C == 'R';
545 static Cursor lexFloatingPointLiteral(Cursor Range, Cursor C, MIToken &Token) {
546 C.advance();
547 // Skip over [0-9]*([eE][-+]?[0-9]+)?
548 while (isdigit(C.peek()))
549 C.advance();
550 if ((C.peek() == 'e' || C.peek() == 'E') &&
551 (isdigit(C.peek(1)) ||
552 ((C.peek(1) == '-' || C.peek(1) == '+') && isdigit(C.peek(2))))) {
553 C.advance(2);
554 while (isdigit(C.peek()))
555 C.advance();
557 Token.reset(MIToken::FloatingPointLiteral, Range.upto(C));
558 return C;
561 static Cursor maybeLexHexadecimalLiteral(Cursor C, MIToken &Token) {
562 if (C.peek() != '0' || (C.peek(1) != 'x' && C.peek(1) != 'X'))
563 return None;
564 Cursor Range = C;
565 C.advance(2);
566 unsigned PrefLen = 2;
567 if (isValidHexFloatingPointPrefix(C.peek())) {
568 C.advance();
569 PrefLen++;
571 while (isxdigit(C.peek()))
572 C.advance();
573 StringRef StrVal = Range.upto(C);
574 if (StrVal.size() <= PrefLen)
575 return None;
576 if (PrefLen == 2)
577 Token.reset(MIToken::HexLiteral, Range.upto(C));
578 else // It must be 3, which means that there was a floating-point prefix.
579 Token.reset(MIToken::FloatingPointLiteral, Range.upto(C));
580 return C;
583 static Cursor maybeLexNumericalLiteral(Cursor C, MIToken &Token) {
584 if (!isdigit(C.peek()) && (C.peek() != '-' || !isdigit(C.peek(1))))
585 return None;
586 auto Range = C;
587 C.advance();
588 while (isdigit(C.peek()))
589 C.advance();
590 if (C.peek() == '.')
591 return lexFloatingPointLiteral(Range, C, Token);
592 StringRef StrVal = Range.upto(C);
593 Token.reset(MIToken::IntegerLiteral, StrVal).setIntegerValue(APSInt(StrVal));
594 return C;
597 static MIToken::TokenKind getMetadataKeywordKind(StringRef Identifier) {
598 return StringSwitch<MIToken::TokenKind>(Identifier)
599 .Case("!tbaa", MIToken::md_tbaa)
600 .Case("!alias.scope", MIToken::md_alias_scope)
601 .Case("!noalias", MIToken::md_noalias)
602 .Case("!range", MIToken::md_range)
603 .Case("!DIExpression", MIToken::md_diexpr)
604 .Case("!DILocation", MIToken::md_dilocation)
605 .Default(MIToken::Error);
608 static Cursor maybeLexExclaim(Cursor C, MIToken &Token,
609 ErrorCallbackType ErrorCallback) {
610 if (C.peek() != '!')
611 return None;
612 auto Range = C;
613 C.advance(1);
614 if (isdigit(C.peek()) || !isIdentifierChar(C.peek())) {
615 Token.reset(MIToken::exclaim, Range.upto(C));
616 return C;
618 while (isIdentifierChar(C.peek()))
619 C.advance();
620 StringRef StrVal = Range.upto(C);
621 Token.reset(getMetadataKeywordKind(StrVal), StrVal);
622 if (Token.isError())
623 ErrorCallback(Token.location(),
624 "use of unknown metadata keyword '" + StrVal + "'");
625 return C;
628 static MIToken::TokenKind symbolToken(char C) {
629 switch (C) {
630 case ',':
631 return MIToken::comma;
632 case '.':
633 return MIToken::dot;
634 case '=':
635 return MIToken::equal;
636 case ':':
637 return MIToken::colon;
638 case '(':
639 return MIToken::lparen;
640 case ')':
641 return MIToken::rparen;
642 case '{':
643 return MIToken::lbrace;
644 case '}':
645 return MIToken::rbrace;
646 case '+':
647 return MIToken::plus;
648 case '-':
649 return MIToken::minus;
650 case '<':
651 return MIToken::less;
652 case '>':
653 return MIToken::greater;
654 default:
655 return MIToken::Error;
659 static Cursor maybeLexSymbol(Cursor C, MIToken &Token) {
660 MIToken::TokenKind Kind;
661 unsigned Length = 1;
662 if (C.peek() == ':' && C.peek(1) == ':') {
663 Kind = MIToken::coloncolon;
664 Length = 2;
665 } else
666 Kind = symbolToken(C.peek());
667 if (Kind == MIToken::Error)
668 return None;
669 auto Range = C;
670 C.advance(Length);
671 Token.reset(Kind, Range.upto(C));
672 return C;
675 static Cursor maybeLexNewline(Cursor C, MIToken &Token) {
676 if (!isNewlineChar(C.peek()))
677 return None;
678 auto Range = C;
679 C.advance();
680 Token.reset(MIToken::Newline, Range.upto(C));
681 return C;
684 static Cursor maybeLexEscapedIRValue(Cursor C, MIToken &Token,
685 ErrorCallbackType ErrorCallback) {
686 if (C.peek() != '`')
687 return None;
688 auto Range = C;
689 C.advance();
690 auto StrRange = C;
691 while (C.peek() != '`') {
692 if (C.isEOF() || isNewlineChar(C.peek())) {
693 ErrorCallback(
694 C.location(),
695 "end of machine instruction reached before the closing '`'");
696 Token.reset(MIToken::Error, Range.remaining());
697 return C;
699 C.advance();
701 StringRef Value = StrRange.upto(C);
702 C.advance();
703 Token.reset(MIToken::QuotedIRValue, Range.upto(C)).setStringValue(Value);
704 return C;
707 StringRef llvm::lexMIToken(StringRef Source, MIToken &Token,
708 ErrorCallbackType ErrorCallback) {
709 auto C = skipComment(skipWhitespace(Cursor(Source)));
710 if (C.isEOF()) {
711 Token.reset(MIToken::Eof, C.remaining());
712 return C.remaining();
715 C = skipMachineOperandComment(C);
717 if (Cursor R = maybeLexMachineBasicBlock(C, Token, ErrorCallback))
718 return R.remaining();
719 if (Cursor R = maybeLexIdentifier(C, Token))
720 return R.remaining();
721 if (Cursor R = maybeLexJumpTableIndex(C, Token))
722 return R.remaining();
723 if (Cursor R = maybeLexStackObject(C, Token))
724 return R.remaining();
725 if (Cursor R = maybeLexFixedStackObject(C, Token))
726 return R.remaining();
727 if (Cursor R = maybeLexConstantPoolItem(C, Token))
728 return R.remaining();
729 if (Cursor R = maybeLexSubRegisterIndex(C, Token, ErrorCallback))
730 return R.remaining();
731 if (Cursor R = maybeLexIRBlock(C, Token, ErrorCallback))
732 return R.remaining();
733 if (Cursor R = maybeLexIRValue(C, Token, ErrorCallback))
734 return R.remaining();
735 if (Cursor R = maybeLexRegister(C, Token, ErrorCallback))
736 return R.remaining();
737 if (Cursor R = maybeLexGlobalValue(C, Token, ErrorCallback))
738 return R.remaining();
739 if (Cursor R = maybeLexExternalSymbol(C, Token, ErrorCallback))
740 return R.remaining();
741 if (Cursor R = maybeLexMCSymbol(C, Token, ErrorCallback))
742 return R.remaining();
743 if (Cursor R = maybeLexHexadecimalLiteral(C, Token))
744 return R.remaining();
745 if (Cursor R = maybeLexNumericalLiteral(C, Token))
746 return R.remaining();
747 if (Cursor R = maybeLexExclaim(C, Token, ErrorCallback))
748 return R.remaining();
749 if (Cursor R = maybeLexSymbol(C, Token))
750 return R.remaining();
751 if (Cursor R = maybeLexNewline(C, Token))
752 return R.remaining();
753 if (Cursor R = maybeLexEscapedIRValue(C, Token, ErrorCallback))
754 return R.remaining();
755 if (Cursor R = maybeLexStringConstant(C, Token, ErrorCallback))
756 return R.remaining();
758 Token.reset(MIToken::Error, C.remaining());
759 ErrorCallback(C.location(),
760 Twine("unexpected character '") + Twine(C.peek()) + "'");
761 return C.remaining();