]>
Commit | Line | Data |
---|---|---|
2f1a1aea | 1 | // recursive descent parser for systemtap scripts |
3e6a17ee | 2 | // Copyright (C) 2005-2013 Red Hat Inc. |
77a5c1f9 | 3 | // Copyright (C) 2006 Intel Corporation. |
5811366a | 4 | // Copyright (C) 2007 Bull S.A.S |
69c68955 FCE |
5 | // |
6 | // This file is part of systemtap, and is free software. You can | |
7 | // redistribute it and/or modify it under the terms of the GNU General | |
8 | // Public License (GPL); either version 2, or (at your option) any | |
9 | // later version. | |
2f1a1aea | 10 | |
2b066ec1 | 11 | #include "config.h" |
2f1a1aea FCE |
12 | #include "staptree.h" |
13 | #include "parse.h" | |
177a8ead | 14 | #include "session.h" |
3f99432c FCE |
15 | #include "util.h" |
16 | ||
2b066ec1 | 17 | #include <iostream> |
eacb10ce | 18 | |
2b066ec1 | 19 | #include <fstream> |
2f1a1aea | 20 | #include <cctype> |
9c0c0e46 | 21 | #include <cstdlib> |
29e64872 | 22 | #include <cassert> |
9c0c0e46 FCE |
23 | #include <cerrno> |
24 | #include <climits> | |
57b73400 | 25 | #include <sstream> |
f74fb737 | 26 | #include <cstring> |
3f99432c | 27 | #include <cctype> |
eacb10ce FCE |
28 | #include <iterator> |
29 | ||
7a468d68 FCE |
30 | extern "C" { |
31 | #include <fnmatch.h> | |
32 | } | |
2f1a1aea FCE |
33 | |
34 | using namespace std; | |
35 | ||
c18f07f8 JS |
36 | |
37 | class lexer | |
38 | { | |
39 | public: | |
fee28e5c | 40 | bool ate_comment; // current token follows a comment |
b5477cd9 | 41 | bool ate_whitespace; // the most recent token followed whitespace |
534aad8b SM |
42 | bool saw_tokens; // the lexer found tokens (before preprocessing occurred) |
43 | ||
b5477cd9 | 44 | token* scan (); |
c18f07f8 JS |
45 | lexer (istream&, const string&, systemtap_session&); |
46 | void set_current_file (stapfile* f); | |
47 | ||
2524d1fd SM |
48 | static set<string> keywords; |
49 | static set<string> atwords; | |
c18f07f8 JS |
50 | private: |
51 | inline int input_get (); | |
52 | inline int input_peek (unsigned n=0); | |
53 | void input_put (const string&, const token*); | |
54 | string input_name; | |
55 | string input_contents; | |
56 | const char *input_pointer; // index into input_contents | |
57 | const char *input_end; | |
58 | unsigned cursor_suspend_count; | |
59 | unsigned cursor_suspend_line; | |
60 | unsigned cursor_suspend_column; | |
61 | unsigned cursor_line; | |
62 | unsigned cursor_column; | |
63 | systemtap_session& session; | |
64 | stapfile* current_file; | |
c18f07f8 JS |
65 | }; |
66 | ||
67 | ||
68 | class parser | |
69 | { | |
70 | public: | |
4cd32d8c | 71 | parser (systemtap_session& s, const string& n, istream& i, bool p); |
c18f07f8 JS |
72 | ~parser (); |
73 | ||
74 | stapfile* parse (); | |
fe410f52 | 75 | stapfile* parse_library_macros (); |
c18f07f8 JS |
76 | |
77 | private: | |
78 | typedef enum { | |
79 | PP_NONE, | |
80 | PP_KEEP_THEN, | |
81 | PP_SKIP_THEN, | |
82 | PP_KEEP_ELSE, | |
83 | PP_SKIP_ELSE, | |
84 | } pp_state_t; | |
85 | ||
534aad8b SM |
86 | struct pp1_activation; |
87 | ||
fe410f52 SM |
88 | struct pp_macrodecl : public macrodecl { |
89 | pp1_activation* parent_act; // used for param bindings | |
90 | virtual bool is_closure() { return parent_act != 0; } | |
91 | pp_macrodecl () : macrodecl(), parent_act(0) { } | |
534aad8b SM |
92 | }; |
93 | ||
c18f07f8 JS |
94 | systemtap_session& session; |
95 | string input_name; | |
c18f07f8 JS |
96 | lexer input; |
97 | bool privileged; | |
98 | parse_context context; | |
99 | ||
534aad8b SM |
100 | // preprocessing subordinate, first pass (macros) |
101 | struct pp1_activation { | |
102 | const token* tok; | |
103 | unsigned cursor; // position within macro body | |
104 | map<string, pp_macrodecl*> params; | |
534aad8b | 105 | |
fe410f52 | 106 | macrodecl* curr_macro; |
534aad8b | 107 | |
fe410f52 SM |
108 | pp1_activation (const token tok, macrodecl* curr_macro) |
109 | : tok(new token(tok)), cursor(0), curr_macro(curr_macro) { } | |
534aad8b SM |
110 | ~pp1_activation (); |
111 | }; | |
112 | ||
fe410f52 | 113 | map<string, macrodecl*> pp1_namespace; |
534aad8b SM |
114 | vector<pp1_activation*> pp1_state; |
115 | const token* next_pp1 (); | |
116 | const token* scan_pp1 (); | |
117 | const token* slurp_pp1_param (vector<const token*>& param); | |
118 | const token* slurp_pp1_body (vector<const token*>& body); | |
119 | ||
120 | // preprocessing subordinate, final pass (conditionals) | |
c18f07f8 | 121 | vector<pair<const token*, pp_state_t> > pp_state; |
b5477cd9 | 122 | const token* scan_pp (); |
c18f07f8 JS |
123 | const token* skip_pp (); |
124 | ||
125 | // scanning state | |
b5477cd9 SM |
126 | const token* next (); |
127 | const token* peek (); | |
c18f07f8 | 128 | |
731a5359 MW |
129 | // Advance past and throw away current token after peek () or next (). |
130 | void swallow (); | |
131 | ||
a07a2c28 | 132 | const token* systemtap_v_seen; |
c18f07f8 JS |
133 | const token* last_t; // the last value returned by peek() or next() |
134 | const token* next_t; // lookahead token | |
135 | ||
731a5359 MW |
136 | // expectations, these swallow the token |
137 | void expect_known (token_type tt, string const & expected); | |
138 | void expect_unknown (token_type tt, string & target); | |
139 | void expect_unknown2 (token_type tt1, token_type tt2, string & target); | |
140 | ||
141 | // convenience forms, these also swallow the token | |
142 | void expect_op (string const & expected); | |
143 | void expect_kw (string const & expected); | |
144 | void expect_number (int64_t & expected); | |
145 | void expect_ident_or_keyword (string & target); | |
146 | ||
147 | // convenience forms, which return true or false, these don't swallow token | |
c18f07f8 JS |
148 | bool peek_op (string const & op); |
149 | bool peek_kw (string const & kw); | |
150 | ||
731a5359 MW |
151 | // convenience forms, which return the token |
152 | const token* expect_kw_token (string const & expected); | |
153 | const token* expect_ident_or_atword (string & target); | |
154 | ||
c18f07f8 JS |
155 | void print_error (const parse_error& pe); |
156 | unsigned num_errors; | |
157 | ||
158 | private: // nonterminals | |
159 | void parse_probe (vector<probe*>&, vector<probe_alias*>&); | |
160 | void parse_global (vector<vardecl*>&, vector<probe*>&); | |
161 | void parse_functiondecl (vector<functiondecl*>&); | |
162 | embeddedcode* parse_embeddedcode (); | |
163 | probe_point* parse_probe_point (); | |
d24f1ff4 SM |
164 | literal_string* consume_string_literals (const token*); |
165 | literal_string* parse_literal_string (); | |
c18f07f8 JS |
166 | literal* parse_literal (); |
167 | block* parse_stmt_block (); | |
168 | try_block* parse_try_block (); | |
169 | statement* parse_statement (); | |
170 | if_statement* parse_if_statement (); | |
171 | for_loop* parse_for_loop (); | |
172 | for_loop* parse_while_loop (); | |
173 | foreach_loop* parse_foreach_loop (); | |
174 | expr_statement* parse_expr_statement (); | |
175 | return_statement* parse_return_statement (); | |
176 | delete_statement* parse_delete_statement (); | |
177 | next_statement* parse_next_statement (); | |
178 | break_statement* parse_break_statement (); | |
179 | continue_statement* parse_continue_statement (); | |
180 | indexable* parse_indexable (); | |
181 | const token *parse_hist_op_or_bare_name (hist_op *&hop, string &name); | |
182 | target_symbol *parse_target_symbol (const token* t); | |
8cc799a5 | 183 | expression* parse_entry_op (const token* t); |
c18f07f8 | 184 | expression* parse_defined_op (const token* t); |
3689db05 | 185 | expression* parse_perf_op (const token* t); |
c18f07f8 JS |
186 | expression* parse_expression (); |
187 | expression* parse_assignment (); | |
188 | expression* parse_ternary (); | |
189 | expression* parse_logical_or (); | |
190 | expression* parse_logical_and (); | |
191 | expression* parse_boolean_or (); | |
192 | expression* parse_boolean_xor (); | |
193 | expression* parse_boolean_and (); | |
194 | expression* parse_array_in (); | |
93daaca8 | 195 | expression* parse_comparison_or_regex_query (); |
c18f07f8 JS |
196 | expression* parse_shift (); |
197 | expression* parse_concatenation (); | |
198 | expression* parse_additive (); | |
199 | expression* parse_multiplicative (); | |
200 | expression* parse_unary (); | |
201 | expression* parse_crement (); | |
202 | expression* parse_value (); | |
203 | expression* parse_symbol (); | |
204 | ||
205 | void parse_target_symbol_components (target_symbol* e); | |
206 | }; | |
207 | ||
208 | ||
2f1a1aea FCE |
209 | // ------------------------------------------------------------------------ |
210 | ||
c18f07f8 JS |
211 | stapfile* |
212 | parse (systemtap_session& s, istream& i, bool pr) | |
213 | { | |
4cd32d8c | 214 | parser p (s, "<input>", i, pr); |
c18f07f8 JS |
215 | return p.parse (); |
216 | } | |
217 | ||
218 | ||
219 | stapfile* | |
4cd32d8c | 220 | parse (systemtap_session& s, const string& name, bool pr) |
c18f07f8 | 221 | { |
4cd32d8c JS |
222 | ifstream i(name.c_str(), ios::in); |
223 | if (i.fail()) | |
224 | { | |
225 | cerr << (file_exists(name) | |
226 | ? _F("Input file '%s' can't be opened for reading.", name.c_str()) | |
227 | : _F("Input file '%s' is missing.", name.c_str())) | |
228 | << endl; | |
229 | return 0; | |
230 | } | |
231 | ||
232 | parser p (s, name, i, pr); | |
c18f07f8 JS |
233 | return p.parse (); |
234 | } | |
235 | ||
fe410f52 SM |
236 | stapfile* |
237 | parse_library_macros (systemtap_session& s, const string& name) | |
238 | { | |
239 | ifstream i(name.c_str(), ios::in); | |
240 | if (i.fail()) | |
241 | { | |
242 | cerr << (file_exists(name) | |
243 | ? _F("Input file '%s' can't be opened for reading.", name.c_str()) | |
244 | : _F("Input file '%s' is missing.", name.c_str())) | |
245 | << endl; | |
246 | return 0; | |
247 | } | |
248 | ||
249 | parser p (s, name, i, false); // TODOXX pr is ...? should path be full?? | |
250 | return p.parse_library_macros (); | |
251 | } | |
252 | ||
c18f07f8 | 253 | // ------------------------------------------------------------------------ |
bb2e3076 FCE |
254 | |
255 | ||
4cd32d8c JS |
256 | parser::parser (systemtap_session& s, const string &n, istream& i, bool p): |
257 | session (s), input_name (n), input (i, input_name, s), privileged (p), | |
a07a2c28 | 258 | context(con_unknown), systemtap_v_seen(0), last_t (0), next_t (0), num_errors (0) |
4cd32d8c JS |
259 | { |
260 | } | |
2f1a1aea FCE |
261 | |
262 | parser::~parser() | |
263 | { | |
2f1a1aea FCE |
264 | } |
265 | ||
d7f3e0c5 GH |
266 | static string |
267 | tt2str(token_type tt) | |
268 | { | |
269 | switch (tt) | |
270 | { | |
271 | case tok_junk: return "junk"; | |
272 | case tok_identifier: return "identifier"; | |
273 | case tok_operator: return "operator"; | |
274 | case tok_string: return "string"; | |
275 | case tok_number: return "number"; | |
276 | case tok_embedded: return "embedded-code"; | |
6e213f58 | 277 | case tok_keyword: return "keyword"; |
d7f3e0c5 GH |
278 | } |
279 | return "unknown token"; | |
280 | } | |
82919855 | 281 | |
0323ed4d WC |
282 | ostream& |
283 | operator << (ostream& o, const source_loc& loc) | |
284 | { | |
a704a23b | 285 | o << loc.file->name << ":" |
0323ed4d WC |
286 | << loc.line << ":" |
287 | << loc.column; | |
288 | ||
289 | return o; | |
290 | } | |
291 | ||
56099f08 FCE |
292 | ostream& |
293 | operator << (ostream& o, const token& t) | |
294 | { | |
d7f3e0c5 | 295 | o << tt2str(t.type); |
56099f08 | 296 | |
6e213f58 | 297 | if (t.type != tok_embedded && t.type != tok_keyword) // XXX: other types? |
56099f08 | 298 | { |
24cb178f FCE |
299 | o << " '"; |
300 | for (unsigned i=0; i<t.content.length(); i++) | |
301 | { | |
302 | char c = t.content[i]; | |
303 | o << (isprint (c) ? c : '?'); | |
304 | } | |
305 | o << "'"; | |
56099f08 | 306 | } |
56099f08 | 307 | |
dff50e09 | 308 | o << " at " |
0323ed4d | 309 | << t.location; |
56099f08 FCE |
310 | |
311 | return o; | |
312 | } | |
313 | ||
314 | ||
dff50e09 | 315 | void |
2f1a1aea FCE |
316 | parser::print_error (const parse_error &pe) |
317 | { | |
16fc963f | 318 | const token *tok = pe.tok ? pe.tok : last_t; |
6b7d7185 | 319 | session.print_error(pe, tok, input_name); |
2f1a1aea FCE |
320 | num_errors ++; |
321 | } | |
322 | ||
323 | ||
2f1a1aea | 324 | |
c434ec7e FCE |
325 | |
326 | template <typename OPERAND> | |
327 | bool eval_comparison (const OPERAND& lhs, const token* op, const OPERAND& rhs) | |
328 | { | |
329 | if (op->type == tok_operator && op->content == "<=") | |
330 | { return lhs <= rhs; } | |
331 | else if (op->type == tok_operator && op->content == ">=") | |
332 | { return lhs >= rhs; } | |
333 | else if (op->type == tok_operator && op->content == "<") | |
334 | { return lhs < rhs; } | |
335 | else if (op->type == tok_operator && op->content == ">") | |
336 | { return lhs > rhs; } | |
337 | else if (op->type == tok_operator && op->content == "==") | |
338 | { return lhs == rhs; } | |
339 | else if (op->type == tok_operator && op->content == "!=") | |
340 | { return lhs != rhs; } | |
341 | else | |
2677d2fb | 342 | throw parse_error (_("expected comparison operator"), op); |
c434ec7e FCE |
343 | } |
344 | ||
345 | ||
534aad8b SM |
346 | // Here, we perform on-the-fly preprocessing in two passes. |
347 | ||
348 | // First pass - macro declaration and expansion. | |
349 | // | |
350 | // The basic form of a declaration is @define SIGNATURE %( BODY %) | |
351 | // where SIGNATURE is of the form macro_name (a, b, c, ...) | |
352 | // and BODY can obtain the parameter contents as @a, @b, @c, .... | |
353 | // Note that parameterless macros can also be declared. | |
354 | // | |
3932c705 | 355 | // Macro definitions may not be nested. |
534aad8b SM |
356 | // A macro is available textually after it has been defined. |
357 | // | |
358 | // The basic form of a macro invocation | |
359 | // for a parameterless macro is @macro_name, | |
360 | // for a macro with parameters is @macro_name(param_1, param_2, ...). | |
361 | // | |
26718dbe SM |
362 | // NB: this means that a parameterless macro @foo called as @foo(a, b, c) |
363 | // leaves its 'parameters' alone, rather than consuming them to result | |
364 | // in a "too many parameters error". This may be useful in the unusual | |
365 | // case of wanting @foo to expand to the name of a function. | |
534aad8b SM |
366 | // |
367 | // Invocations of unknown macros are left unexpanded, to allow | |
368 | // the continued use of constructs such as @cast, @var, etc. | |
369 | ||
fe410f52 | 370 | macrodecl::~macrodecl () |
534aad8b SM |
371 | { |
372 | delete tok; | |
373 | for (vector<const token*>::iterator it = body.begin(); | |
374 | it != body.end(); it++) | |
375 | delete *it; | |
376 | } | |
377 | ||
378 | parser::pp1_activation::~pp1_activation () | |
379 | { | |
380 | delete tok; | |
fe410f52 | 381 | if (curr_macro->is_closure()) return; // body is shared with an earlier declaration |
534aad8b SM |
382 | for (map<string, pp_macrodecl*>::iterator it = params.begin(); |
383 | it != params.end(); it++) | |
384 | delete it->second; | |
385 | } | |
386 | ||
387 | // Grab a token from the current input source (main file or macro body): | |
388 | const token* | |
389 | parser::next_pp1 () | |
390 | { | |
391 | if (pp1_state.empty()) | |
392 | return input.scan (); | |
393 | ||
394 | // otherwise, we're inside a macro | |
395 | pp1_activation* act = pp1_state.back(); | |
396 | unsigned& cursor = act->cursor; | |
397 | if (cursor < act->curr_macro->body.size()) | |
398 | { | |
399 | token* t = new token(*act->curr_macro->body[cursor]); | |
0002fc51 | 400 | t->chain = new token(*act->tok); // mark chained token |
534aad8b SM |
401 | cursor++; |
402 | return t; | |
403 | } | |
404 | else | |
405 | return 0; // reached end of macro body | |
406 | } | |
407 | ||
408 | const token* | |
409 | parser::scan_pp1 () | |
410 | { | |
411 | while (true) | |
412 | { | |
413 | const token* t = next_pp1 (); | |
414 | if (t == 0) // EOF or end of macro body | |
415 | { | |
416 | if (pp1_state.empty()) // actual EOF | |
417 | return 0; | |
418 | ||
419 | // Exit macro and loop around to look for the next token. | |
420 | pp1_activation* act = pp1_state.back(); | |
421 | pp1_state.pop_back(); delete act; | |
422 | continue; | |
423 | } | |
424 | ||
425 | // macro definition | |
426 | if (t->type == tok_operator && t->content == "@define") | |
427 | { | |
428 | if (!pp1_state.empty()) | |
429 | throw parse_error (_("'@define' forbidden inside macro body"), t); | |
430 | delete t; | |
431 | ||
432 | // handle macro definition | |
433 | // (1) consume macro signature | |
3932c705 | 434 | t = input.scan(); |
534aad8b SM |
435 | if (! (t && t->type == tok_identifier)) |
436 | throw parse_error (_("expected identifier"), t); | |
437 | string name = t->content; | |
438 | ||
439 | // check for redefinition of existing macro | |
440 | if (pp1_namespace.find(name) != pp1_namespace.end()) | |
78ab2de3 SM |
441 | { |
442 | parse_error er (_F("attempt to redefine macro '@%s' in the same file", name.c_str ()), t); | |
443 | ||
444 | // Also point to pp1_namespace[name]->tok, the site of | |
445 | // the original definition: | |
446 | er.chain = new parse_error (_F("macro '@%s' first defined here", | |
447 | name.c_str()), pp1_namespace[name]->tok); | |
448 | throw er; | |
449 | } | |
450 | ||
451 | // XXX: the above restriction was mostly necessary due to | |
452 | // wanting to leave open the possibility of | |
453 | // statically-scoped semantics in the future. | |
26718dbe SM |
454 | |
455 | // XXX: this cascades into further parse errors as the | |
456 | // parser tries to parse the remaining definition... (e.g. | |
457 | // it can't tell that the macro body isn't a conditional, | |
458 | // that the uses of parameters aren't nonexistent | |
459 | // macros.....) | |
534aad8b SM |
460 | if (name == "define") |
461 | throw parse_error (_("attempt to redefine '@define'"), t); | |
462 | if (input.atwords.count("@" + name)) | |
463 | session.print_warning (_F("macro redefines built-in operator '@%s'", name.c_str()), t); | |
464 | ||
fe410f52 SM |
465 | macrodecl* decl = (pp1_namespace[name] = new macrodecl); |
466 | decl->tok = t; | |
534aad8b SM |
467 | |
468 | // determine if the macro takes parameters | |
4ac28d7e | 469 | bool saw_params = false; |
3932c705 SM |
470 | t = input.scan(); |
471 | if (t && t->type == tok_operator && t->content == "(") | |
4ac28d7e SM |
472 | { |
473 | saw_params = true; | |
474 | do | |
475 | { | |
476 | delete t; | |
477 | ||
478 | t = input.scan (); | |
479 | if (! (t && t->type == tok_identifier)) | |
480 | throw parse_error(_("expected identifier"), t); | |
fe410f52 | 481 | decl->formal_args.push_back(t->content); |
4ac28d7e SM |
482 | delete t; |
483 | ||
484 | t = input.scan (); | |
485 | if (t && t->type == tok_operator && t->content == ",") | |
486 | { | |
487 | continue; | |
488 | } | |
489 | else if (t && t->type == tok_operator && t->content == ")") | |
490 | { | |
491 | delete t; | |
492 | t = input.scan(); | |
493 | break; | |
494 | } | |
495 | else | |
496 | { | |
497 | throw parse_error (_("expected ',' or ')'"), t); | |
498 | } | |
499 | } | |
500 | while (true); | |
501 | } | |
534aad8b SM |
502 | |
503 | // (2) identify & consume macro body | |
3932c705 | 504 | if (! (t && t->type == tok_operator && t->content == "%(")) |
4ac28d7e SM |
505 | { |
506 | if (saw_params) | |
507 | throw parse_error (_("expected '%('"), t); | |
508 | else | |
509 | throw parse_error (_("expected '%(' or '('"), t); | |
510 | } | |
3932c705 | 511 | delete t; |
534aad8b | 512 | |
3932c705 SM |
513 | t = slurp_pp1_body (decl->body); |
514 | if (!t) | |
515 | throw parse_error (_("incomplete macro definition - missing '%)'"), decl->tok); | |
516 | delete t; | |
534aad8b SM |
517 | |
518 | // Now loop around to look for a real token. | |
519 | continue; | |
520 | } | |
521 | ||
522 | // (potential) macro invocation | |
523 | if (t->type == tok_operator && t->content[0] == '@') | |
524 | { | |
525 | string name = t->content.substr(1); // strip initial '@' | |
526 | ||
527 | // check if name refers to a real parameter or macro | |
fe410f52 | 528 | macrodecl* decl; |
534aad8b SM |
529 | pp1_activation* act = pp1_state.empty() ? 0 : pp1_state.back(); |
530 | if (act && act->params.find(name) != act->params.end()) | |
531 | decl = act->params[name]; | |
fe410f52 SM |
532 | else if (!(act && act->curr_macro->context == ctx_library) |
533 | && pp1_namespace.find(name) != pp1_namespace.end()) | |
534aad8b | 534 | decl = pp1_namespace[name]; |
fe410f52 SM |
535 | else if (session.library_macros.find(name) |
536 | != session.library_macros.end()) | |
537 | decl = session.library_macros[name]; | |
534aad8b SM |
538 | else // this is an ordinary @operator |
539 | return t; | |
540 | ||
541 | // handle macro invocation | |
542 | pp1_activation *new_act = new pp1_activation(*t, decl); | |
fe410f52 | 543 | unsigned num_params = decl->formal_args.size(); |
534aad8b SM |
544 | |
545 | // (1a) restore parameter invocation closure | |
fe410f52 | 546 | if (num_params == 0 && decl->is_closure()) |
534aad8b SM |
547 | { |
548 | // NB: decl->parent_act is always safe since the | |
549 | // parameter decl (if any) comes from an activation | |
550 | // record which deeper in the stack than new_act. | |
551 | ||
552 | // decl is a macro parameter which must be evaluated in | |
553 | // the context of the original point of invocation: | |
fe410f52 | 554 | new_act->params = ((pp_macrodecl*)decl)->parent_act->params; |
534aad8b SM |
555 | goto expand; |
556 | } | |
557 | ||
558 | // (1b) consume macro parameters (if any) | |
559 | if (num_params == 0) | |
560 | goto expand; | |
561 | ||
562 | // for simplicity, we do not allow macro constructs here | |
563 | // -- if we did, we'd have to recursively call scan_pp1() | |
564 | t = next_pp1 (); | |
39566df2 | 565 | if (! (t && t->type == tok_operator && t->content == "(")) |
534aad8b SM |
566 | { |
567 | delete new_act; | |
52c2652f | 568 | throw parse_error (_NF |
534aad8b | 569 | ("expected '(' in invocation of macro '@%s'" |
f499dee5 | 570 | " taking %d parameter", |
534aad8b | 571 | "expected '(' in invocation of macro '@%s'" |
f499dee5 | 572 | " taking %d parameters", |
52c2652f | 573 | num_params, name.c_str(), num_params), t); |
534aad8b SM |
574 | } |
575 | ||
576 | // XXX perhaps parse/count the full number of params, | |
577 | // so we can say "expected x, found y params" on error? | |
578 | for (unsigned i = 0; i < num_params; i++) | |
579 | { | |
580 | delete t; | |
581 | ||
582 | // create parameter closure | |
fe410f52 | 583 | string param_name = decl->formal_args[i]; |
534aad8b | 584 | pp_macrodecl* p = (new_act->params[param_name] |
fe410f52 SM |
585 | = new pp_macrodecl); |
586 | p->tok = new token(*new_act->tok); | |
587 | p->parent_act = act; | |
534aad8b SM |
588 | // NB: *new_act->tok points to invocation, act is NULL at top level |
589 | ||
590 | t = slurp_pp1_param (p->body); | |
591 | ||
592 | // check correct usage of ',' or ')' | |
593 | if (t == 0) // hit unexpected EOF or end of macro | |
594 | { | |
595 | // XXX could we pop the stack and continue parsing | |
596 | // the invocation, allowing macros to construct new | |
597 | // invocations in piecemeal fashion?? | |
598 | const token* orig_t = new token(*new_act->tok); | |
599 | delete new_act; | |
600 | throw parse_error (_("could not find end of macro invocation"), orig_t); | |
601 | } | |
602 | if (t->type == tok_operator && t->content == ",") | |
603 | { | |
604 | if (i + 1 == num_params) | |
605 | { | |
606 | delete new_act; | |
607 | throw parse_error (_F("too many parameters for macro '@%s' (expected %d)", name.c_str(), num_params), t); | |
608 | } | |
609 | } | |
610 | else if (t->type == tok_operator && t->content == ")") | |
611 | { | |
612 | if (i + 1 != num_params) | |
613 | { | |
614 | delete new_act; | |
615 | throw parse_error (_F("too few parameters for macro '@%s' (expected %d)", name.c_str(), num_params), t); | |
616 | } | |
617 | } | |
618 | else | |
619 | { | |
620 | // XXX this is, incidentally, impossible | |
621 | delete new_act; | |
622 | throw parse_error(_("expected ',' or ')' after macro parameter"), t); | |
623 | } | |
624 | } | |
625 | ||
626 | delete t; | |
627 | ||
628 | // (2) set up macro expansion | |
629 | expand: | |
630 | pp1_state.push_back (new_act); | |
631 | ||
632 | // Now loop around to look for a real token. | |
633 | continue; | |
634 | } | |
635 | ||
636 | // Otherwise, we have an ordinary token. | |
637 | return t; | |
638 | } | |
639 | } | |
640 | ||
641 | // Consume a single macro invocation's parameters, heeding nested ( ) | |
642 | // brackets and stopping on an unbalanced ')' or an unbracketed ',' | |
643 | // (and returning the final separator token). | |
644 | const token* | |
645 | parser::slurp_pp1_param (vector<const token*>& param) | |
646 | { | |
647 | const token* t = 0; | |
648 | unsigned nesting = 0; | |
649 | do | |
650 | { | |
651 | t = next_pp1 (); | |
652 | ||
653 | if (!t) | |
654 | break; | |
655 | if (t->type == tok_operator && t->content == "(") | |
656 | ++nesting; | |
657 | else if (nesting && t->type == tok_operator && t->content == ")") | |
658 | --nesting; | |
659 | else if (!nesting && t->type == tok_operator | |
660 | && (t->content == ")" || t->content == ",")) | |
661 | break; | |
662 | param.push_back(t); | |
663 | } | |
664 | while (true); | |
665 | return t; // report ")" or "," or NULL | |
666 | } | |
667 | ||
668 | ||
669 | // Consume a macro declaration's body, heeding nested %( %) brackets. | |
670 | const token* | |
671 | parser::slurp_pp1_body (vector<const token*>& body) | |
672 | { | |
673 | const token* t = 0; | |
674 | unsigned nesting = 0; | |
675 | do | |
676 | { | |
677 | t = next_pp1 (); | |
678 | ||
679 | if (!t) | |
680 | break; | |
681 | if (t->type == tok_operator && t->content == "%(") | |
682 | ++nesting; | |
683 | else if (nesting && t->type == tok_operator && t->content == "%)") | |
684 | --nesting; | |
685 | else if (!nesting && t->type == tok_operator && t->content == "%)") | |
686 | break; | |
687 | body.push_back(t); | |
688 | } | |
689 | while (true); | |
690 | return t; // report final "%)" or NULL | |
691 | } | |
692 | ||
fe410f52 SM |
693 | // Used for parsing .stpm files. |
694 | stapfile* | |
695 | parser::parse_library_macros () | |
696 | { | |
697 | stapfile* f = new stapfile; | |
698 | input.set_current_file (f); | |
699 | ||
700 | try | |
701 | { | |
702 | const token* t = scan_pp1 (); | |
703 | ||
704 | // Currently we only take objection to macro invocations if they | |
705 | // produce a non-whitespace token after being expanded. | |
706 | ||
707 | // XXX should we prevent macro invocations even if they expand to empty?? | |
708 | ||
709 | if (t != 0) | |
710 | throw parse_error (_F("library macro file '%s' contains non-@define construct", input_name.c_str()), t); | |
711 | ||
712 | // We need to first check whether *any* of the macros are duplicates, | |
713 | // then commit to including the entire file in the global namespace | |
714 | // (or not). Yuck. | |
715 | for (map<string, macrodecl*>::iterator it = pp1_namespace.begin(); | |
716 | it != pp1_namespace.end(); it++) | |
717 | { | |
718 | string name = it->first; | |
719 | ||
720 | if (session.library_macros.find(name) != session.library_macros.end()) | |
721 | { | |
78ab2de3 SM |
722 | parse_error er(_F("duplicate definition of library macro '@%s'", name.c_str()), it->second->tok); |
723 | er.chain = new parse_error (_F("macro '@%s' first defined here", name.c_str()), session.library_macros[name]->tok); | |
724 | print_error (er); | |
fe410f52 | 725 | |
78ab2de3 | 726 | delete er.chain; |
fe410f52 SM |
727 | delete f; |
728 | return 0; | |
729 | } | |
730 | } | |
731 | ||
732 | } | |
733 | catch (const parse_error& pe) | |
734 | { | |
735 | print_error (pe); | |
736 | delete f; | |
737 | return 0; | |
738 | } | |
739 | ||
740 | // If no errors, include the entire file. Note how this is outside | |
741 | // of the try-catch block -- no errors possible. | |
742 | for (map<string, macrodecl*>::iterator it = pp1_namespace.begin(); | |
743 | it != pp1_namespace.end(); it++) | |
744 | { | |
745 | string name = it->first; | |
746 | ||
747 | session.library_macros[name] = it->second; | |
748 | session.library_macros[name]->context = ctx_library; | |
fe410f52 SM |
749 | } |
750 | ||
751 | return f; | |
752 | } | |
753 | ||
534aad8b SM |
754 | // Second pass - preprocessor conditional expansion. |
755 | // | |
177a8ead | 756 | // The basic form is %( CONDITION %? THEN-TOKENS %: ELSE-TOKENS %) |
44ce8ed5 FCE |
757 | // where CONDITION is: kernel_v[r] COMPARISON-OP "version-string" |
758 | // or: arch COMPARISON-OP "arch-string" | |
db135493 | 759 | // or: systemtap_v COMPARISON-OP "version-string" |
2e6dd9d0 | 760 | // or: systemtap_privilege COMPARISON-OP "privilege-string" |
561079c8 | 761 | // or: CONFIG_foo COMPARISON-OP "config-string" |
717a457b | 762 | // or: CONFIG_foo COMPARISON-OP number |
4227f98d | 763 | // or: CONFIG_foo COMPARISON-OP CONFIG_bar |
5811366a FCE |
764 | // or: "string1" COMPARISON-OP "string2" |
765 | // or: number1 COMPARISON-OP number2 | |
44ce8ed5 | 766 | // The %: ELSE-TOKENS part is optional. |
177a8ead FCE |
767 | // |
768 | // e.g. %( kernel_v > "2.5" %? "foo" %: "baz" %) | |
c434ec7e | 769 | // e.g. %( arch != "i?86" %? "foo" %: "baz" %) |
561079c8 | 770 | // e.g. %( CONFIG_foo %? "foo" %: "baz" %) |
177a8ead FCE |
771 | // |
772 | // Up to an entire %( ... %) expression is processed by a single call | |
773 | // to this function. Tokens included by any nested conditions are | |
774 | // enqueued in a private vector. | |
775 | ||
776 | bool eval_pp_conditional (systemtap_session& s, | |
777 | const token* l, const token* op, const token* r) | |
778 | { | |
44ce8ed5 | 779 | if (l->type == tok_identifier && (l->content == "kernel_v" || |
db135493 FCE |
780 | l->content == "kernel_vr" || |
781 | l->content == "systemtap_v")) | |
44ce8ed5 | 782 | { |
db135493 | 783 | if (! (r->type == tok_string)) |
2677d2fb | 784 | throw parse_error (_("expected string literal"), r); |
db135493 | 785 | |
44ce8ed5 | 786 | string target_kernel_vr = s.kernel_release; |
197a4d62 | 787 | string target_kernel_v = s.kernel_base_release; |
db135493 | 788 | string target; |
dff50e09 | 789 | |
db135493 FCE |
790 | if (l->content == "kernel_v") target = target_kernel_v; |
791 | else if (l->content == "kernel_vr") target = target_kernel_vr; | |
792 | else if (l->content == "systemtap_v") target = s.compatible; | |
793 | else assert (0); | |
7a468d68 | 794 | |
7a468d68 FCE |
795 | string query = r->content; |
796 | bool rhs_wildcard = (strpbrk (query.c_str(), "*?[") != 0); | |
797 | ||
44ce8ed5 FCE |
798 | // collect acceptable strverscmp results. |
799 | int rvc_ok1, rvc_ok2; | |
7a468d68 | 800 | bool wc_ok = false; |
44ce8ed5 FCE |
801 | if (op->type == tok_operator && op->content == "<=") |
802 | { rvc_ok1 = -1; rvc_ok2 = 0; } | |
803 | else if (op->type == tok_operator && op->content == ">=") | |
804 | { rvc_ok1 = 1; rvc_ok2 = 0; } | |
805 | else if (op->type == tok_operator && op->content == "<") | |
806 | { rvc_ok1 = -1; rvc_ok2 = -1; } | |
807 | else if (op->type == tok_operator && op->content == ">") | |
808 | { rvc_ok1 = 1; rvc_ok2 = 1; } | |
809 | else if (op->type == tok_operator && op->content == "==") | |
7a468d68 | 810 | { rvc_ok1 = 0; rvc_ok2 = 0; wc_ok = true; } |
44ce8ed5 | 811 | else if (op->type == tok_operator && op->content == "!=") |
7a468d68 | 812 | { rvc_ok1 = -1; rvc_ok2 = 1; wc_ok = true; } |
44ce8ed5 | 813 | else |
2677d2fb | 814 | throw parse_error (_("expected comparison operator"), op); |
7a468d68 FCE |
815 | |
816 | if ((!wc_ok) && rhs_wildcard) | |
2677d2fb | 817 | throw parse_error (_("wildcard not allowed with order comparison operators"), op); |
7a468d68 FCE |
818 | |
819 | if (rhs_wildcard) | |
820 | { | |
821 | int rvc_result = fnmatch (query.c_str(), target.c_str(), | |
822 | FNM_NOESCAPE); // spooky | |
823 | bool badness = (rvc_result == 0) ^ (op->content == "=="); | |
824 | return !badness; | |
825 | } | |
826 | else | |
827 | { | |
828 | int rvc_result = strverscmp (target.c_str(), query.c_str()); | |
829 | // normalize rvc_result | |
830 | if (rvc_result < 0) rvc_result = -1; | |
831 | if (rvc_result > 0) rvc_result = 1; | |
832 | return (rvc_result == rvc_ok1 || rvc_result == rvc_ok2); | |
833 | } | |
44ce8ed5 | 834 | } |
2e6dd9d0 SM |
835 | else if (l->type == tok_identifier && l->content == "systemtap_privilege") |
836 | { | |
837 | string target_privilege = | |
cba5b802 SM |
838 | /* XXX perhaps include a "guru" state */ |
839 | pr_contains(s.privilege, pr_stapdev) ? "stapdev" | |
840 | : pr_contains(s.privilege, pr_stapsys) ? "stapsys" | |
841 | : pr_contains(s.privilege, pr_stapusr) ? "stapusr" | |
2e6dd9d0 SM |
842 | : "none"; /* should be impossible -- s.privilege always one of above */ |
843 | assert(target_privilege != "none"); | |
844 | ||
845 | if (! (r->type == tok_string)) | |
846 | throw parse_error (_("expected string literal"), r); | |
847 | string query_privilege = r->content; | |
848 | ||
849 | bool nomatch = (target_privilege != query_privilege); | |
850 | ||
851 | bool result; | |
852 | if (op->type == tok_operator && op->content == "==") | |
853 | result = !nomatch; | |
854 | else if (op->type == tok_operator && op->content == "!=") | |
855 | result = nomatch; | |
856 | else | |
857 | throw parse_error (_("expected '==' or '!='"), op); | |
cba5b802 | 858 | /* XXX perhaps allow <= >= and similar comparisons */ |
2e6dd9d0 SM |
859 | |
860 | return result; | |
861 | } | |
44ce8ed5 FCE |
862 | else if (l->type == tok_identifier && l->content == "arch") |
863 | { | |
864 | string target_architecture = s.architecture; | |
865 | if (! (r->type == tok_string)) | |
2677d2fb | 866 | throw parse_error (_("expected string literal"), r); |
44ce8ed5 | 867 | string query_architecture = r->content; |
dff50e09 | 868 | |
7a468d68 FCE |
869 | int nomatch = fnmatch (query_architecture.c_str(), |
870 | target_architecture.c_str(), | |
871 | FNM_NOESCAPE); // still spooky | |
872 | ||
561079c8 FCE |
873 | bool result; |
874 | if (op->type == tok_operator && op->content == "==") | |
875 | result = !nomatch; | |
876 | else if (op->type == tok_operator && op->content == "!=") | |
877 | result = nomatch; | |
878 | else | |
2677d2fb | 879 | throw parse_error (_("expected '==' or '!='"), op); |
561079c8 | 880 | |
d9677d7b DS |
881 | return result; |
882 | } | |
883 | else if (l->type == tok_identifier && l->content == "runtime") | |
884 | { | |
885 | if (! (r->type == tok_string)) | |
886 | throw parse_error (_("expected string literal"), r); | |
887 | ||
888 | string query_runtime = r->content; | |
889 | string target_runtime; | |
890 | ||
891 | target_runtime = (s.runtime_mode == systemtap_session::dyninst_runtime | |
892 | ? "dyninst" : "kernel"); | |
893 | int nomatch = fnmatch (query_runtime.c_str(), | |
894 | target_runtime.c_str(), | |
895 | FNM_NOESCAPE); // still spooky | |
896 | ||
897 | bool result; | |
898 | if (op->type == tok_operator && op->content == "==") | |
899 | result = !nomatch; | |
900 | else if (op->type == tok_operator && op->content == "!=") | |
901 | result = nomatch; | |
902 | else | |
903 | throw parse_error (_("expected '==' or '!='"), op); | |
904 | ||
561079c8 FCE |
905 | return result; |
906 | } | |
60d98537 | 907 | else if (l->type == tok_identifier && startswith(l->content, "CONFIG_")) |
561079c8 | 908 | { |
717a457b MW |
909 | if (r->type == tok_string) |
910 | { | |
911 | string lhs = s.kernel_config[l->content]; // may be empty | |
912 | string rhs = r->content; | |
561079c8 | 913 | |
717a457b | 914 | int nomatch = fnmatch (rhs.c_str(), lhs.c_str(), FNM_NOESCAPE); // still spooky |
561079c8 | 915 | |
717a457b MW |
916 | bool result; |
917 | if (op->type == tok_operator && op->content == "==") | |
918 | result = !nomatch; | |
919 | else if (op->type == tok_operator && op->content == "!=") | |
920 | result = nomatch; | |
921 | else | |
2677d2fb | 922 | throw parse_error (_("expected '==' or '!='"), op); |
dff50e09 | 923 | |
717a457b MW |
924 | return result; |
925 | } | |
926 | else if (r->type == tok_number) | |
927 | { | |
928 | const char* startp = s.kernel_config[l->content].c_str (); | |
929 | char* endp = (char*) startp; | |
930 | errno = 0; | |
931 | int64_t lhs = (int64_t) strtoll (startp, & endp, 0); | |
932 | if (errno == ERANGE || errno == EINVAL || *endp != '\0') | |
933 | throw parse_error ("Config option value not a number", l); | |
934 | ||
935 | int64_t rhs = lex_cast<int64_t>(r->content); | |
936 | return eval_comparison (lhs, op, rhs); | |
937 | } | |
4227f98d | 938 | else if (r->type == tok_identifier |
60d98537 | 939 | && startswith(r->content, "CONFIG_")) |
4227f98d MW |
940 | { |
941 | // First try to convert both to numbers, | |
942 | // otherwise threat both as strings. | |
943 | const char* startp = s.kernel_config[l->content].c_str (); | |
944 | char* endp = (char*) startp; | |
945 | errno = 0; | |
946 | int64_t val = (int64_t) strtoll (startp, & endp, 0); | |
947 | if (errno != ERANGE && errno != EINVAL && *endp == '\0') | |
948 | { | |
949 | int64_t lhs = val; | |
950 | startp = s.kernel_config[r->content].c_str (); | |
951 | endp = (char*) startp; | |
952 | errno = 0; | |
953 | int64_t rhs = (int64_t) strtoll (startp, & endp, 0); | |
954 | if (errno != ERANGE && errno != EINVAL && *endp == '\0') | |
955 | return eval_comparison (lhs, op, rhs); | |
956 | } | |
957 | ||
958 | string lhs = s.kernel_config[l->content]; | |
959 | string rhs = s.kernel_config[r->content]; | |
960 | return eval_comparison (lhs, op, rhs); | |
961 | } | |
717a457b | 962 | else |
ce0f6648 | 963 | throw parse_error (_("expected string, number literal or other CONFIG_... as right side operand"), r); |
dff50e09 | 964 | } |
c434ec7e | 965 | else if (l->type == tok_string && r->type == tok_string) |
5811366a | 966 | { |
c434ec7e FCE |
967 | string lhs = l->content; |
968 | string rhs = r->content; | |
969 | return eval_comparison (lhs, op, rhs); | |
970 | // NB: no wildcarding option here | |
971 | } | |
972 | else if (l->type == tok_number && r->type == tok_number) | |
973 | { | |
974 | int64_t lhs = lex_cast<int64_t>(l->content); | |
975 | int64_t rhs = lex_cast<int64_t>(r->content); | |
976 | return eval_comparison (lhs, op, rhs); | |
7a468d68 | 977 | // NB: no wildcarding option here |
5811366a FCE |
978 | } |
979 | else if (l->type == tok_string && r->type == tok_number | |
980 | && op->type == tok_operator) | |
2677d2fb | 981 | throw parse_error (_("expected string literal as right value"), r); |
5811366a FCE |
982 | else if (l->type == tok_number && r->type == tok_string |
983 | && op->type == tok_operator) | |
2677d2fb | 984 | throw parse_error (_("expected number literal as right value"), r); |
c434ec7e | 985 | |
177a8ead | 986 | else |
d9677d7b DS |
987 | throw parse_error (_("expected 'arch', 'kernel_v', 'kernel_vr', 'systemtap_v',\n" |
988 | " 'runtime', 'systemtap_privilege', 'CONFIG_...', or\n" | |
989 | " comparison between strings or integers"), l); | |
177a8ead FCE |
990 | } |
991 | ||
992 | ||
5811366a | 993 | // Only tokens corresponding to the TRUE statement must be expanded |
177a8ead | 994 | const token* |
b5477cd9 | 995 | parser::scan_pp () |
177a8ead FCE |
996 | { |
997 | while (true) | |
998 | { | |
e92f2566 JS |
999 | pp_state_t pp = PP_NONE; |
1000 | if (!pp_state.empty()) | |
1001 | pp = pp_state.back().second; | |
1002 | ||
1003 | const token* t = 0; | |
1004 | if (pp == PP_SKIP_THEN || pp == PP_SKIP_ELSE) | |
1005 | t = skip_pp (); | |
1006 | else | |
534aad8b | 1007 | t = scan_pp1 (); |
e92f2566 JS |
1008 | |
1009 | if (t == 0) // EOF | |
177a8ead | 1010 | { |
e92f2566 JS |
1011 | if (pp != PP_NONE) |
1012 | { | |
1013 | t = pp_state.back().first; | |
1014 | pp_state.pop_back(); // so skip_some doesn't keep trying to close this | |
ce0f6648 | 1015 | //TRANSLATORS: 'conditional' meaning 'conditional preprocessing' |
2677d2fb | 1016 | throw parse_error (_("incomplete conditional at end of file"), t); |
e92f2566 | 1017 | } |
177a8ead FCE |
1018 | return t; |
1019 | } | |
1020 | ||
e92f2566 JS |
1021 | // misplaced preprocessor "then" |
1022 | if (t->type == tok_operator && t->content == "%?") | |
2677d2fb | 1023 | throw parse_error (_("incomplete conditional - missing '%('"), t); |
e92f2566 JS |
1024 | |
1025 | // preprocessor "else" | |
1026 | if (t->type == tok_operator && t->content == "%:") | |
1027 | { | |
1028 | if (pp == PP_NONE) | |
2677d2fb | 1029 | throw parse_error (_("incomplete conditional - missing '%('"), t); |
e92f2566 | 1030 | if (pp == PP_KEEP_ELSE || pp == PP_SKIP_ELSE) |
2677d2fb | 1031 | throw parse_error (_("invalid conditional - duplicate '%:'"), t); |
1d94e4e5 SM |
1032 | // XXX: here and elsewhere, error cascades might be avoided |
1033 | // by dropping tokens until we reach the closing %) | |
e92f2566 JS |
1034 | |
1035 | pp_state.back().second = (pp == PP_KEEP_THEN) ? | |
1036 | PP_SKIP_ELSE : PP_KEEP_ELSE; | |
1037 | delete t; | |
1038 | continue; | |
1039 | } | |
1040 | ||
1041 | // preprocessor close | |
1042 | if (t->type == tok_operator && t->content == "%)") | |
1043 | { | |
1044 | if (pp == PP_NONE) | |
2677d2fb | 1045 | throw parse_error (_("incomplete conditional - missing '%('"), t); |
e92f2566 | 1046 | delete pp_state.back().first; |
a07a2c28 | 1047 | delete t; //this is the closing bracket |
e92f2566 JS |
1048 | pp_state.pop_back(); |
1049 | continue; | |
1050 | } | |
dff50e09 | 1051 | |
177a8ead FCE |
1052 | if (! (t->type == tok_operator && t->content == "%(")) // ordinary token |
1053 | return t; | |
1054 | ||
1055 | // We have a %( - it's time to throw a preprocessing party! | |
1056 | ||
2d7881bf PP |
1057 | bool result = false; |
1058 | bool and_result = true; | |
1059 | const token *n = NULL; | |
1060 | do { | |
1061 | const token *l, *op, *r; | |
534aad8b SM |
1062 | l = scan_pp1 (); |
1063 | op = scan_pp1 (); | |
1064 | r = scan_pp1 (); | |
2d7881bf | 1065 | if (l == 0 || op == 0 || r == 0) |
2677d2fb | 1066 | throw parse_error (_("incomplete condition after '%('"), t); |
2d7881bf PP |
1067 | // NB: consider generalizing to consume all tokens until %?, and |
1068 | // passing that as a vector to an evaluator. | |
1069 | ||
1070 | // Do not evaluate the condition if we haven't expanded everything. | |
1071 | // This may occur when having several recursive conditionals. | |
1072 | and_result &= eval_pp_conditional (session, l, op, r); | |
a07a2c28 LB |
1073 | if(l->content=="systemtap_v") |
1074 | systemtap_v_seen=r; | |
1075 | ||
1076 | else | |
1077 | delete r; | |
1078 | ||
2d7881bf PP |
1079 | delete l; |
1080 | delete op; | |
2d7881bf PP |
1081 | delete n; |
1082 | ||
534aad8b | 1083 | n = scan_pp1 (); |
2d7881bf PP |
1084 | if (n && n->type == tok_operator && n->content == "&&") |
1085 | continue; | |
1086 | result |= and_result; | |
1087 | and_result = true; | |
1088 | if (! (n && n->type == tok_operator && n->content == "||")) | |
1089 | break; | |
1090 | } while (true); | |
3f847830 FCE |
1091 | |
1092 | /* | |
1093 | clog << "PP eval (" << *t << ") == " << result << endl; | |
1094 | */ | |
1095 | ||
e92f2566 | 1096 | const token *m = n; |
177a8ead | 1097 | if (! (m && m->type == tok_operator && m->content == "%?")) |
2677d2fb | 1098 | throw parse_error (_("expected '%?' marker for conditional"), t); |
70c743d8 | 1099 | delete m; // "%?" |
177a8ead | 1100 | |
e92f2566 JS |
1101 | pp = result ? PP_KEEP_THEN : PP_SKIP_THEN; |
1102 | pp_state.push_back (make_pair (t, pp)); | |
3f847830 | 1103 | |
e92f2566 JS |
1104 | // Now loop around to look for a real token. |
1105 | } | |
1106 | } | |
3f847830 | 1107 | |
3f847830 | 1108 | |
e92f2566 JS |
1109 | // Skip over tokens and any errors, heeding |
1110 | // only nested preprocessor starts and ends. | |
1111 | const token* | |
1112 | parser::skip_pp () | |
1113 | { | |
1114 | const token* t = 0; | |
1115 | unsigned nesting = 0; | |
1116 | do | |
1117 | { | |
1118 | try | |
1119 | { | |
534aad8b | 1120 | t = scan_pp1 (); |
177a8ead | 1121 | } |
e92f2566 | 1122 | catch (const parse_error &e) |
70c743d8 | 1123 | { |
e92f2566 | 1124 | continue; |
70c743d8 | 1125 | } |
e92f2566 JS |
1126 | if (!t) |
1127 | break; | |
1128 | if (t->type == tok_operator && t->content == "%(") | |
1129 | ++nesting; | |
1130 | else if (nesting && t->type == tok_operator && t->content == "%)") | |
1131 | --nesting; | |
1132 | else if (!nesting && t->type == tok_operator && | |
1133 | (t->content == "%:" || t->content == "%?" || t->content == "%)")) | |
1134 | break; | |
1135 | delete t; | |
177a8ead | 1136 | } |
e92f2566 JS |
1137 | while (true); |
1138 | return t; | |
177a8ead FCE |
1139 | } |
1140 | ||
1141 | ||
2f1a1aea | 1142 | const token* |
b5477cd9 | 1143 | parser::next () |
2f1a1aea FCE |
1144 | { |
1145 | if (! next_t) | |
b5477cd9 | 1146 | next_t = scan_pp (); |
2f1a1aea | 1147 | if (! next_t) |
2677d2fb | 1148 | throw parse_error (_("unexpected end-of-file")); |
2f1a1aea | 1149 | |
2f1a1aea FCE |
1150 | last_t = next_t; |
1151 | // advance by zeroing next_t | |
1152 | next_t = 0; | |
1153 | return last_t; | |
1154 | } | |
1155 | ||
1156 | ||
1157 | const token* | |
b5477cd9 | 1158 | parser::peek () |
2f1a1aea FCE |
1159 | { |
1160 | if (! next_t) | |
b5477cd9 | 1161 | next_t = scan_pp (); |
2f1a1aea FCE |
1162 | |
1163 | // don't advance by zeroing next_t | |
1164 | last_t = next_t; | |
1165 | return next_t; | |
1166 | } | |
1167 | ||
1168 | ||
731a5359 MW |
1169 | void |
1170 | parser::swallow () | |
1171 | { | |
1172 | // can only swallow something last peeked or nexted token. | |
1173 | assert (last_t != 0); | |
1174 | delete last_t; | |
1175 | // advance by zeroing next_t | |
1176 | last_t = next_t = 0; | |
1177 | } | |
1178 | ||
1179 | ||
d7f3e0c5 GH |
1180 | static inline bool |
1181 | tok_is(token const * t, token_type tt, string const & expected) | |
1182 | { | |
1183 | return t && t->type == tt && t->content == expected; | |
1184 | } | |
1185 | ||
1186 | ||
731a5359 | 1187 | void |
d7f3e0c5 GH |
1188 | parser::expect_known (token_type tt, string const & expected) |
1189 | { | |
1190 | const token *t = next(); | |
57b73400 | 1191 | if (! (t && t->type == tt && t->content == expected)) |
ce0f6648 | 1192 | throw parse_error (_F("expected '%s'", expected.c_str())); |
731a5359 | 1193 | swallow (); // We are done with it, content was copied. |
d7f3e0c5 GH |
1194 | } |
1195 | ||
1196 | ||
731a5359 | 1197 | void |
d7f3e0c5 GH |
1198 | parser::expect_unknown (token_type tt, string & target) |
1199 | { | |
1200 | const token *t = next(); | |
1201 | if (!(t && t->type == tt)) | |
2677d2fb | 1202 | throw parse_error (_("expected ") + tt2str(tt)); |
d7f3e0c5 | 1203 | target = t->content; |
731a5359 | 1204 | swallow (); // We are done with it, content was copied. |
d7f3e0c5 GH |
1205 | } |
1206 | ||
1207 | ||
731a5359 | 1208 | void |
493ee224 DS |
1209 | parser::expect_unknown2 (token_type tt1, token_type tt2, string & target) |
1210 | { | |
1211 | const token *t = next(); | |
1212 | if (!(t && (t->type == tt1 || t->type == tt2))) | |
dd90d565 | 1213 | throw parse_error (_F("expected %s or %s", tt2str(tt1).c_str(), tt2str(tt2).c_str())); |
493ee224 | 1214 | target = t->content; |
731a5359 | 1215 | swallow (); // We are done with it, content was copied. |
493ee224 DS |
1216 | } |
1217 | ||
1218 | ||
731a5359 | 1219 | void |
d7f3e0c5 GH |
1220 | parser::expect_op (std::string const & expected) |
1221 | { | |
731a5359 | 1222 | expect_known (tok_operator, expected); |
d7f3e0c5 GH |
1223 | } |
1224 | ||
1225 | ||
731a5359 | 1226 | void |
d7f3e0c5 GH |
1227 | parser::expect_kw (std::string const & expected) |
1228 | { | |
731a5359 | 1229 | expect_known (tok_keyword, expected); |
d7f3e0c5 GH |
1230 | } |
1231 | ||
dff50e09 | 1232 | const token* |
731a5359 MW |
1233 | parser::expect_kw_token (std::string const & expected) |
1234 | { | |
1235 | const token *t = next(); | |
1236 | if (! (t && t->type == tok_keyword && t->content == expected)) | |
1237 | throw parse_error (_F("expected '%s'", expected.c_str())); | |
1238 | return t; | |
1239 | } | |
1240 | ||
1241 | void | |
e38723d2 | 1242 | parser::expect_number (int64_t & value) |
57b73400 | 1243 | { |
e38723d2 MH |
1244 | bool neg = false; |
1245 | const token *t = next(); | |
1246 | if (t->type == tok_operator && t->content == "-") | |
1247 | { | |
1248 | neg = true; | |
731a5359 | 1249 | swallow (); |
e38723d2 MH |
1250 | t = next (); |
1251 | } | |
1252 | if (!(t && t->type == tok_number)) | |
2677d2fb | 1253 | throw parse_error (_("expected number")); |
e38723d2 MH |
1254 | |
1255 | const char* startp = t->content.c_str (); | |
1256 | char* endp = (char*) startp; | |
1257 | ||
1258 | // NB: we allow controlled overflow from LLONG_MIN .. ULLONG_MAX | |
1259 | // Actually, this allows all the way from -ULLONG_MAX to ULLONG_MAX, | |
1260 | // since the lexer only gives us positive digit strings, but we'll | |
1261 | // limit it to LLONG_MIN when a '-' operator is fed into the literal. | |
1262 | errno = 0; | |
1263 | value = (int64_t) strtoull (startp, & endp, 0); | |
1264 | if (errno == ERANGE || errno == EINVAL || *endp != '\0' | |
1265 | || (neg && (unsigned long long) value > 9223372036854775808ULL) | |
1266 | || (unsigned long long) value > 18446744073709551615ULL | |
1267 | || value < -9223372036854775807LL-1) | |
2677d2fb | 1268 | throw parse_error (_("number invalid or out of range")); |
dff50e09 | 1269 | |
e38723d2 MH |
1270 | if (neg) |
1271 | value = -value; | |
1272 | ||
731a5359 | 1273 | swallow (); // We are done with it, content was parsed and copied into value. |
57b73400 GH |
1274 | } |
1275 | ||
d7f3e0c5 | 1276 | |
dff50e09 | 1277 | const token* |
50cc7cd5 | 1278 | parser::expect_ident_or_atword (std::string & target) |
d7f3e0c5 | 1279 | { |
06219d6f SM |
1280 | const token *t = next(); |
1281 | ||
1282 | // accept identifiers and operators beginning in '@': | |
1283 | if (!t || (t->type != tok_identifier | |
1284 | && (t->type != tok_operator || t->content[0] != '@'))) | |
1285 | // XXX currently this is only called from parse_hist_op_or_bare_name(), | |
1286 | // so the message is accurate, but keep an eye out in the future: | |
dd90d565 | 1287 | throw parse_error (_F("expected %s or statistical operation", tt2str(tok_identifier).c_str())); |
06219d6f SM |
1288 | |
1289 | target = t->content; | |
1290 | return t; | |
d7f3e0c5 GH |
1291 | } |
1292 | ||
1293 | ||
731a5359 | 1294 | void |
493ee224 DS |
1295 | parser::expect_ident_or_keyword (std::string & target) |
1296 | { | |
731a5359 | 1297 | expect_unknown2 (tok_identifier, tok_keyword, target); |
493ee224 DS |
1298 | } |
1299 | ||
1300 | ||
dff50e09 | 1301 | bool |
d7f3e0c5 GH |
1302 | parser::peek_op (std::string const & op) |
1303 | { | |
1304 | return tok_is (peek(), tok_operator, op); | |
1305 | } | |
1306 | ||
1307 | ||
dff50e09 | 1308 | bool |
d7f3e0c5 GH |
1309 | parser::peek_kw (std::string const & kw) |
1310 | { | |
1311 | return tok_is (peek(), tok_identifier, kw); | |
1312 | } | |
1313 | ||
1314 | ||
1315 | ||
66c7d4c1 | 1316 | lexer::lexer (istream& input, const string& in, systemtap_session& s): |
03ba36d9 SM |
1317 | ate_comment(false), ate_whitespace(false), saw_tokens(false), |
1318 | input_name (in), input_pointer (0), input_end (0), cursor_suspend_count(0), | |
1319 | cursor_suspend_line (1), cursor_suspend_column (1), cursor_line (1), | |
1320 | cursor_column (1), session(s), current_file (0) | |
eacb10ce | 1321 | { |
66c7d4c1 | 1322 | getline(input, input_contents, '\0'); |
2203b032 | 1323 | |
66c7d4c1 JS |
1324 | input_pointer = input_contents.data(); |
1325 | input_end = input_contents.data() + input_contents.size(); | |
1326 | ||
1327 | if (keywords.empty()) | |
1328 | { | |
3a7ec735 FCE |
1329 | // NB: adding new keywords is highly disruptive to the language, |
1330 | // in particular to existing scripts that could be suddenly | |
1331 | // broken. If done at all, it has to be s.compatible-sensitive, | |
1332 | // and broadly advertised. | |
66c7d4c1 JS |
1333 | keywords.insert("probe"); |
1334 | keywords.insert("global"); | |
1335 | keywords.insert("function"); | |
1336 | keywords.insert("if"); | |
1337 | keywords.insert("else"); | |
1338 | keywords.insert("for"); | |
1339 | keywords.insert("foreach"); | |
1340 | keywords.insert("in"); | |
1341 | keywords.insert("limit"); | |
1342 | keywords.insert("return"); | |
1343 | keywords.insert("delete"); | |
1344 | keywords.insert("while"); | |
1345 | keywords.insert("break"); | |
1346 | keywords.insert("continue"); | |
1347 | keywords.insert("next"); | |
1348 | keywords.insert("string"); | |
1349 | keywords.insert("long"); | |
f4fe2e93 FCE |
1350 | keywords.insert("try"); |
1351 | keywords.insert("catch"); | |
66c7d4c1 | 1352 | } |
2524d1fd SM |
1353 | |
1354 | if (atwords.empty()) | |
1355 | { | |
1356 | // NB: adding new @words is mildly disruptive to existing | |
1357 | // scripts that define macros with the same name, but not | |
1358 | // really. The user will merely receive a warning that they are | |
1359 | // redefining an existing operator. | |
1360 | atwords.insert("@cast"); | |
1361 | atwords.insert("@defined"); | |
1362 | atwords.insert("@entry"); | |
1363 | atwords.insert("@var"); | |
1364 | atwords.insert("@avg"); | |
1365 | atwords.insert("@count"); | |
1366 | atwords.insert("@sum"); | |
1367 | atwords.insert("@min"); | |
1368 | atwords.insert("@max"); | |
1369 | atwords.insert("@hist_linear"); | |
1370 | atwords.insert("@hist_log"); | |
1371 | } | |
eacb10ce | 1372 | } |
2f1a1aea | 1373 | |
66c7d4c1 | 1374 | set<string> lexer::keywords; |
2524d1fd | 1375 | set<string> lexer::atwords; |
66c7d4c1 | 1376 | |
1b1b4ceb RA |
1377 | void |
1378 | lexer::set_current_file (stapfile* f) | |
1379 | { | |
1380 | current_file = f; | |
2203b032 JS |
1381 | if (f) |
1382 | { | |
1383 | f->file_contents = input_contents; | |
1384 | f->name = input_name; | |
1385 | } | |
1b1b4ceb | 1386 | } |
bb2e3076 FCE |
1387 | |
1388 | int | |
1389 | lexer::input_peek (unsigned n) | |
1390 | { | |
66c7d4c1 JS |
1391 | if (input_pointer + n >= input_end) |
1392 | return -1; // EOF | |
1393 | return (unsigned char)*(input_pointer + n); | |
bb2e3076 FCE |
1394 | } |
1395 | ||
1396 | ||
dff50e09 | 1397 | int |
2f1a1aea FCE |
1398 | lexer::input_get () |
1399 | { | |
66c7d4c1 | 1400 | int c = input_peek(); |
bb2e3076 FCE |
1401 | if (c < 0) return c; // EOF |
1402 | ||
66c7d4c1 JS |
1403 | ++input_pointer; |
1404 | ||
3f99432c | 1405 | if (cursor_suspend_count) |
9300f661 JS |
1406 | { |
1407 | // Track effect of input_put: preserve previous cursor/line_column | |
1408 | // until all of its characters are consumed. | |
1409 | if (--cursor_suspend_count == 0) | |
1410 | { | |
1411 | cursor_line = cursor_suspend_line; | |
1412 | cursor_column = cursor_suspend_column; | |
1413 | } | |
1414 | } | |
3f99432c | 1415 | else |
2f1a1aea | 1416 | { |
3f99432c FCE |
1417 | // update source cursor |
1418 | if (c == '\n') | |
1419 | { | |
1420 | cursor_line ++; | |
1421 | cursor_column = 1; | |
1422 | } | |
1423 | else | |
1424 | cursor_column ++; | |
2f1a1aea | 1425 | } |
2f1a1aea | 1426 | |
eacb10ce | 1427 | // clog << "[" << (char)c << "]"; |
2f1a1aea FCE |
1428 | return c; |
1429 | } | |
1430 | ||
1431 | ||
3f99432c | 1432 | void |
9300f661 | 1433 | lexer::input_put (const string& chars, const token* t) |
3f99432c | 1434 | { |
66c7d4c1 JS |
1435 | size_t pos = input_pointer - input_contents.data(); |
1436 | // clog << "[put:" << chars << " @" << pos << "]"; | |
1437 | input_contents.insert (pos, chars); | |
eacb10ce | 1438 | cursor_suspend_count += chars.size(); |
9300f661 JS |
1439 | cursor_suspend_line = cursor_line; |
1440 | cursor_suspend_column = cursor_column; | |
1441 | cursor_line = t->location.line; | |
1442 | cursor_column = t->location.column; | |
66c7d4c1 JS |
1443 | input_pointer = input_contents.data() + pos; |
1444 | input_end = input_contents.data() + input_contents.size(); | |
3f99432c FCE |
1445 | } |
1446 | ||
1447 | ||
2f1a1aea | 1448 | token* |
b5477cd9 | 1449 | lexer::scan () |
2f1a1aea | 1450 | { |
fee28e5c | 1451 | ate_comment = false; // reset for each new token |
b5477cd9 | 1452 | ate_whitespace = false; // reset for each new token |
534aad8b SM |
1453 | |
1454 | // XXX be very sure to restore old_saw_tokens if we return without a token: | |
1455 | bool old_saw_tokens = saw_tokens; | |
1456 | saw_tokens = true; | |
1457 | ||
2f1a1aea | 1458 | token* n = new token; |
2203b032 | 1459 | n->location.file = current_file; |
534aad8b | 1460 | n->chain = NULL; // important safety dance |
2f1a1aea | 1461 | |
9300f661 JS |
1462 | skip: |
1463 | bool suspended = (cursor_suspend_count > 0); | |
2f1a1aea FCE |
1464 | n->location.line = cursor_line; |
1465 | n->location.column = cursor_column; | |
1466 | ||
1467 | int c = input_get(); | |
3f99432c | 1468 | // clog << "{" << (char)c << (char)c2 << "}"; |
2f1a1aea FCE |
1469 | if (c < 0) |
1470 | { | |
1471 | delete n; | |
534aad8b | 1472 | saw_tokens = old_saw_tokens; |
2f1a1aea FCE |
1473 | return 0; |
1474 | } | |
1475 | ||
1476 | if (isspace (c)) | |
b5477cd9 SM |
1477 | { |
1478 | ate_whitespace = true; | |
1479 | goto skip; | |
1480 | } | |
2f1a1aea | 1481 | |
66c7d4c1 JS |
1482 | int c2 = input_peek (); |
1483 | ||
3f99432c FCE |
1484 | // Paste command line arguments as character streams into |
1485 | // the beginning of a token. $1..$999 go through as raw | |
1486 | // characters; @1..@999 are quoted/escaped as strings. | |
1487 | // $# and @# expand to the number of arguments, similarly | |
1488 | // raw or quoted. | |
9300f661 | 1489 | if ((c == '$' || c == '@') && (c2 == '#')) |
3f99432c | 1490 | { |
9300f661 JS |
1491 | n->content.push_back (c); |
1492 | n->content.push_back (c2); | |
3f99432c | 1493 | input_get(); // swallow '#' |
9300f661 | 1494 | if (suspended) |
16fc963f SM |
1495 | { |
1496 | n->make_junk(_("invalid nested substitution of command line arguments")); | |
1497 | return n; | |
1498 | } | |
9300f661 JS |
1499 | size_t num_args = session.args.size (); |
1500 | input_put ((c == '$') ? lex_cast (num_args) : lex_cast_qstring (num_args), n); | |
1501 | n->content.clear(); | |
1502 | goto skip; | |
3f99432c | 1503 | } |
9300f661 | 1504 | else if ((c == '$' || c == '@') && (isdigit (c2))) |
3f99432c | 1505 | { |
9300f661 | 1506 | n->content.push_back (c); |
3f99432c FCE |
1507 | unsigned idx = 0; |
1508 | do | |
1509 | { | |
1510 | input_get (); | |
1511 | idx = (idx * 10) + (c2 - '0'); | |
9300f661 | 1512 | n->content.push_back (c2); |
3f99432c FCE |
1513 | c2 = input_peek (); |
1514 | } while (c2 > 0 && | |
dff50e09 | 1515 | isdigit (c2) && |
3f99432c | 1516 | idx <= session.args.size()); // prevent overflow |
16fc963f SM |
1517 | if (suspended) |
1518 | { | |
1519 | n->make_junk(_("invalid nested substitution of command line arguments")); | |
1520 | return n; | |
1521 | } | |
3f99432c FCE |
1522 | if (idx == 0 || |
1523 | idx-1 >= session.args.size()) | |
16fc963f SM |
1524 | { |
1525 | n->make_junk(_F("command line argument index %lu out of range [1-%lu]", | |
1526 | (unsigned long) idx, (unsigned long) session.args.size())); | |
1527 | return n; | |
1528 | } | |
9300f661 JS |
1529 | const string& arg = session.args[idx-1]; |
1530 | input_put ((c == '$') ? arg : lex_cast_qstring (arg), n); | |
1531 | n->content.clear(); | |
1532 | goto skip; | |
3f99432c FCE |
1533 | } |
1534 | ||
b5477cd9 | 1535 | else if (isalpha (c) || c == '$' || c == '@' || c == '_') |
2f1a1aea FCE |
1536 | { |
1537 | n->type = tok_identifier; | |
1538 | n->content = (char) c; | |
b5477cd9 | 1539 | while (isalnum (c2) || c2 == '_' || c2 == '$') |
2f1a1aea | 1540 | { |
3f99432c FCE |
1541 | input_get (); |
1542 | n->content.push_back (c2); | |
1543 | c2 = input_peek (); | |
6e213f58 | 1544 | } |
213bee8f | 1545 | |
66c7d4c1 | 1546 | if (keywords.count(n->content)) |
3f99432c | 1547 | n->type = tok_keyword; |
06219d6f | 1548 | else if (n->content[0] == '@') |
dd90d565 | 1549 | // makes it easier to detect illegal use of @words: |
06219d6f | 1550 | n->type = tok_operator; |
dff50e09 | 1551 | |
2f1a1aea FCE |
1552 | return n; |
1553 | } | |
1554 | ||
3a20432b | 1555 | else if (isdigit (c)) // positive literal |
2f1a1aea | 1556 | { |
2f1a1aea | 1557 | n->type = tok_number; |
9c0c0e46 FCE |
1558 | n->content = (char) c; |
1559 | ||
66c7d4c1 | 1560 | while (isalnum (c2)) |
2f1a1aea | 1561 | { |
9c0c0e46 FCE |
1562 | // NB: isalnum is very permissive. We rely on strtol, called in |
1563 | // parser::parse_literal below, to confirm that the number string | |
1564 | // is correctly formatted and in range. | |
1565 | ||
66c7d4c1 JS |
1566 | input_get (); |
1567 | n->content.push_back (c2); | |
1568 | c2 = input_peek (); | |
2f1a1aea FCE |
1569 | } |
1570 | return n; | |
1571 | } | |
1572 | ||
1573 | else if (c == '\"') | |
1574 | { | |
1575 | n->type = tok_string; | |
1576 | while (1) | |
1577 | { | |
1578 | c = input_get (); | |
1579 | ||
3f99432c | 1580 | if (c < 0 || c == '\n') |
2f1a1aea | 1581 | { |
16fc963f SM |
1582 | n->make_junk(_("Could not find matching closing quote")); |
1583 | return n; | |
2f1a1aea FCE |
1584 | } |
1585 | if (c == '\"') // closing double-quotes | |
1586 | break; | |
3f99432c | 1587 | else if (c == '\\') // see also input_put |
dff50e09 | 1588 | { |
7d46afb8 GH |
1589 | c = input_get (); |
1590 | switch (c) | |
1591 | { | |
ef8a6134 SM |
1592 | case 'x': |
1593 | if (strverscmp(session.compatible.c_str(), "2.3") < 0) | |
1594 | goto the_default; | |
7d46afb8 GH |
1595 | case 'a': |
1596 | case 'b': | |
1597 | case 't': | |
1598 | case 'n': | |
1599 | case 'v': | |
1600 | case 'f': | |
1601 | case 'r': | |
f03954fd | 1602 | case '0' ... '7': // NB: need only match the first digit |
7d46afb8 | 1603 | case '\\': |
7d46afb8 | 1604 | // Pass these escapes through to the string value |
dff50e09 | 1605 | // being parsed; it will be emitted into a C literal. |
c7c8d469 FCE |
1606 | // XXX: PR13371: perhaps we should evaluate them here |
1607 | // (and re-quote them during translate.cxx emission). | |
7d46afb8 GH |
1608 | n->content.push_back('\\'); |
1609 | ||
3f99432c | 1610 | // fall through |
ef8a6134 SM |
1611 | default: the_default: |
1612 | n->content.push_back(c); | |
1613 | break; | |
7d46afb8 | 1614 | } |
2f1a1aea FCE |
1615 | } |
1616 | else | |
1617 | n->content.push_back(c); | |
1618 | } | |
1619 | return n; | |
1620 | } | |
1621 | ||
1622 | else if (ispunct (c)) | |
1623 | { | |
bb2e3076 | 1624 | int c3 = input_peek (1); |
2f1a1aea | 1625 | |
3a20432b FCE |
1626 | // NB: if we were to recognize negative numeric literals here, |
1627 | // we'd introduce another grammar ambiguity: | |
1628 | // 1-1 would be parsed as tok_number(1) and tok_number(-1) | |
1629 | // instead of tok_number(1) tok_operator('-') tok_number(1) | |
1630 | ||
66c7d4c1 | 1631 | if (c == '#') // shell comment |
2f1a1aea FCE |
1632 | { |
1633 | unsigned this_line = cursor_line; | |
bb2e3076 FCE |
1634 | do { c = input_get (); } |
1635 | while (c >= 0 && cursor_line == this_line); | |
fee28e5c | 1636 | ate_comment = true; |
b5477cd9 | 1637 | ate_whitespace = true; |
2f1a1aea FCE |
1638 | goto skip; |
1639 | } | |
66c7d4c1 | 1640 | else if ((c == '/' && c2 == '/')) // C++ comment |
63a7c90e FCE |
1641 | { |
1642 | unsigned this_line = cursor_line; | |
bb2e3076 FCE |
1643 | do { c = input_get (); } |
1644 | while (c >= 0 && cursor_line == this_line); | |
fee28e5c | 1645 | ate_comment = true; |
b5477cd9 | 1646 | ate_whitespace = true; |
63a7c90e FCE |
1647 | goto skip; |
1648 | } | |
1649 | else if (c == '/' && c2 == '*') // C comment | |
1650 | { | |
66c7d4c1 JS |
1651 | (void) input_get (); // swallow '*' already in c2 |
1652 | c = input_get (); | |
63a7c90e | 1653 | c2 = input_get (); |
bb2e3076 | 1654 | while (c2 >= 0) |
63a7c90e | 1655 | { |
66c7d4c1 JS |
1656 | if (c == '*' && c2 == '/') |
1657 | break; | |
63a7c90e FCE |
1658 | c = c2; |
1659 | c2 = input_get (); | |
63a7c90e | 1660 | } |
fee28e5c | 1661 | ate_comment = true; |
b5477cd9 | 1662 | ate_whitespace = true; |
bb2e3076 | 1663 | goto skip; |
63a7c90e | 1664 | } |
54dfabe9 FCE |
1665 | else if (c == '%' && c2 == '{') // embedded code |
1666 | { | |
1667 | n->type = tok_embedded; | |
1668 | (void) input_get (); // swallow '{' already in c2 | |
66c7d4c1 JS |
1669 | c = input_get (); |
1670 | c2 = input_get (); | |
1671 | while (c2 >= 0) | |
54dfabe9 | 1672 | { |
66c7d4c1 JS |
1673 | if (c == '%' && c2 == '}') |
1674 | return n; | |
ebbf9df4 FCE |
1675 | if (c == '}' && c2 == '%') // possible typo |
1676 | session.print_warning (_("possible erroneous closing '}%', use '%}'?"), n); | |
54dfabe9 | 1677 | n->content += c; |
66c7d4c1 JS |
1678 | c = c2; |
1679 | c2 = input_get (); | |
54dfabe9 | 1680 | } |
72cdb9cd | 1681 | |
16fc963f SM |
1682 | n->make_junk(_("Could not find matching '%}' to close embedded function block")); |
1683 | return n; | |
54dfabe9 | 1684 | } |
2f1a1aea | 1685 | |
bb2e3076 FCE |
1686 | // We're committed to recognizing at least the first character |
1687 | // as an operator. | |
2f1a1aea | 1688 | n->type = tok_operator; |
66c7d4c1 | 1689 | n->content = c; |
2f1a1aea | 1690 | |
bb2e3076 | 1691 | // match all valid operators, in decreasing size order |
66c7d4c1 JS |
1692 | if ((c == '<' && c2 == '<' && c3 == '<') || |
1693 | (c == '<' && c2 == '<' && c3 == '=') || | |
1694 | (c == '>' && c2 == '>' && c3 == '=')) | |
82919855 | 1695 | { |
66c7d4c1 JS |
1696 | n->content += c2; |
1697 | n->content += c3; | |
bb2e3076 FCE |
1698 | input_get (); input_get (); // swallow other two characters |
1699 | } | |
66c7d4c1 JS |
1700 | else if ((c == '=' && c2 == '=') || |
1701 | (c == '!' && c2 == '=') || | |
1702 | (c == '<' && c2 == '=') || | |
1703 | (c == '>' && c2 == '=') || | |
93daaca8 SM |
1704 | (c == '=' && c2 == '~') || |
1705 | (c == '!' && c2 == '~') || | |
66c7d4c1 JS |
1706 | (c == '+' && c2 == '=') || |
1707 | (c == '-' && c2 == '=') || | |
1708 | (c == '*' && c2 == '=') || | |
1709 | (c == '/' && c2 == '=') || | |
1710 | (c == '%' && c2 == '=') || | |
1711 | (c == '&' && c2 == '=') || | |
1712 | (c == '^' && c2 == '=') || | |
1713 | (c == '|' && c2 == '=') || | |
1714 | (c == '.' && c2 == '=') || | |
1715 | (c == '&' && c2 == '&') || | |
1716 | (c == '|' && c2 == '|') || | |
1717 | (c == '+' && c2 == '+') || | |
1718 | (c == '-' && c2 == '-') || | |
1719 | (c == '-' && c2 == '>') || | |
1720 | (c == '<' && c2 == '<') || | |
1721 | (c == '>' && c2 == '>') || | |
177a8ead | 1722 | // preprocessor tokens |
66c7d4c1 JS |
1723 | (c == '%' && c2 == '(') || |
1724 | (c == '%' && c2 == '?') || | |
1725 | (c == '%' && c2 == ':') || | |
1726 | (c == '%' && c2 == ')')) | |
bb2e3076 | 1727 | { |
66c7d4c1 | 1728 | n->content += c2; |
bb2e3076 | 1729 | input_get (); // swallow other character |
dff50e09 | 1730 | } |
2f1a1aea FCE |
1731 | |
1732 | return n; | |
1733 | } | |
1734 | ||
1735 | else | |
1736 | { | |
1737 | n->type = tok_junk; | |
e3795795 FCE |
1738 | ostringstream s; |
1739 | s << "\\x" << hex << setw(2) << setfill('0') << c; | |
1740 | n->content = s.str(); | |
16fc963f | 1741 | n->msg = ""; // signal parser to emit "expected X, found junk" type error |
2f1a1aea FCE |
1742 | return n; |
1743 | } | |
1744 | } | |
1745 | ||
16fc963f SM |
1746 | // ------------------------------------------------------------------------ |
1747 | ||
1748 | void | |
1749 | token::make_junk (const string new_msg) | |
1750 | { | |
1751 | type = tok_junk; | |
1752 | msg = new_msg; | |
1753 | } | |
2f1a1aea FCE |
1754 | |
1755 | // ------------------------------------------------------------------------ | |
1756 | ||
1757 | stapfile* | |
1758 | parser::parse () | |
1759 | { | |
1760 | stapfile* f = new stapfile; | |
1b1b4ceb | 1761 | input.set_current_file (f); |
56099f08 FCE |
1762 | |
1763 | bool empty = true; | |
1764 | ||
2f1a1aea FCE |
1765 | while (1) |
1766 | { | |
1767 | try | |
1768 | { | |
a07a2c28 | 1769 | systemtap_v_seen = 0; |
2f1a1aea | 1770 | const token* t = peek (); |
534aad8b | 1771 | if (! t) // nice clean EOF, modulo any preprocessing that occurred |
2f1a1aea FCE |
1772 | break; |
1773 | ||
56099f08 | 1774 | empty = false; |
6e213f58 DS |
1775 | if (t->type == tok_keyword && t->content == "probe") |
1776 | { | |
1777 | context = con_probe; | |
1778 | parse_probe (f->probes, f->aliases); | |
1779 | } | |
1780 | else if (t->type == tok_keyword && t->content == "global") | |
1781 | { | |
1782 | context = con_global; | |
4b5f3e45 | 1783 | parse_global (f->globals, f->probes); |
6e213f58 DS |
1784 | } |
1785 | else if (t->type == tok_keyword && t->content == "function") | |
1786 | { | |
1787 | context = con_function; | |
1788 | parse_functiondecl (f->functions); | |
1789 | } | |
54dfabe9 | 1790 | else if (t->type == tok_embedded) |
6e213f58 DS |
1791 | { |
1792 | context = con_embedded; | |
1793 | f->embeds.push_back (parse_embeddedcode ()); | |
1794 | } | |
2f1a1aea | 1795 | else |
6e213f58 DS |
1796 | { |
1797 | context = con_unknown; | |
2677d2fb | 1798 | throw parse_error (_("expected 'probe', 'global', 'function', or '%{'")); |
6e213f58 | 1799 | } |
2f1a1aea FCE |
1800 | } |
1801 | catch (parse_error& pe) | |
1802 | { | |
1803 | print_error (pe); | |
16fc963f SM |
1804 | |
1805 | // XXX: do we want tok_junk to be able to force skip_some behaviour? | |
cd7116b8 | 1806 | if (pe.skip_some) // for recovery |
46954f1d FCE |
1807 | // Quietly swallow all tokens until the next keyword we can start parsing from. |
1808 | while (1) | |
1809 | try | |
1810 | { | |
cd7116b8 FCE |
1811 | { |
1812 | const token* t = peek (); | |
1813 | if (! t) | |
1814 | break; | |
46954f1d FCE |
1815 | if (t->type == tok_keyword && t->content == "probe") break; |
1816 | else if (t->type == tok_keyword && t->content == "global") break; | |
1817 | else if (t->type == tok_keyword && t->content == "function") break; | |
1818 | else if (t->type == tok_embedded) break; | |
731a5359 | 1819 | swallow (); // swallow it |
cd7116b8 | 1820 | } |
46954f1d FCE |
1821 | } |
1822 | catch (parse_error& pe2) | |
1823 | { | |
1824 | // parse error during recovery ... ugh | |
1825 | print_error (pe2); | |
1826 | } | |
177a8ead | 1827 | } |
2f1a1aea FCE |
1828 | } |
1829 | ||
56099f08 FCE |
1830 | if (empty) |
1831 | { | |
534aad8b SM |
1832 | // vary message depending on whether file was *actually* empty: |
1833 | cerr << (input.saw_tokens | |
1834 | ? _F("Input file '%s' is empty after preprocessing.", input_name.c_str()) | |
4cd32d8c | 1835 | : _F("Input file '%s' is empty.", input_name.c_str())) |
534aad8b | 1836 | << endl; |
56099f08 | 1837 | delete f; |
2203b032 | 1838 | f = 0; |
56099f08 FCE |
1839 | } |
1840 | else if (num_errors > 0) | |
2f1a1aea | 1841 | { |
52c2652f | 1842 | cerr << _NF("%d parse error.", "%d parse errors.", num_errors, num_errors) << endl; |
2f1a1aea | 1843 | delete f; |
2203b032 | 1844 | f = 0; |
2f1a1aea | 1845 | } |
dff50e09 | 1846 | |
2203b032 | 1847 | input.set_current_file(0); |
2f1a1aea FCE |
1848 | return f; |
1849 | } | |
1850 | ||
1851 | ||
20c6c071 | 1852 | void |
54dfabe9 FCE |
1853 | parser::parse_probe (std::vector<probe *> & probe_ret, |
1854 | std::vector<probe_alias *> & alias_ret) | |
2f1a1aea | 1855 | { |
82919855 | 1856 | const token* t0 = next (); |
6e213f58 | 1857 | if (! (t0->type == tok_keyword && t0->content == "probe")) |
2677d2fb | 1858 | throw parse_error (_("expected 'probe'")); |
82919855 | 1859 | |
20c6c071 GH |
1860 | vector<probe_point *> aliases; |
1861 | vector<probe_point *> locations; | |
1862 | ||
1863 | bool equals_ok = true; | |
82919855 | 1864 | |
97266278 LG |
1865 | int epilogue_alias = 0; |
1866 | ||
2f1a1aea FCE |
1867 | while (1) |
1868 | { | |
b4ceace2 | 1869 | probe_point * pp = parse_probe_point (); |
dff50e09 | 1870 | |
b4ceace2 | 1871 | const token* t = peek (); |
dff50e09 | 1872 | if (equals_ok && t |
b4ceace2 FCE |
1873 | && t->type == tok_operator && t->content == "=") |
1874 | { | |
1ad820e3 | 1875 | if (pp->optional || pp->sufficient) |
2677d2fb | 1876 | throw parse_error (_("probe point alias name cannot be optional nor sufficient"), pp->components.front()->tok); |
b4ceace2 | 1877 | aliases.push_back(pp); |
731a5359 | 1878 | swallow (); |
b4ceace2 FCE |
1879 | continue; |
1880 | } | |
dff50e09 | 1881 | else if (equals_ok && t |
97266278 LG |
1882 | && t->type == tok_operator && t->content == "+=") |
1883 | { | |
1ad820e3 | 1884 | if (pp->optional || pp->sufficient) |
2677d2fb | 1885 | throw parse_error (_("probe point alias name cannot be optional nor sufficient"), pp->components.front()->tok); |
97266278 LG |
1886 | aliases.push_back(pp); |
1887 | epilogue_alias = 1; | |
731a5359 | 1888 | swallow (); |
97266278 LG |
1889 | continue; |
1890 | } | |
b4ceace2 FCE |
1891 | else if (t && t->type == tok_operator && t->content == ",") |
1892 | { | |
1893 | locations.push_back(pp); | |
1894 | equals_ok = false; | |
731a5359 | 1895 | swallow (); |
b4ceace2 FCE |
1896 | continue; |
1897 | } | |
1898 | else if (t && t->type == tok_operator && t->content == "{") | |
1899 | { | |
1900 | locations.push_back(pp); | |
1901 | break; | |
1902 | } | |
2f1a1aea | 1903 | else |
2677d2fb | 1904 | throw parse_error (_("expected probe point specifier")); |
2f1a1aea | 1905 | } |
20c6c071 | 1906 | |
20c6c071 GH |
1907 | if (aliases.empty()) |
1908 | { | |
54dfabe9 FCE |
1909 | probe* p = new probe; |
1910 | p->tok = t0; | |
1911 | p->locations = locations; | |
1912 | p->body = parse_stmt_block (); | |
37ebca01 | 1913 | p->privileged = privileged; |
a07a2c28 | 1914 | p->systemtap_v_conditional = systemtap_v_seen; |
54dfabe9 | 1915 | probe_ret.push_back (p); |
20c6c071 GH |
1916 | } |
1917 | else | |
1918 | { | |
54dfabe9 | 1919 | probe_alias* p = new probe_alias (aliases); |
97266278 LG |
1920 | if(epilogue_alias) |
1921 | p->epilogue_style = true; | |
1922 | else | |
1923 | p->epilogue_style = false; | |
54dfabe9 FCE |
1924 | p->tok = t0; |
1925 | p->locations = locations; | |
1926 | p->body = parse_stmt_block (); | |
37ebca01 | 1927 | p->privileged = privileged; |
a07a2c28 | 1928 | p->systemtap_v_conditional = systemtap_v_seen; |
54dfabe9 | 1929 | alias_ret.push_back (p); |
20c6c071 | 1930 | } |
54dfabe9 | 1931 | } |
20c6c071 | 1932 | |
54dfabe9 FCE |
1933 | |
1934 | embeddedcode* | |
1935 | parser::parse_embeddedcode () | |
1936 | { | |
1937 | embeddedcode* e = new embeddedcode; | |
1938 | const token* t = next (); | |
1939 | if (t->type != tok_embedded) | |
2677d2fb | 1940 | throw parse_error (_("expected '%{'")); |
24cb178f FCE |
1941 | |
1942 | if (! privileged) | |
efb02738 | 1943 | throw parse_error (_("embedded code in unprivileged script; need stap -g"), |
cd7116b8 | 1944 | false /* don't skip tokens for parse resumption */); |
54dfabe9 FCE |
1945 | |
1946 | e->tok = t; | |
1947 | e->code = t->content; | |
1948 | return e; | |
2f1a1aea FCE |
1949 | } |
1950 | ||
1951 | ||
1952 | block* | |
56099f08 | 1953 | parser::parse_stmt_block () |
2f1a1aea FCE |
1954 | { |
1955 | block* pb = new block; | |
1956 | ||
56099f08 FCE |
1957 | const token* t = next (); |
1958 | if (! (t->type == tok_operator && t->content == "{")) | |
2677d2fb | 1959 | throw parse_error (_("expected '{'")); |
56099f08 FCE |
1960 | |
1961 | pb->tok = t; | |
2b066ec1 | 1962 | |
2f1a1aea FCE |
1963 | while (1) |
1964 | { | |
46954f1d FCE |
1965 | t = peek (); |
1966 | if (t && t->type == tok_operator && t->content == "}") | |
1967 | { | |
731a5359 | 1968 | swallow (); |
46954f1d FCE |
1969 | break; |
1970 | } | |
1971 | pb->statements.push_back (parse_statement ()); | |
2f1a1aea FCE |
1972 | } |
1973 | ||
1974 | return pb; | |
1975 | } | |
1976 | ||
1977 | ||
f4fe2e93 FCE |
1978 | try_block* |
1979 | parser::parse_try_block () | |
1980 | { | |
1981 | try_block* pb = new try_block; | |
1982 | ||
731a5359 | 1983 | pb->tok = expect_kw_token ("try"); |
f4fe2e93 FCE |
1984 | pb->try_block = parse_stmt_block(); |
1985 | expect_kw ("catch"); | |
1986 | ||
1987 | const token* t = peek (); | |
3819d181 | 1988 | if (t != NULL && t->type == tok_operator && t->content == "(") |
f4fe2e93 | 1989 | { |
731a5359 | 1990 | swallow (); // swallow the '(' |
f4fe2e93 FCE |
1991 | |
1992 | t = next(); | |
1993 | if (! (t->type == tok_identifier)) | |
2677d2fb | 1994 | throw parse_error (_("expected identifier")); |
f4fe2e93 FCE |
1995 | symbol* sym = new symbol; |
1996 | sym->tok = t; | |
1997 | sym->name = t->content; | |
1998 | pb->catch_error_var = sym; | |
1999 | ||
2000 | expect_op (")"); | |
2001 | } | |
2002 | else | |
2003 | pb->catch_error_var = 0; | |
2004 | ||
2005 | pb->catch_block = parse_stmt_block(); | |
2006 | ||
2007 | return pb; | |
2008 | } | |
2009 | ||
2010 | ||
2011 | ||
2f1a1aea FCE |
2012 | statement* |
2013 | parser::parse_statement () | |
2014 | { | |
40b71c47 | 2015 | statement *ret; |
2f1a1aea FCE |
2016 | const token* t = peek (); |
2017 | if (t && t->type == tok_operator && t->content == ";") | |
f946b10f | 2018 | return new null_statement (next ()); |
dff50e09 | 2019 | else if (t && t->type == tok_operator && t->content == "{") |
40b71c47 | 2020 | return parse_stmt_block (); // Don't squash semicolons. |
f4fe2e93 FCE |
2021 | else if (t && t->type == tok_keyword && t->content == "try") |
2022 | return parse_try_block (); // Don't squash semicolons. | |
6e213f58 | 2023 | else if (t && t->type == tok_keyword && t->content == "if") |
40b71c47 | 2024 | return parse_if_statement (); // Don't squash semicolons. |
6e213f58 | 2025 | else if (t && t->type == tok_keyword && t->content == "for") |
40b71c47 | 2026 | return parse_for_loop (); // Don't squash semicolons. |
6e213f58 | 2027 | else if (t && t->type == tok_keyword && t->content == "foreach") |
40b71c47 MW |
2028 | return parse_foreach_loop (); // Don't squash semicolons. |
2029 | else if (t && t->type == tok_keyword && t->content == "while") | |
2030 | return parse_while_loop (); // Don't squash semicolons. | |
6e213f58 | 2031 | else if (t && t->type == tok_keyword && t->content == "return") |
40b71c47 | 2032 | ret = parse_return_statement (); |
6e213f58 | 2033 | else if (t && t->type == tok_keyword && t->content == "delete") |
40b71c47 | 2034 | ret = parse_delete_statement (); |
6e213f58 | 2035 | else if (t && t->type == tok_keyword && t->content == "break") |
40b71c47 | 2036 | ret = parse_break_statement (); |
6e213f58 | 2037 | else if (t && t->type == tok_keyword && t->content == "continue") |
40b71c47 | 2038 | ret = parse_continue_statement (); |
6e213f58 | 2039 | else if (t && t->type == tok_keyword && t->content == "next") |
40b71c47 | 2040 | ret = parse_next_statement (); |
2f1a1aea FCE |
2041 | else if (t && (t->type == tok_operator || // expressions are flexible |
2042 | t->type == tok_identifier || | |
2043 | t->type == tok_number || | |
7d902887 FCE |
2044 | t->type == tok_string || |
2045 | t->type == tok_embedded )) | |
40b71c47 | 2046 | ret = parse_expr_statement (); |
54dfabe9 | 2047 | // XXX: consider generally accepting tok_embedded here too |
2f1a1aea | 2048 | else |
2677d2fb | 2049 | throw parse_error (_("expected statement")); |
40b71c47 MW |
2050 | |
2051 | // Squash "empty" trailing colons after any "non-block-like" statement. | |
2052 | t = peek (); | |
2053 | if (t && t->type == tok_operator && t->content == ";") | |
2054 | { | |
731a5359 | 2055 | swallow (); // Silently eat trailing ; after statement |
40b71c47 MW |
2056 | } |
2057 | ||
2058 | return ret; | |
2f1a1aea FCE |
2059 | } |
2060 | ||
2061 | ||
56099f08 | 2062 | void |
78f6bba6 | 2063 | parser::parse_global (vector <vardecl*>& globals, vector<probe*>&) |
2f1a1aea | 2064 | { |
82919855 | 2065 | const token* t0 = next (); |
6e213f58 | 2066 | if (! (t0->type == tok_keyword && t0->content == "global")) |
2677d2fb | 2067 | throw parse_error (_("expected 'global'")); |
731a5359 | 2068 | swallow (); |
82919855 | 2069 | |
56099f08 FCE |
2070 | while (1) |
2071 | { | |
2072 | const token* t = next (); | |
2073 | if (! (t->type == tok_identifier)) | |
2677d2fb | 2074 | throw parse_error (_("expected identifier")); |
56099f08 | 2075 | |
2b066ec1 FCE |
2076 | for (unsigned i=0; i<globals.size(); i++) |
2077 | if (globals[i]->name == t->content) | |
2677d2fb | 2078 | throw parse_error (_("duplicate global name")); |
dff50e09 | 2079 | |
24cb178f FCE |
2080 | vardecl* d = new vardecl; |
2081 | d->name = t->content; | |
2082 | d->tok = t; | |
a07a2c28 | 2083 | d->systemtap_v_conditional = systemtap_v_seen; |
24cb178f | 2084 | globals.push_back (d); |
56099f08 | 2085 | |
82919855 | 2086 | t = peek (); |
ef474d24 | 2087 | |
74e6cc92 CM |
2088 | if(t && t->type == tok_operator && t->content == "%") //wrapping |
2089 | { | |
2090 | d->wrap = true; | |
731a5359 | 2091 | swallow (); |
74e6cc92 CM |
2092 | t = peek(); |
2093 | } | |
2094 | ||
ef474d24 JS |
2095 | if (t && t->type == tok_operator && t->content == "[") // array size |
2096 | { | |
2097 | int64_t size; | |
731a5359 | 2098 | swallow (); |
ef474d24 JS |
2099 | expect_number(size); |
2100 | if (size <= 0 || size > 1000000) // arbitrary max | |
2677d2fb | 2101 | throw parse_error(_("array size out of range")); |
ef474d24 JS |
2102 | d->maxsize = (int)size; |
2103 | expect_known(tok_operator, "]"); | |
2104 | t = peek (); | |
2105 | } | |
2106 | ||
4b5f3e45 | 2107 | if (t && t->type == tok_operator && t->content == "=") // initialization |
ef474d24 JS |
2108 | { |
2109 | if (!d->compatible_arity(0)) | |
2677d2fb | 2110 | throw parse_error(_("only scalar globals can be initialized")); |
58701b78 | 2111 | d->set_arity(0, t); |
731a5359 | 2112 | next (); // Don't swallow, set_arity() used the peeked token. |
ef474d24 JS |
2113 | d->init = parse_literal (); |
2114 | d->type = d->init->type; | |
2115 | t = peek (); | |
2116 | } | |
4b5f3e45 | 2117 | |
c3799d72 | 2118 | if (t && t->type == tok_operator && t->content == ";") // termination |
950da622 | 2119 | { |
731a5359 | 2120 | swallow (); |
950da622 MW |
2121 | break; |
2122 | } | |
c3799d72 | 2123 | |
4b5f3e45 | 2124 | if (t && t->type == tok_operator && t->content == ",") // next global |
82919855 | 2125 | { |
731a5359 | 2126 | swallow (); |
82919855 FCE |
2127 | continue; |
2128 | } | |
56099f08 | 2129 | else |
82919855 | 2130 | break; |
56099f08 FCE |
2131 | } |
2132 | } | |
2133 | ||
2134 | ||
24cb178f FCE |
2135 | void |
2136 | parser::parse_functiondecl (std::vector<functiondecl*>& functions) | |
56099f08 | 2137 | { |
82919855 | 2138 | const token* t = next (); |
6e213f58 | 2139 | if (! (t->type == tok_keyword && t->content == "function")) |
2677d2fb | 2140 | throw parse_error (_("expected 'function'")); |
731a5359 | 2141 | swallow (); |
56099f08 | 2142 | |
82919855 | 2143 | t = next (); |
6e213f58 DS |
2144 | if (! (t->type == tok_identifier) |
2145 | && ! (t->type == tok_keyword | |
2146 | && (t->content == "string" || t->content == "long"))) | |
2677d2fb | 2147 | throw parse_error (_("expected identifier")); |
24cb178f FCE |
2148 | |
2149 | for (unsigned i=0; i<functions.size(); i++) | |
2150 | if (functions[i]->name == t->content) | |
2677d2fb | 2151 | throw parse_error (_("duplicate function name")); |
24cb178f FCE |
2152 | |
2153 | functiondecl *fd = new functiondecl (); | |
56099f08 FCE |
2154 | fd->name = t->content; |
2155 | fd->tok = t; | |
2156 | ||
2157 | t = next (); | |
6a505121 FCE |
2158 | if (t->type == tok_operator && t->content == ":") |
2159 | { | |
731a5359 | 2160 | swallow (); |
6a505121 | 2161 | t = next (); |
6e213f58 | 2162 | if (t->type == tok_keyword && t->content == "string") |
6a505121 | 2163 | fd->type = pe_string; |
6e213f58 | 2164 | else if (t->type == tok_keyword && t->content == "long") |
6a505121 | 2165 | fd->type = pe_long; |
2677d2fb | 2166 | else throw parse_error (_("expected 'string' or 'long'")); |
731a5359 | 2167 | swallow (); |
6a505121 FCE |
2168 | |
2169 | t = next (); | |
2170 | } | |
2171 | ||
56099f08 | 2172 | if (! (t->type == tok_operator && t->content == "(")) |
2677d2fb | 2173 | throw parse_error (_("expected '('")); |
731a5359 | 2174 | swallow (); |
56099f08 FCE |
2175 | |
2176 | while (1) | |
2177 | { | |
2178 | t = next (); | |
2179 | ||
100a540e | 2180 | // permit zero-argument functions |
56099f08 | 2181 | if (t->type == tok_operator && t->content == ")") |
731a5359 MW |
2182 | { |
2183 | swallow (); | |
2184 | break; | |
2185 | } | |
56099f08 | 2186 | else if (! (t->type == tok_identifier)) |
2677d2fb | 2187 | throw parse_error (_("expected identifier")); |
56099f08 FCE |
2188 | vardecl* vd = new vardecl; |
2189 | vd->name = t->content; | |
2190 | vd->tok = t; | |
2191 | fd->formal_args.push_back (vd); | |
a07a2c28 | 2192 | fd->systemtap_v_conditional = systemtap_v_seen; |
56099f08 FCE |
2193 | |
2194 | t = next (); | |
6a505121 FCE |
2195 | if (t->type == tok_operator && t->content == ":") |
2196 | { | |
731a5359 | 2197 | swallow (); |
6a505121 | 2198 | t = next (); |
6e213f58 | 2199 | if (t->type == tok_keyword && t->content == "string") |
6a505121 | 2200 | vd->type = pe_string; |
6e213f58 | 2201 | else if (t->type == tok_keyword && t->content == "long") |
6a505121 | 2202 | vd->type = pe_long; |
2677d2fb | 2203 | else throw parse_error (_("expected 'string' or 'long'")); |
731a5359 | 2204 | swallow (); |
6a505121 FCE |
2205 | t = next (); |
2206 | } | |
56099f08 | 2207 | if (t->type == tok_operator && t->content == ")") |
731a5359 MW |
2208 | { |
2209 | swallow (); | |
2210 | break; | |
2211 | } | |
56099f08 | 2212 | if (t->type == tok_operator && t->content == ",") |
731a5359 MW |
2213 | { |
2214 | swallow (); | |
2215 | continue; | |
2216 | } | |
56099f08 | 2217 | else |
2677d2fb | 2218 | throw parse_error (_("expected ',' or ')'")); |
56099f08 FCE |
2219 | } |
2220 | ||
54dfabe9 FCE |
2221 | t = peek (); |
2222 | if (t && t->type == tok_embedded) | |
2223 | fd->body = parse_embeddedcode (); | |
2224 | else | |
2225 | fd->body = parse_stmt_block (); | |
24cb178f FCE |
2226 | |
2227 | functions.push_back (fd); | |
2f1a1aea FCE |
2228 | } |
2229 | ||
2230 | ||
9c0c0e46 FCE |
2231 | probe_point* |
2232 | parser::parse_probe_point () | |
2f1a1aea | 2233 | { |
9c0c0e46 | 2234 | probe_point* pl = new probe_point; |
2f1a1aea | 2235 | |
9c0c0e46 | 2236 | while (1) |
2f1a1aea | 2237 | { |
b5477cd9 | 2238 | const token* t = next (); |
6e213f58 DS |
2239 | if (! (t->type == tok_identifier |
2240 | // we must allow ".return" and ".function", which are keywords | |
b5477cd9 SM |
2241 | || t->type == tok_keyword |
2242 | // we must allow "*", due to being an operator | |
2243 | || (t->type == tok_operator && t->content == "*"))) | |
2677d2fb | 2244 | throw parse_error (_("expected identifier or '*'")); |
9c0c0e46 | 2245 | |
b5477cd9 SM |
2246 | // loop which reconstitutes an identifier with wildcards |
2247 | string content = t->content; | |
2248 | while (1) | |
2249 | { | |
2250 | const token* u = peek(); | |
3819d181 MW |
2251 | if (u == NULL) |
2252 | break; | |
b5477cd9 SM |
2253 | // ensure pieces of the identifier are adjacent: |
2254 | if (input.ate_whitespace) | |
2255 | break; | |
2256 | // ensure pieces of the identifier are valid: | |
2257 | if (! (u->type == tok_identifier | |
2258 | // we must allow arbitrary keywords with a wildcard | |
2259 | || u->type == tok_keyword | |
2260 | // we must allow "*", due to being an operator | |
2261 | || (u->type == tok_operator && u->content == "*"))) | |
2262 | break; | |
2263 | ||
2264 | // append u to t | |
2265 | content = content + u->content; | |
2266 | ||
2267 | // consume u | |
731a5359 | 2268 | swallow (); |
b5477cd9 | 2269 | } |
534aad8b SM |
2270 | // get around const-ness of t: |
2271 | token* new_t = new token(*t); | |
b5477cd9 SM |
2272 | new_t->content = content; |
2273 | delete t; t = new_t; | |
9c0c0e46 FCE |
2274 | |
2275 | probe_point::component* c = new probe_point::component; | |
2276 | c->functor = t->content; | |
f1a0157a | 2277 | c->tok = t; |
9c0c0e46 | 2278 | pl->components.push_back (c); |
6e3347a9 | 2279 | // NB we may add c->arg soon |
9c0c0e46 FCE |
2280 | |
2281 | t = peek (); | |
a477f3f1 | 2282 | |
6e3347a9 | 2283 | // consume optional parameter |
9c0c0e46 FCE |
2284 | if (t && t->type == tok_operator && t->content == "(") |
2285 | { | |
731a5359 | 2286 | swallow (); // consume "(" |
9c0c0e46 FCE |
2287 | c->arg = parse_literal (); |
2288 | ||
2289 | t = next (); | |
2290 | if (! (t->type == tok_operator && t->content == ")")) | |
2677d2fb | 2291 | throw parse_error (_("expected ')'")); |
731a5359 | 2292 | swallow (); |
9c0c0e46 FCE |
2293 | |
2294 | t = peek (); | |
9c0c0e46 | 2295 | } |
9c0c0e46 FCE |
2296 | |
2297 | if (t && t->type == tok_operator && t->content == ".") | |
6e3347a9 | 2298 | { |
731a5359 | 2299 | swallow (); |
6e3347a9 FCE |
2300 | continue; |
2301 | } | |
2302 | ||
f1a0157a | 2303 | // We only fall through here at the end of a probe point (past |
6e3347a9 FCE |
2304 | // all the dotted/parametrized components). |
2305 | ||
d898100a FCE |
2306 | if (t && t->type == tok_operator && |
2307 | (t->content == "?" || t->content == "!")) | |
6e3347a9 FCE |
2308 | { |
2309 | pl->optional = true; | |
d898100a FCE |
2310 | if (t->content == "!") pl->sufficient = true; |
2311 | // NB: sufficient implies optional | |
731a5359 | 2312 | swallow (); |
6e3347a9 FCE |
2313 | t = peek (); |
2314 | // fall through | |
cbbe8080 MH |
2315 | } |
2316 | ||
2317 | if (t && t->type == tok_keyword && t->content == "if") | |
2318 | { | |
731a5359 | 2319 | swallow (); |
cbbe8080 | 2320 | t = peek (); |
00917a8a | 2321 | if (!(t && t->type == tok_operator && t->content == "(")) |
2677d2fb | 2322 | throw parse_error (_("expected '('")); |
731a5359 | 2323 | swallow (); |
cbbe8080 MH |
2324 | |
2325 | pl->condition = parse_expression (); | |
2326 | ||
2327 | t = peek (); | |
00917a8a | 2328 | if (!(t && t->type == tok_operator && t->content == ")")) |
2677d2fb | 2329 | throw parse_error (_("expected ')'")); |
731a5359 | 2330 | swallow (); |
cbbe8080 MH |
2331 | t = peek (); |
2332 | // fall through | |
6e3347a9 FCE |
2333 | } |
2334 | ||
dff50e09 | 2335 | if (t && t->type == tok_operator |
6e3347a9 FCE |
2336 | && (t->content == "{" || t->content == "," || |
2337 | t->content == "=" || t->content == "+=" )) | |
2338 | break; | |
dff50e09 | 2339 | |
2677d2fb | 2340 | throw parse_error (_("expected one of '. , ( ? ! { = +='")); |
2f1a1aea FCE |
2341 | } |
2342 | ||
2343 | return pl; | |
2344 | } | |
2345 | ||
2346 | ||
d24f1ff4 SM |
2347 | literal_string* |
2348 | parser::consume_string_literals(const token *t) | |
2349 | { | |
2350 | literal_string *ls = new literal_string (t->content); | |
2351 | ||
2352 | // PR11208: check if the next token is also a string literal; | |
2353 | // auto-concatenate it. This is complicated to the extent that we | |
2354 | // need to skip intermediate whitespace. | |
2355 | // | |
2356 | // NB for versions prior to 2.0: but don't skip over intervening comments | |
2357 | const token *n = peek(); | |
2358 | while (n != NULL && n->type == tok_string | |
2359 | && ! (strverscmp(session.compatible.c_str(), "2.0") < 0 | |
2360 | && input.ate_comment)) | |
2361 | { | |
2362 | ls->value.append(next()->content); // consume and append the token | |
2363 | n = peek(); | |
2364 | } | |
2365 | return ls; | |
2366 | } | |
2367 | ||
2368 | ||
2369 | // Parse a string literal and perform backslash escaping on the contents: | |
2370 | literal_string* | |
2371 | parser::parse_literal_string () | |
2372 | { | |
2373 | const token* t = next (); | |
2374 | literal_string* l; | |
2375 | if (t->type == tok_string) | |
2376 | l = consume_string_literals (t); | |
2377 | else | |
6a420ae9 | 2378 | throw parse_error (_("expected literal string")); |
d24f1ff4 SM |
2379 | |
2380 | l->tok = t; | |
2381 | return l; | |
2382 | } | |
2383 | ||
2384 | ||
2f1a1aea FCE |
2385 | literal* |
2386 | parser::parse_literal () | |
2387 | { | |
2388 | const token* t = next (); | |
56099f08 | 2389 | literal* l; |
2f1a1aea | 2390 | if (t->type == tok_string) |
c5be7511 | 2391 | { |
d24f1ff4 | 2392 | l = consume_string_literals (t); |
c5be7511 | 2393 | } |
16e8f21f | 2394 | else |
9c0c0e46 | 2395 | { |
16e8f21f JS |
2396 | bool neg = false; |
2397 | if (t->type == tok_operator && t->content == "-") | |
2398 | { | |
2399 | neg = true; | |
731a5359 | 2400 | swallow (); |
16e8f21f JS |
2401 | t = next (); |
2402 | } | |
2403 | ||
2404 | if (t->type == tok_number) | |
2405 | { | |
2406 | const char* startp = t->content.c_str (); | |
2407 | char* endp = (char*) startp; | |
2408 | ||
2409 | // NB: we allow controlled overflow from LLONG_MIN .. ULLONG_MAX | |
2410 | // Actually, this allows all the way from -ULLONG_MAX to ULLONG_MAX, | |
79e6d33f JS |
2411 | // since the lexer only gives us positive digit strings, but we'll |
2412 | // limit it to LLONG_MIN when a '-' operator is fed into the literal. | |
16e8f21f JS |
2413 | errno = 0; |
2414 | long long value = (long long) strtoull (startp, & endp, 0); | |
16e8f21f | 2415 | if (errno == ERANGE || errno == EINVAL || *endp != '\0' |
79e6d33f | 2416 | || (neg && (unsigned long long) value > 9223372036854775808ULL) |
16e8f21f JS |
2417 | || (unsigned long long) value > 18446744073709551615ULL |
2418 | || value < -9223372036854775807LL-1) | |
2677d2fb | 2419 | throw parse_error (_("number invalid or out of range")); |
16e8f21f | 2420 | |
79e6d33f JS |
2421 | if (neg) |
2422 | value = -value; | |
2423 | ||
16e8f21f JS |
2424 | l = new literal_number (value); |
2425 | } | |
2426 | else | |
2677d2fb | 2427 | throw parse_error (_("expected literal string or number")); |
9c0c0e46 | 2428 | } |
56099f08 FCE |
2429 | |
2430 | l->tok = t; | |
2431 | return l; | |
2f1a1aea FCE |
2432 | } |
2433 | ||
2434 | ||
2435 | if_statement* | |
2436 | parser::parse_if_statement () | |
2437 | { | |
2438 | const token* t = next (); | |
6e213f58 | 2439 | if (! (t->type == tok_keyword && t->content == "if")) |
2677d2fb | 2440 | throw parse_error (_("expected 'if'")); |
56099f08 FCE |
2441 | if_statement* s = new if_statement; |
2442 | s->tok = t; | |
2443 | ||
2444 | t = next (); | |
2f1a1aea | 2445 | if (! (t->type == tok_operator && t->content == "(")) |
2677d2fb | 2446 | throw parse_error (_("expected '('")); |
731a5359 | 2447 | swallow (); |
2f1a1aea | 2448 | |
2f1a1aea FCE |
2449 | s->condition = parse_expression (); |
2450 | ||
2451 | t = next (); | |
2452 | if (! (t->type == tok_operator && t->content == ")")) | |
2677d2fb | 2453 | throw parse_error (_("expected ')'")); |
731a5359 | 2454 | swallow (); |
2f1a1aea FCE |
2455 | |
2456 | s->thenblock = parse_statement (); | |
2457 | ||
2458 | t = peek (); | |
6e213f58 | 2459 | if (t && t->type == tok_keyword && t->content == "else") |
2f1a1aea | 2460 | { |
731a5359 | 2461 | swallow (); |
2f1a1aea FCE |
2462 | s->elseblock = parse_statement (); |
2463 | } | |
ed10c639 FCE |
2464 | else |
2465 | s->elseblock = 0; // in case not otherwise initialized | |
2f1a1aea FCE |
2466 | |
2467 | return s; | |
2468 | } | |
2469 | ||
2470 | ||
69c68955 FCE |
2471 | expr_statement* |
2472 | parser::parse_expr_statement () | |
2473 | { | |
2474 | expr_statement *es = new expr_statement; | |
2475 | const token* t = peek (); | |
5e58d11c MW |
2476 | if (t == NULL) |
2477 | throw parse_error (_("expression statement expected")); | |
731a5359 MW |
2478 | // Copy, we only peeked, parse_expression might swallow. |
2479 | es->tok = new token (*t); | |
69c68955 FCE |
2480 | es->value = parse_expression (); |
2481 | return es; | |
2482 | } | |
2483 | ||
2484 | ||
56099f08 FCE |
2485 | return_statement* |
2486 | parser::parse_return_statement () | |
2487 | { | |
2488 | const token* t = next (); | |
6e213f58 | 2489 | if (! (t->type == tok_keyword && t->content == "return")) |
2677d2fb | 2490 | throw parse_error (_("expected 'return'")); |
6e213f58 | 2491 | if (context != con_function) |
2677d2fb | 2492 | throw parse_error (_("found 'return' not in function context")); |
56099f08 FCE |
2493 | return_statement* s = new return_statement; |
2494 | s->tok = t; | |
2495 | s->value = parse_expression (); | |
2496 | return s; | |
2497 | } | |
2498 | ||
2499 | ||
2500 | delete_statement* | |
2501 | parser::parse_delete_statement () | |
2502 | { | |
2503 | const token* t = next (); | |
6e213f58 | 2504 | if (! (t->type == tok_keyword && t->content == "delete")) |
2677d2fb | 2505 | throw parse_error (_("expected 'delete'")); |
56099f08 FCE |
2506 | delete_statement* s = new delete_statement; |
2507 | s->tok = t; | |
2508 | s->value = parse_expression (); | |
2509 | return s; | |
2510 | } | |
2511 | ||
2512 | ||
f3c26ea5 FCE |
2513 | next_statement* |
2514 | parser::parse_next_statement () | |
2515 | { | |
2516 | const token* t = next (); | |
6e213f58 | 2517 | if (! (t->type == tok_keyword && t->content == "next")) |
2677d2fb | 2518 | throw parse_error (_("expected 'next'")); |
6e213f58 | 2519 | if (context != con_probe) |
2677d2fb | 2520 | throw parse_error (_("found 'next' not in probe context")); |
f3c26ea5 FCE |
2521 | next_statement* s = new next_statement; |
2522 | s->tok = t; | |
2523 | return s; | |
2524 | } | |
2525 | ||
2526 | ||
2527 | break_statement* | |
2528 | parser::parse_break_statement () | |
2529 | { | |
2530 | const token* t = next (); | |
6e213f58 | 2531 | if (! (t->type == tok_keyword && t->content == "break")) |
2677d2fb | 2532 | throw parse_error (_("expected 'break'")); |
f3c26ea5 FCE |
2533 | break_statement* s = new break_statement; |
2534 | s->tok = t; | |
2535 | return s; | |
2536 | } | |
2537 | ||
2538 | ||
2539 | continue_statement* | |
2540 | parser::parse_continue_statement () | |
2541 | { | |
2542 | const token* t = next (); | |
6e213f58 | 2543 | if (! (t->type == tok_keyword && t->content == "continue")) |
2677d2fb | 2544 | throw parse_error (_("expected 'continue'")); |
f3c26ea5 FCE |
2545 | continue_statement* s = new continue_statement; |
2546 | s->tok = t; | |
2547 | return s; | |
2548 | } | |
2549 | ||
2550 | ||
69c68955 FCE |
2551 | for_loop* |
2552 | parser::parse_for_loop () | |
2553 | { | |
f3c26ea5 | 2554 | const token* t = next (); |
6e213f58 | 2555 | if (! (t->type == tok_keyword && t->content == "for")) |
2677d2fb | 2556 | throw parse_error (_("expected 'for'")); |
f3c26ea5 FCE |
2557 | for_loop* s = new for_loop; |
2558 | s->tok = t; | |
2559 | ||
2560 | t = next (); | |
2561 | if (! (t->type == tok_operator && t->content == "(")) | |
2677d2fb | 2562 | throw parse_error (_("expected '('")); |
731a5359 | 2563 | swallow (); |
f3c26ea5 FCE |
2564 | |
2565 | // initializer + ";" | |
2566 | t = peek (); | |
2567 | if (t && t->type == tok_operator && t->content == ";") | |
2568 | { | |
cbfbbf69 | 2569 | s->init = 0; |
731a5359 | 2570 | swallow (); |
f3c26ea5 FCE |
2571 | } |
2572 | else | |
2573 | { | |
2574 | s->init = parse_expr_statement (); | |
2575 | t = next (); | |
2576 | if (! (t->type == tok_operator && t->content == ";")) | |
2677d2fb | 2577 | throw parse_error (_("expected ';'")); |
731a5359 | 2578 | swallow (); |
f3c26ea5 FCE |
2579 | } |
2580 | ||
2581 | // condition + ";" | |
2582 | t = peek (); | |
2583 | if (t && t->type == tok_operator && t->content == ";") | |
2584 | { | |
2585 | literal_number* l = new literal_number(1); | |
2586 | s->cond = l; | |
2587 | s->cond->tok = next (); | |
2588 | } | |
2589 | else | |
2590 | { | |
2591 | s->cond = parse_expression (); | |
2592 | t = next (); | |
2593 | if (! (t->type == tok_operator && t->content == ";")) | |
2677d2fb | 2594 | throw parse_error (_("expected ';'")); |
731a5359 | 2595 | swallow (); |
f3c26ea5 | 2596 | } |
dff50e09 | 2597 | |
f3c26ea5 FCE |
2598 | // increment + ")" |
2599 | t = peek (); | |
2600 | if (t && t->type == tok_operator && t->content == ")") | |
2601 | { | |
cbfbbf69 | 2602 | s->incr = 0; |
731a5359 | 2603 | swallow (); |
f3c26ea5 FCE |
2604 | } |
2605 | else | |
2606 | { | |
2607 | s->incr = parse_expr_statement (); | |
2608 | t = next (); | |
2609 | if (! (t->type == tok_operator && t->content == ")")) | |
2677d2fb | 2610 | throw parse_error (_("expected ')'")); |
731a5359 | 2611 | swallow (); |
f3c26ea5 FCE |
2612 | } |
2613 | ||
2614 | // block | |
2615 | s->block = parse_statement (); | |
2616 | ||
2617 | return s; | |
2618 | } | |
2619 | ||
2620 | ||
2621 | for_loop* | |
2622 | parser::parse_while_loop () | |
2623 | { | |
2624 | const token* t = next (); | |
6e213f58 | 2625 | if (! (t->type == tok_keyword && t->content == "while")) |
2677d2fb | 2626 | throw parse_error (_("expected 'while'")); |
f3c26ea5 FCE |
2627 | for_loop* s = new for_loop; |
2628 | s->tok = t; | |
2629 | ||
2630 | t = next (); | |
2631 | if (! (t->type == tok_operator && t->content == "(")) | |
2677d2fb | 2632 | throw parse_error (_("expected '('")); |
731a5359 | 2633 | swallow (); |
f3c26ea5 FCE |
2634 | |
2635 | // dummy init and incr fields | |
cbfbbf69 FCE |
2636 | s->init = 0; |
2637 | s->incr = 0; | |
f3c26ea5 FCE |
2638 | |
2639 | // condition | |
2640 | s->cond = parse_expression (); | |
2641 | ||
f3c26ea5 FCE |
2642 | t = next (); |
2643 | if (! (t->type == tok_operator && t->content == ")")) | |
2677d2fb | 2644 | throw parse_error (_("expected ')'")); |
731a5359 | 2645 | swallow (); |
dff50e09 | 2646 | |
f3c26ea5 FCE |
2647 | // block |
2648 | s->block = parse_statement (); | |
2649 | ||
2650 | return s; | |
69c68955 FCE |
2651 | } |
2652 | ||
2653 | ||
2654 | foreach_loop* | |
2655 | parser::parse_foreach_loop () | |
2656 | { | |
2657 | const token* t = next (); | |
6e213f58 | 2658 | if (! (t->type == tok_keyword && t->content == "foreach")) |
2677d2fb | 2659 | throw parse_error (_("expected 'foreach'")); |
69c68955 FCE |
2660 | foreach_loop* s = new foreach_loop; |
2661 | s->tok = t; | |
93484556 | 2662 | s->sort_direction = 0; |
fd5689dc | 2663 | s->sort_aggr = sc_none; |
c261711d | 2664 | s->value = NULL; |
27f21e8c | 2665 | s->limit = NULL; |
69c68955 FCE |
2666 | |
2667 | t = next (); | |
2668 | if (! (t->type == tok_operator && t->content == "(")) | |
2677d2fb | 2669 | throw parse_error (_("expected '('")); |
731a5359 | 2670 | swallow (); |
69c68955 | 2671 | |
c261711d JS |
2672 | symbol* lookahead_sym = NULL; |
2673 | int lookahead_sort = 0; | |
2674 | ||
2675 | t = peek (); | |
2676 | if (t && t->type == tok_identifier) | |
2677 | { | |
2678 | next (); | |
2679 | lookahead_sym = new symbol; | |
2680 | lookahead_sym->tok = t; | |
2681 | lookahead_sym->name = t->content; | |
2682 | ||
2683 | t = peek (); | |
2684 | if (t && t->type == tok_operator && | |
2685 | (t->content == "+" || t->content == "-")) | |
2686 | { | |
c261711d | 2687 | lookahead_sort = (t->content == "+") ? 1 : -1; |
731a5359 | 2688 | swallow (); |
c261711d JS |
2689 | } |
2690 | ||
2691 | t = peek (); | |
2692 | if (t && t->type == tok_operator && t->content == "=") | |
2693 | { | |
731a5359 | 2694 | swallow (); |
c261711d JS |
2695 | s->value = lookahead_sym; |
2696 | if (lookahead_sort) | |
2697 | { | |
2698 | s->sort_direction = lookahead_sort; | |
2699 | s->sort_column = 0; | |
2700 | } | |
2701 | lookahead_sym = NULL; | |
2702 | } | |
2703 | } | |
2704 | ||
69c68955 FCE |
2705 | // see also parse_array_in |
2706 | ||
2707 | bool parenthesized = false; | |
2708 | t = peek (); | |
c261711d | 2709 | if (!lookahead_sym && t && t->type == tok_operator && t->content == "[") |
69c68955 | 2710 | { |
731a5359 | 2711 | swallow (); |
69c68955 FCE |
2712 | parenthesized = true; |
2713 | } | |
2714 | ||
c261711d JS |
2715 | if (lookahead_sym) |
2716 | { | |
2717 | s->indexes.push_back (lookahead_sym); | |
2718 | if (lookahead_sort) | |
2719 | { | |
2720 | s->sort_direction = lookahead_sort; | |
2721 | s->sort_column = 1; | |
2722 | } | |
2723 | lookahead_sym = NULL; | |
2724 | } | |
2725 | else while (1) | |
69c68955 FCE |
2726 | { |
2727 | t = next (); | |
2728 | if (! (t->type == tok_identifier)) | |
2677d2fb | 2729 | throw parse_error (_("expected identifier")); |
69c68955 FCE |
2730 | symbol* sym = new symbol; |
2731 | sym->tok = t; | |
2732 | sym->name = t->content; | |
2733 | s->indexes.push_back (sym); | |
2734 | ||
93484556 FCE |
2735 | t = peek (); |
2736 | if (t && t->type == tok_operator && | |
2737 | (t->content == "+" || t->content == "-")) | |
2738 | { | |
2739 | if (s->sort_direction) | |
2677d2fb | 2740 | throw parse_error (_("multiple sort directives")); |
93484556 FCE |
2741 | s->sort_direction = (t->content == "+") ? 1 : -1; |
2742 | s->sort_column = s->indexes.size(); | |
731a5359 | 2743 | swallow (); |
93484556 FCE |
2744 | } |
2745 | ||
69c68955 FCE |
2746 | if (parenthesized) |
2747 | { | |
93484556 | 2748 | t = peek (); |
69c68955 FCE |
2749 | if (t && t->type == tok_operator && t->content == ",") |
2750 | { | |
731a5359 | 2751 | swallow (); |
69c68955 FCE |
2752 | continue; |
2753 | } | |
2754 | else if (t && t->type == tok_operator && t->content == "]") | |
2755 | { | |
731a5359 | 2756 | swallow (); |
69c68955 FCE |
2757 | break; |
2758 | } | |
dff50e09 | 2759 | else |
2677d2fb | 2760 | throw parse_error (_("expected ',' or ']'")); |
69c68955 FCE |
2761 | } |
2762 | else | |
2763 | break; // expecting only one expression | |
2764 | } | |
2765 | ||
2766 | t = next (); | |
6e213f58 | 2767 | if (! (t->type == tok_keyword && t->content == "in")) |
2677d2fb | 2768 | throw parse_error (_("expected 'in'")); |
731a5359 | 2769 | swallow (); |
dff50e09 | 2770 | |
d02548c0 | 2771 | s->base = parse_indexable(); |
69c68955 | 2772 | |
fd5689dc FCE |
2773 | // check for atword, see also expect_ident_or_atword, |
2774 | t = peek (); | |
2775 | if (t && t->type == tok_operator && t->content[0] == '@') | |
2776 | { | |
2777 | if (t->content == "@avg") s->sort_aggr = sc_average; | |
2778 | else if (t->content == "@min") s->sort_aggr = sc_min; | |
2779 | else if (t->content == "@max") s->sort_aggr = sc_max; | |
2780 | else if (t->content == "@count") s->sort_aggr = sc_count; | |
2781 | else if (t->content == "@sum") s->sort_aggr = sc_sum; | |
2782 | else throw parse_error(_("expected statistical operation")); | |
2783 | swallow(); | |
2784 | ||
2785 | t = peek (); | |
2786 | if (! (t && t->type == tok_operator && (t->content == "+" || t->content == "-"))) | |
2787 | throw parse_error(_("expected sort directive")); | |
2788 | } | |
2789 | ||
93484556 FCE |
2790 | t = peek (); |
2791 | if (t && t->type == tok_operator && | |
2792 | (t->content == "+" || t->content == "-")) | |
2793 | { | |
2794 | if (s->sort_direction) | |
2677d2fb | 2795 | throw parse_error (_("multiple sort directives")); |
93484556 FCE |
2796 | s->sort_direction = (t->content == "+") ? 1 : -1; |
2797 | s->sort_column = 0; | |
731a5359 | 2798 | swallow (); |
93484556 FCE |
2799 | } |
2800 | ||
27f21e8c DS |
2801 | t = peek (); |
2802 | if (tok_is(t, tok_keyword, "limit")) | |
2803 | { | |
731a5359 | 2804 | swallow (); // get past the "limit" |
27f21e8c DS |
2805 | s->limit = parse_expression (); |
2806 | } | |
2807 | ||
69c68955 FCE |
2808 | t = next (); |
2809 | if (! (t->type == tok_operator && t->content == ")")) | |
2810 | throw parse_error ("expected ')'"); | |
731a5359 | 2811 | swallow (); |
69c68955 FCE |
2812 | |
2813 | s->block = parse_statement (); | |
2814 | return s; | |
2815 | } | |
2816 | ||
2817 | ||
2f1a1aea FCE |
2818 | expression* |
2819 | parser::parse_expression () | |
2820 | { | |
2821 | return parse_assignment (); | |
2822 | } | |
2823 | ||
2f1a1aea FCE |
2824 | |
2825 | expression* | |
2826 | parser::parse_assignment () | |
2827 | { | |
2828 | expression* op1 = parse_ternary (); | |
2829 | ||
2830 | const token* t = peek (); | |
82919855 | 2831 | // right-associative operators |
dff50e09 | 2832 | if (t && t->type == tok_operator |
2f1a1aea | 2833 | && (t->content == "=" || |
82919855 | 2834 | t->content == "<<<" || |
2f1a1aea | 2835 | t->content == "+=" || |
bb2e3076 FCE |
2836 | t->content == "-=" || |
2837 | t->content == "*=" || | |
2838 | t->content == "/=" || | |
2839 | t->content == "%=" || | |
2840 | t->content == "<<=" || | |
2841 | t->content == ">>=" || | |
2842 | t->content == "&=" || | |
2843 | t->content == "^=" || | |
2844 | t->content == "|=" || | |
d5d7c2cc | 2845 | t->content == ".=" || |
dff50e09 | 2846 | false)) |
2f1a1aea | 2847 | { |
bb2e3076 | 2848 | // NB: lvalueness is checked during elaboration / translation |
2f1a1aea | 2849 | assignment* e = new assignment; |
56099f08 | 2850 | e->left = op1; |
2f1a1aea | 2851 | e->op = t->content; |
56099f08 | 2852 | e->tok = t; |
2f1a1aea | 2853 | next (); |
82919855 | 2854 | e->right = parse_expression (); |
56099f08 | 2855 | op1 = e; |
2f1a1aea | 2856 | } |
56099f08 FCE |
2857 | |
2858 | return op1; | |
2f1a1aea FCE |
2859 | } |
2860 | ||
2861 | ||
2862 | expression* | |
2863 | parser::parse_ternary () | |
2864 | { | |
2865 | expression* op1 = parse_logical_or (); | |
2866 | ||
2867 | const token* t = peek (); | |
2868 | if (t && t->type == tok_operator && t->content == "?") | |
2869 | { | |
2f1a1aea | 2870 | ternary_expression* e = new ternary_expression; |
56099f08 | 2871 | e->tok = t; |
2f1a1aea | 2872 | e->cond = op1; |
56099f08 FCE |
2873 | next (); |
2874 | e->truevalue = parse_expression (); // XXX | |
2f1a1aea FCE |
2875 | |
2876 | t = next (); | |
2877 | if (! (t->type == tok_operator && t->content == ":")) | |
2677d2fb | 2878 | throw parse_error (_("expected ':'")); |
731a5359 | 2879 | swallow (); |
2f1a1aea | 2880 | |
56099f08 | 2881 | e->falsevalue = parse_expression (); // XXX |
2f1a1aea FCE |
2882 | return e; |
2883 | } | |
2884 | else | |
2885 | return op1; | |
2886 | } | |
2887 | ||
2888 | ||
2889 | expression* | |
2890 | parser::parse_logical_or () | |
2891 | { | |
2892 | expression* op1 = parse_logical_and (); | |
dff50e09 | 2893 | |
2f1a1aea | 2894 | const token* t = peek (); |
56099f08 | 2895 | while (t && t->type == tok_operator && t->content == "||") |
2f1a1aea | 2896 | { |
2f1a1aea | 2897 | logical_or_expr* e = new logical_or_expr; |
56099f08 FCE |
2898 | e->tok = t; |
2899 | e->op = t->content; | |
2f1a1aea | 2900 | e->left = op1; |
56099f08 FCE |
2901 | next (); |
2902 | e->right = parse_logical_and (); | |
2903 | op1 = e; | |
2904 | t = peek (); | |
2f1a1aea | 2905 | } |
56099f08 FCE |
2906 | |
2907 | return op1; | |
2f1a1aea FCE |
2908 | } |
2909 | ||
2910 | ||
2911 | expression* | |
2912 | parser::parse_logical_and () | |
2913 | { | |
bb2e3076 | 2914 | expression* op1 = parse_boolean_or (); |
2f1a1aea FCE |
2915 | |
2916 | const token* t = peek (); | |
56099f08 | 2917 | while (t && t->type == tok_operator && t->content == "&&") |
2f1a1aea | 2918 | { |
2f1a1aea FCE |
2919 | logical_and_expr *e = new logical_and_expr; |
2920 | e->left = op1; | |
56099f08 FCE |
2921 | e->op = t->content; |
2922 | e->tok = t; | |
2923 | next (); | |
bb2e3076 FCE |
2924 | e->right = parse_boolean_or (); |
2925 | op1 = e; | |
2926 | t = peek (); | |
2927 | } | |
2928 | ||
2929 | return op1; | |
2930 | } | |
2931 | ||
2932 | ||
2933 | expression* | |
2934 | parser::parse_boolean_or () | |
2935 | { | |
2936 | expression* op1 = parse_boolean_xor (); | |
2937 | ||
2938 | const token* t = peek (); | |
2939 | while (t && t->type == tok_operator && t->content == "|") | |
2940 | { | |
2941 | binary_expression* e = new binary_expression; | |
2942 | e->left = op1; | |
2943 | e->op = t->content; | |
2944 | e->tok = t; | |
2945 | next (); | |
2946 | e->right = parse_boolean_xor (); | |
2947 | op1 = e; | |
2948 | t = peek (); | |
2949 | } | |
2950 | ||
2951 | return op1; | |
2952 | } | |
2953 | ||
2954 | ||
2955 | expression* | |
2956 | parser::parse_boolean_xor () | |
2957 | { | |
2958 | expression* op1 = parse_boolean_and (); | |
2959 | ||
2960 | const token* t = peek (); | |
2961 | while (t && t->type == tok_operator && t->content == "^") | |
2962 | { | |
2963 | binary_expression* e = new binary_expression; | |
2964 | e->left = op1; | |
2965 | e->op = t->content; | |
2966 | e->tok = t; | |
2967 | next (); | |
2968 | e->right = parse_boolean_and (); | |
2969 | op1 = e; | |
2970 | t = peek (); | |
2971 | } | |
2972 | ||
2973 | return op1; | |
2974 | } | |
2975 | ||
2976 | ||
2977 | expression* | |
2978 | parser::parse_boolean_and () | |
2979 | { | |
2980 | expression* op1 = parse_array_in (); | |
2981 | ||
2982 | const token* t = peek (); | |
2983 | while (t && t->type == tok_operator && t->content == "&") | |
2984 | { | |
2985 | binary_expression* e = new binary_expression; | |
2986 | e->left = op1; | |
2987 | e->op = t->content; | |
2988 | e->tok = t; | |
2989 | next (); | |
56099f08 FCE |
2990 | e->right = parse_array_in (); |
2991 | op1 = e; | |
2992 | t = peek (); | |
2f1a1aea | 2993 | } |
56099f08 FCE |
2994 | |
2995 | return op1; | |
2f1a1aea FCE |
2996 | } |
2997 | ||
2998 | ||
2999 | expression* | |
3000 | parser::parse_array_in () | |
3001 | { | |
ce10591c | 3002 | // This is a very tricky case. All these are legit expressions: |
69c68955 | 3003 | // "a in b" "a+0 in b" "[a,b] in c" "[c,(d+0)] in b" |
ce10591c FCE |
3004 | vector<expression*> indexes; |
3005 | bool parenthesized = false; | |
2f1a1aea FCE |
3006 | |
3007 | const token* t = peek (); | |
69c68955 | 3008 | if (t && t->type == tok_operator && t->content == "[") |
ce10591c | 3009 | { |
731a5359 | 3010 | swallow (); |
ce10591c FCE |
3011 | parenthesized = true; |
3012 | } | |
3013 | ||
3014 | while (1) | |
3015 | { | |
93daaca8 | 3016 | expression* op1 = parse_comparison_or_regex_query (); |
ce10591c FCE |
3017 | indexes.push_back (op1); |
3018 | ||
3019 | if (parenthesized) | |
3020 | { | |
3021 | const token* t = peek (); | |
3022 | if (t && t->type == tok_operator && t->content == ",") | |
3023 | { | |
731a5359 | 3024 | swallow (); |
ce10591c FCE |
3025 | continue; |
3026 | } | |
69c68955 | 3027 | else if (t && t->type == tok_operator && t->content == "]") |
ce10591c | 3028 | { |
731a5359 | 3029 | swallow (); |
ce10591c FCE |
3030 | break; |
3031 | } | |
dff50e09 | 3032 | else |
2677d2fb | 3033 | throw parse_error (_("expected ',' or ']'")); |
ce10591c FCE |
3034 | } |
3035 | else | |
3036 | break; // expecting only one expression | |
3037 | } | |
3038 | ||
3039 | t = peek (); | |
6e213f58 | 3040 | if (t && t->type == tok_keyword && t->content == "in") |
2f1a1aea | 3041 | { |
2f1a1aea | 3042 | array_in *e = new array_in; |
56099f08 | 3043 | e->tok = t; |
731a5359 | 3044 | next (); |
ce10591c FCE |
3045 | |
3046 | arrayindex* a = new arrayindex; | |
3047 | a->indexes = indexes; | |
d02548c0 | 3048 | a->base = parse_indexable(); |
d15d767c | 3049 | a->tok = a->base->tok; |
ce10591c | 3050 | e->operand = a; |
2f1a1aea FCE |
3051 | return e; |
3052 | } | |
ce10591c FCE |
3053 | else if (indexes.size() == 1) // no "in" - need one expression only |
3054 | return indexes[0]; | |
2f1a1aea | 3055 | else |
2677d2fb | 3056 | throw parse_error (_("unexpected comma-separated expression list")); |
2f1a1aea FCE |
3057 | } |
3058 | ||
3059 | ||
3060 | expression* | |
93daaca8 | 3061 | parser::parse_comparison_or_regex_query () |
2f1a1aea | 3062 | { |
bb2e3076 | 3063 | expression* op1 = parse_shift (); |
2f1a1aea | 3064 | |
557abe61 | 3065 | // XXX precedence -- perhaps a =~ b == c =~ d --> (a =~ b) == (c =~ d) ? |
93daaca8 SM |
3066 | const token *t = peek(); |
3067 | if (t && t->type == tok_operator | |
3068 | && (t->content == "=~" || | |
3069 | t->content == "!~")) | |
3070 | { | |
3071 | regex_query* r = new regex_query; | |
3072 | r->left = op1; | |
3073 | r->op = t->content; | |
3074 | r->tok = t; | |
3075 | next (); | |
d3bc48f0 | 3076 | r->right = parse_literal_string(); |
93daaca8 SM |
3077 | op1 = r; |
3078 | t = peek (); | |
3079 | } | |
3080 | else while (t && t->type == tok_operator | |
553d27a5 FCE |
3081 | && (t->content == ">" || |
3082 | t->content == "<" || | |
3083 | t->content == "==" || | |
3084 | t->content == "!=" || | |
3085 | t->content == "<=" || | |
bb2e3076 | 3086 | t->content == ">=")) |
2f1a1aea FCE |
3087 | { |
3088 | comparison* e = new comparison; | |
3089 | e->left = op1; | |
3090 | e->op = t->content; | |
56099f08 | 3091 | e->tok = t; |
2f1a1aea | 3092 | next (); |
bb2e3076 FCE |
3093 | e->right = parse_shift (); |
3094 | op1 = e; | |
3095 | t = peek (); | |
3096 | } | |
3097 | ||
3098 | return op1; | |
3099 | } | |
3100 | ||
3101 | ||
3102 | expression* | |
3103 | parser::parse_shift () | |
3104 | { | |
3105 | expression* op1 = parse_concatenation (); | |
3106 | ||
3107 | const token* t = peek (); | |
dff50e09 | 3108 | while (t && t->type == tok_operator && |
bb2e3076 FCE |
3109 | (t->content == "<<" || t->content == ">>")) |
3110 | { | |
3111 | binary_expression* e = new binary_expression; | |
3112 | e->left = op1; | |
3113 | e->op = t->content; | |
3114 | e->tok = t; | |
3115 | next (); | |
56099f08 FCE |
3116 | e->right = parse_concatenation (); |
3117 | op1 = e; | |
3118 | t = peek (); | |
2f1a1aea | 3119 | } |
56099f08 FCE |
3120 | |
3121 | return op1; | |
2f1a1aea FCE |
3122 | } |
3123 | ||
3124 | ||
3125 | expression* | |
3126 | parser::parse_concatenation () | |
3127 | { | |
3128 | expression* op1 = parse_additive (); | |
3129 | ||
3130 | const token* t = peek (); | |
3131 | // XXX: the actual awk string-concatenation operator is *whitespace*. | |
3132 | // I don't know how to easily to model that here. | |
56099f08 | 3133 | while (t && t->type == tok_operator && t->content == ".") |
2f1a1aea FCE |
3134 | { |
3135 | concatenation* e = new concatenation; | |
3136 | e->left = op1; | |
3137 | e->op = t->content; | |
56099f08 | 3138 | e->tok = t; |
2f1a1aea | 3139 | next (); |
56099f08 FCE |
3140 | e->right = parse_additive (); |
3141 | op1 = e; | |
3142 | t = peek (); | |
2f1a1aea | 3143 | } |
56099f08 FCE |
3144 | |
3145 | return op1; | |
2f1a1aea FCE |
3146 | } |
3147 | ||
3148 | ||
3149 | expression* | |
3150 | parser::parse_additive () | |
3151 | { | |
3152 | expression* op1 = parse_multiplicative (); | |
3153 | ||
3154 | const token* t = peek (); | |
dff50e09 | 3155 | while (t && t->type == tok_operator |
2f1a1aea FCE |
3156 | && (t->content == "+" || t->content == "-")) |
3157 | { | |
3158 | binary_expression* e = new binary_expression; | |
3159 | e->op = t->content; | |
3160 | e->left = op1; | |
56099f08 | 3161 | e->tok = t; |
2f1a1aea | 3162 | next (); |
56099f08 FCE |
3163 | e->right = parse_multiplicative (); |
3164 | op1 = e; | |
3165 | t = peek (); | |
2f1a1aea | 3166 | } |
56099f08 FCE |
3167 | |
3168 | return op1; | |
2f1a1aea FCE |
3169 | } |
3170 | ||
3171 | ||
3172 | expression* | |
3173 | parser::parse_multiplicative () | |
3174 | { | |
3175 | expression* op1 = parse_unary (); | |
3176 | ||
3177 | const token* t = peek (); | |
dff50e09 | 3178 | while (t && t->type == tok_operator |
2f1a1aea FCE |
3179 | && (t->content == "*" || t->content == "/" || t->content == "%")) |
3180 | { | |
3181 | binary_expression* e = new binary_expression; | |
3182 | e->op = t->content; | |
3183 | e->left = op1; | |
56099f08 | 3184 | e->tok = t; |
2f1a1aea | 3185 | next (); |
56099f08 FCE |
3186 | e->right = parse_unary (); |
3187 | op1 = e; | |
3188 | t = peek (); | |
2f1a1aea | 3189 | } |
56099f08 FCE |
3190 | |
3191 | return op1; | |
2f1a1aea FCE |
3192 | } |
3193 | ||
3194 | ||
3195 | expression* | |
3196 | parser::parse_unary () | |
3197 | { | |
3198 | const token* t = peek (); | |
dff50e09 FCE |
3199 | if (t && t->type == tok_operator |
3200 | && (t->content == "+" || | |
3201 | t->content == "-" || | |
bb2e3076 FCE |
3202 | t->content == "!" || |
3203 | t->content == "~" || | |
3204 | false)) | |
2f1a1aea FCE |
3205 | { |
3206 | unary_expression* e = new unary_expression; | |
3207 | e->op = t->content; | |
56099f08 | 3208 | e->tok = t; |
2f1a1aea | 3209 | next (); |
1cb79a72 | 3210 | e->operand = parse_unary (); |
2f1a1aea FCE |
3211 | return e; |
3212 | } | |
3213 | else | |
bb2e3076 | 3214 | return parse_crement (); |
2f1a1aea FCE |
3215 | } |
3216 | ||
3217 | ||
3218 | expression* | |
3219 | parser::parse_crement () // as in "increment" / "decrement" | |
3220 | { | |
cbfbbf69 FCE |
3221 | // NB: Ideally, we'd parse only a symbol as an operand to the |
3222 | // *crement operators, instead of a general expression value. We'd | |
3223 | // need more complex lookahead code to tell apart the postfix cases. | |
3224 | // So we just punt, and leave it to pass-3 to signal errors on | |
3225 | // cases like "4++". | |
3226 | ||
2f1a1aea | 3227 | const token* t = peek (); |
dff50e09 | 3228 | if (t && t->type == tok_operator |
2f1a1aea FCE |
3229 | && (t->content == "++" || t->content == "--")) |
3230 | { | |
3231 | pre_crement* e = new pre_crement; | |
3232 | e->op = t->content; | |
56099f08 | 3233 | e->tok = t; |
2f1a1aea FCE |
3234 | next (); |
3235 | e->operand = parse_value (); | |
3236 | return e; | |
3237 | } | |
3238 | ||
3239 | // post-crement or non-crement | |
3240 | expression *op1 = parse_value (); | |
dff50e09 | 3241 | |
2f1a1aea | 3242 | t = peek (); |
dff50e09 | 3243 | if (t && t->type == tok_operator |
2f1a1aea FCE |
3244 | && (t->content == "++" || t->content == "--")) |
3245 | { | |
3246 | post_crement* e = new post_crement; | |
3247 | e->op = t->content; | |
56099f08 | 3248 | e->tok = t; |
2f1a1aea FCE |
3249 | next (); |
3250 | e->operand = op1; | |
3251 | return e; | |
3252 | } | |
3253 | else | |
3254 | return op1; | |
3255 | } | |
3256 | ||
3257 | ||
3258 | expression* | |
3259 | parser::parse_value () | |
3260 | { | |
3261 | const token* t = peek (); | |
3262 | if (! t) | |
2677d2fb | 3263 | throw parse_error (_("expected value")); |
2f1a1aea | 3264 | |
7d902887 FCE |
3265 | if (t->type == tok_embedded) |
3266 | { | |
7d902887 | 3267 | if (! privileged) |
efb02738 | 3268 | throw parse_error (_("embedded expression code in unprivileged script; need stap -g"), false); |
7d902887 FCE |
3269 | |
3270 | embedded_expr *e = new embedded_expr; | |
3271 | e->tok = t; | |
3272 | e->code = t->content; | |
731a5359 | 3273 | next (); |
7d902887 FCE |
3274 | return e; |
3275 | } | |
3276 | ||
2f1a1aea FCE |
3277 | if (t->type == tok_operator && t->content == "(") |
3278 | { | |
731a5359 | 3279 | swallow (); |
2f1a1aea FCE |
3280 | expression* e = parse_expression (); |
3281 | t = next (); | |
3282 | if (! (t->type == tok_operator && t->content == ")")) | |
2677d2fb | 3283 | throw parse_error (_("expected ')'")); |
731a5359 | 3284 | swallow (); |
2f1a1aea FCE |
3285 | return e; |
3286 | } | |
03c75a4a JS |
3287 | else if (t->type == tok_operator && t->content == "&") |
3288 | { | |
731a5359 | 3289 | next (); // Cannot swallow, passing token on... |
d48afc20 | 3290 | return parse_target_symbol (t); |
03c75a4a | 3291 | } |
06219d6f SM |
3292 | else if (t->type == tok_identifier |
3293 | || (t->type == tok_operator && t->content[0] == '@')) | |
2f1a1aea FCE |
3294 | return parse_symbol (); |
3295 | else | |
3296 | return parse_literal (); | |
3297 | } | |
3298 | ||
3299 | ||
d02548c0 GH |
3300 | const token * |
3301 | parser::parse_hist_op_or_bare_name (hist_op *&hop, string &name) | |
3302 | { | |
3303 | hop = NULL; | |
50cc7cd5 | 3304 | const token* t = expect_ident_or_atword (name); |
d02548c0 GH |
3305 | if (name == "@hist_linear" || name == "@hist_log") |
3306 | { | |
3307 | hop = new hist_op; | |
3308 | if (name == "@hist_linear") | |
3309 | hop->htype = hist_linear; | |
3310 | else if (name == "@hist_log") | |
3311 | hop->htype = hist_log; | |
3312 | hop->tok = t; | |
3313 | expect_op("("); | |
3314 | hop->stat = parse_expression (); | |
3315 | int64_t tnum; | |
3316 | if (hop->htype == hist_linear) | |
3317 | { | |
3318 | for (size_t i = 0; i < 3; ++i) | |
3319 | { | |
3320 | expect_op (","); | |
3321 | expect_number (tnum); | |
3322 | hop->params.push_back (tnum); | |
3323 | } | |
3324 | } | |
d02548c0 GH |
3325 | expect_op(")"); |
3326 | } | |
3327 | return t; | |
3328 | } | |
3329 | ||
3330 | ||
3331 | indexable* | |
3332 | parser::parse_indexable () | |
3333 | { | |
3334 | hist_op *hop = NULL; | |
3335 | string name; | |
3336 | const token *tok = parse_hist_op_or_bare_name(hop, name); | |
3337 | if (hop) | |
3338 | return hop; | |
3339 | else | |
3340 | { | |
3341 | symbol* sym = new symbol; | |
3342 | sym->name = name; | |
3343 | sym->tok = tok; | |
3344 | return sym; | |
3345 | } | |
3346 | } | |
3347 | ||
3348 | ||
cc9001af MW |
3349 | // var, indexable[index], func(parms), printf("...", ...), $var,r |
3350 | // @cast, @defined, @entry, @var, $var->member, @stat_op(stat) | |
30263a73 | 3351 | expression* parser::parse_symbol () |
2f1a1aea | 3352 | { |
d02548c0 GH |
3353 | hist_op *hop = NULL; |
3354 | symbol *sym = NULL; | |
d7f3e0c5 | 3355 | string name; |
d02548c0 GH |
3356 | const token *t = parse_hist_op_or_bare_name(hop, name); |
3357 | ||
3358 | if (!hop) | |
0fefb486 | 3359 | { |
dff50e09 | 3360 | // If we didn't get a hist_op, then we did get an identifier. We can |
d02548c0 GH |
3361 | // now scrutinize this identifier for the various magic forms of identifier |
3362 | // (printf, @stat_op, and $var...) | |
3363 | ||
cc9001af MW |
3364 | if (name == "@cast" |
3365 | || name == "@var" | |
3366 | || (name.size() > 0 && name[0] == '$')) | |
30263a73 | 3367 | return parse_target_symbol (t); |
9b5af295 | 3368 | |
db135493 FCE |
3369 | // NB: PR11343: @defined() is not incompatible with earlier versions |
3370 | // of stap, so no need to check session.compatible for 1.2 | |
30263a73 FCE |
3371 | if (name == "@defined") |
3372 | return parse_defined_op (t); | |
8cc799a5 JS |
3373 | |
3374 | if (name == "@entry") | |
3375 | return parse_entry_op (t); | |
3376 | ||
3689db05 SC |
3377 | if (name == "@perf") |
3378 | return parse_perf_op (t); | |
3379 | ||
cc9001af | 3380 | if (name.size() > 0 && name[0] == '@') |
d7f3e0c5 | 3381 | { |
d02548c0 GH |
3382 | stat_op *sop = new stat_op; |
3383 | if (name == "@avg") | |
3384 | sop->ctype = sc_average; | |
3385 | else if (name == "@count") | |
3386 | sop->ctype = sc_count; | |
3387 | else if (name == "@sum") | |
3388 | sop->ctype = sc_sum; | |
3389 | else if (name == "@min") | |
3390 | sop->ctype = sc_min; | |
3391 | else if (name == "@max") | |
3392 | sop->ctype = sc_max; | |
3393 | else | |
cc9001af | 3394 | throw parse_error(_("unknown operator ") + name); |
d02548c0 GH |
3395 | expect_op("("); |
3396 | sop->tok = t; | |
3397 | sop->stat = parse_expression (); | |
3398 | expect_op(")"); | |
3399 | return sop; | |
3400 | } | |
dff50e09 | 3401 | |
d5e178c1 | 3402 | else if (print_format *fmt = print_format::create(t)) |
d02548c0 | 3403 | { |
d02548c0 | 3404 | expect_op("("); |
b15c465c PP |
3405 | if ((name == "print" || name == "println" || |
3406 | name == "sprint" || name == "sprintln") && | |
f34254da | 3407 | (peek_op("@hist_linear") || peek_op("@hist_log"))) |
a4636912 GH |
3408 | { |
3409 | // We have a special case where we recognize | |
3410 | // print(@hist_foo(bar)) as a magic print-the-histogram | |
3411 | // construct. This is sort of gross but it avoids | |
3412 | // promoting histogram references to typeful | |
3413 | // expressions. | |
dff50e09 | 3414 | |
1bbeef03 GH |
3415 | hop = NULL; |
3416 | t = parse_hist_op_or_bare_name(hop, name); | |
3417 | assert(hop); | |
dff50e09 | 3418 | |
1bbeef03 GH |
3419 | // It is, sadly, possible that even while parsing a |
3420 | // hist_op, we *mis-guessed* and the user wishes to | |
3421 | // print(@hist_op(foo)[bucket]), a scalar. In that case | |
3422 | // we must parse the arrayindex and print an expression. | |
839325a1 JS |
3423 | // |
3424 | // XXX: This still fails if the arrayindex is part of a | |
3425 | // larger expression. To really handle everything, we'd | |
3426 | // need to push back all the hist tokens start over. | |
dff50e09 | 3427 | |
1bbeef03 GH |
3428 | if (!peek_op ("[")) |
3429 | fmt->hist = hop; | |
3430 | else | |
3431 | { | |
3432 | // This is simplified version of the | |
3433 | // multi-array-index parser below, because we can | |
3434 | // only ever have one index on a histogram anyways. | |
3435 | expect_op("["); | |
3436 | struct arrayindex* ai = new arrayindex; | |
3437 | ai->tok = t; | |
3438 | ai->base = hop; | |
3439 | ai->indexes.push_back (parse_expression ()); | |
3440 | expect_op("]"); | |
3441 | fmt->args.push_back(ai); | |
839325a1 JS |
3442 | |
3443 | // Consume any subsequent arguments. | |
3444 | while (!peek_op (")")) | |
3445 | { | |
3446 | expect_op(","); | |
3447 | expression *e = parse_expression (); | |
3448 | fmt->args.push_back(e); | |
3449 | } | |
1bbeef03 | 3450 | } |
a4636912 | 3451 | } |
d7f3e0c5 | 3452 | else |
d02548c0 | 3453 | { |
3cb17058 JS |
3454 | int min_args = 0; |
3455 | if (fmt->print_with_format) | |
3456 | { | |
3457 | // Consume and convert a format string. Agreement between the | |
3458 | // format string and the arguments is postponed to the | |
3459 | // typechecking phase. | |
3460 | string tmp; | |
3461 | expect_unknown (tok_string, tmp); | |
3462 | fmt->raw_components = tmp; | |
3463 | fmt->components = print_format::string_to_components (tmp); | |
3464 | } | |
3465 | else if (fmt->print_with_delim) | |
3466 | { | |
3467 | // Consume a delimiter to separate arguments. | |
3468 | fmt->delimiter.clear(); | |
3469 | fmt->delimiter.type = print_format::conv_literal; | |
3470 | expect_unknown (tok_string, fmt->delimiter.literal_string); | |
3471 | min_args = 2; | |
3472 | } | |
3473 | else | |
3474 | { | |
3475 | // If we are not printing with a format string, we must have | |
3476 | // at least one argument (of any type). | |
3477 | expression *e = parse_expression (); | |
3478 | fmt->args.push_back(e); | |
3479 | } | |
3480 | ||
3481 | // Consume any subsequent arguments. | |
3482 | while (min_args || !peek_op (")")) | |
3483 | { | |
3484 | expect_op(","); | |
3485 | expression *e = parse_expression (); | |
3486 | fmt->args.push_back(e); | |
3487 | if (min_args) | |
3488 | --min_args; | |
3489 | } | |
d02548c0 GH |
3490 | } |
3491 | expect_op(")"); | |
3492 | return fmt; | |
3493 | } | |
dff50e09 | 3494 | |
d02548c0 GH |
3495 | else if (peek_op ("(")) // function call |
3496 | { | |
731a5359 | 3497 | swallow (); |
d02548c0 GH |
3498 | struct functioncall* f = new functioncall; |
3499 | f->tok = t; | |
3500 | f->function = name; | |
3501 | // Allow empty actual parameter list | |
3502 | if (peek_op (")")) | |
3503 | { | |
731a5359 | 3504 | swallow (); |
d02548c0 GH |
3505 | return f; |
3506 | } | |
3507 | while (1) | |
3508 | { | |
3509 | f->args.push_back (parse_expression ()); | |
3510 | if (peek_op (")")) | |
3511 | { | |
731a5359 | 3512 | swallow (); |
d02548c0 GH |
3513 | break; |
3514 | } | |
3515 | else if (peek_op (",")) | |
3516 | { | |
731a5359 | 3517 | swallow (); |
d02548c0 GH |
3518 | continue; |
3519 | } | |
3520 | else | |
2677d2fb | 3521 | throw parse_error (_("expected ',' or ')'")); |
d02548c0 GH |
3522 | } |
3523 | return f; | |
3524 | } | |
3525 | ||
3526 | else | |
3527 | { | |
3528 | sym = new symbol; | |
3529 | sym->name = name; | |
3530 | sym->tok = t; | |
d7f3e0c5 | 3531 | } |
0fefb486 | 3532 | } |
dff50e09 FCE |
3533 | |
3534 | // By now, either we had a hist_op in the first place, or else | |
d02548c0 GH |
3535 | // we had a plain word and it was converted to a symbol. |
3536 | ||
70c743d8 | 3537 | assert (!hop != !sym); // logical XOR |
d02548c0 GH |
3538 | |
3539 | // All that remains is to check for array indexing | |
3540 | ||
d7f3e0c5 | 3541 | if (peek_op ("[")) // array |
2f1a1aea | 3542 | { |
731a5359 | 3543 | swallow (); |
2f1a1aea | 3544 | struct arrayindex* ai = new arrayindex; |
d02548c0 GH |
3545 | ai->tok = t; |
3546 | ||
3547 | if (hop) | |
3548 | ai->base = hop; | |
3549 | else | |
3550 | ai->base = sym; | |
3551 | ||
2f1a1aea FCE |
3552 | while (1) |
3553 | { | |
3554 | ai->indexes.push_back (parse_expression ()); | |
d7f3e0c5 | 3555 | if (peek_op ("]")) |
dff50e09 | 3556 | { |
731a5359 | 3557 | swallow (); |
dff50e09 | 3558 | break; |
d7f3e0c5 GH |
3559 | } |
3560 | else if (peek_op (",")) | |
3561 | { | |
731a5359 | 3562 | swallow (); |
d7f3e0c5 GH |
3563 | continue; |
3564 | } | |
2f1a1aea | 3565 | else |
2677d2fb | 3566 | throw parse_error (_("expected ',' or ']'")); |
2f1a1aea FCE |
3567 | } |
3568 | return ai; | |
3569 | } | |
d02548c0 GH |
3570 | |
3571 | // If we got to here, we *should* have a symbol; if we have | |
3572 | // a hist_op on its own, it doesn't count as an expression, | |
3573 | // so we throw a parse error. | |
3574 | ||
3575 | if (hop) | |
2677d2fb | 3576 | throw parse_error(_("base histogram operator where expression expected"), t); |
dff50e09 FCE |
3577 | |
3578 | return sym; | |
2f1a1aea | 3579 | } |
56099f08 | 3580 | |
30263a73 FCE |
3581 | // Parse a @cast or $var. Given head token has already been consumed. |
3582 | target_symbol* parser::parse_target_symbol (const token* t) | |
3583 | { | |
d48afc20 JS |
3584 | bool addressof = false; |
3585 | if (t->type == tok_operator && t->content == "&") | |
3586 | { | |
3587 | addressof = true; | |
3819d181 MW |
3588 | // Don't delete t before trying next token. |
3589 | // We might need it in the error message when there is no next token. | |
3590 | const token *next_t = next (); | |
731a5359 | 3591 | delete t; |
3819d181 | 3592 | t = next_t; |
d48afc20 JS |
3593 | } |
3594 | ||
06219d6f | 3595 | if (t->type == tok_operator && t->content == "@cast") |
30263a73 FCE |
3596 | { |
3597 | cast_op *cop = new cast_op; | |
3598 | cop->tok = t; | |
277c21bc | 3599 | cop->name = t->content; |
30263a73 FCE |
3600 | expect_op("("); |
3601 | cop->operand = parse_expression (); | |
3602 | expect_op(","); | |
7f6b80bd | 3603 | expect_unknown(tok_string, cop->type_name); |
30263a73 FCE |
3604 | if (peek_op (",")) |
3605 | { | |
731a5359 | 3606 | swallow (); |
30263a73 FCE |
3607 | expect_unknown(tok_string, cop->module); |
3608 | } | |
3609 | expect_op(")"); | |
3610 | parse_target_symbol_components(cop); | |
d48afc20 | 3611 | cop->addressof = addressof; |
30263a73 FCE |
3612 | return cop; |
3613 | } | |
3614 | ||
3615 | if (t->type == tok_identifier && t->content[0]=='$') | |
3616 | { | |
3617 | // target_symbol time | |
3618 | target_symbol *tsym = new target_symbol; | |
3619 | tsym->tok = t; | |
277c21bc | 3620 | tsym->name = t->content; |
30263a73 | 3621 | parse_target_symbol_components(tsym); |
d48afc20 | 3622 | tsym->addressof = addressof; |
30263a73 FCE |
3623 | return tsym; |
3624 | } | |
3625 | ||
06219d6f | 3626 | if (t->type == tok_operator && t->content == "@var") |
cc9001af | 3627 | { |
bd1fcbad YZ |
3628 | atvar_op *aop = new atvar_op; |
3629 | aop->tok = t; | |
3630 | aop->name = t->content; | |
cc9001af | 3631 | expect_op("("); |
bd1fcbad YZ |
3632 | expect_unknown(tok_string, aop->target_name); |
3633 | size_t found_at = aop->target_name.find("@"); | |
bfa7e523 | 3634 | if (found_at != string::npos) |
bd1fcbad | 3635 | aop->cu_name = aop->target_name.substr(found_at + 1); |
bfa7e523 | 3636 | else |
bd1fcbad YZ |
3637 | aop->cu_name = ""; |
3638 | if (peek_op (",")) | |
3639 | { | |
3640 | swallow (); | |
3641 | expect_unknown (tok_string, aop->module); | |
3642 | } | |
3643 | else | |
3644 | aop->module = ""; | |
cc9001af | 3645 | expect_op(")"); |
bd1fcbad YZ |
3646 | parse_target_symbol_components(aop); |
3647 | aop->addressof = addressof; | |
3648 | return aop; | |
cc9001af MW |
3649 | } |
3650 | ||
3651 | throw parse_error (_("expected @cast, @var or $var")); | |
30263a73 FCE |
3652 | } |
3653 | ||
3654 | ||
3655 | // Parse a @defined(). Given head token has already been consumed. | |
3656 | expression* parser::parse_defined_op (const token* t) | |
3657 | { | |
3658 | defined_op* dop = new defined_op; | |
3659 | dop->tok = t; | |
3660 | expect_op("("); | |
30263a73 | 3661 | // no need for parse_hist_op... etc., as @defined takes only target_symbols as its operand. |
d48afc20 | 3662 | const token* tt = next (); |
30263a73 FCE |
3663 | dop->operand = parse_target_symbol (tt); |
3664 | expect_op(")"); | |
3665 | return dop; | |
3666 | } | |
3667 | ||
3668 | ||
8cc799a5 JS |
3669 | // Parse a @entry(). Given head token has already been consumed. |
3670 | expression* parser::parse_entry_op (const token* t) | |
3671 | { | |
3672 | entry_op* eop = new entry_op; | |
3673 | eop->tok = t; | |
3674 | expect_op("("); | |
3675 | eop->operand = parse_expression (); | |
3676 | expect_op(")"); | |
3677 | return eop; | |
3678 | } | |
3679 | ||
3680 | ||
3689db05 SC |
3681 | // Parse a @perf(). Given head token has already been consumed. |
3682 | expression* parser::parse_perf_op (const token* t) | |
3683 | { | |
3684 | perf_op* pop = new perf_op; | |
3e6a17ee SC |
3685 | |
3686 | if (strverscmp(session.compatible.c_str(), "2.1") < 0) | |
3687 | throw parse_error (_("expected @cast, @var or $var")); | |
3688 | ||
3689db05 SC |
3689 | pop->tok = t; |
3690 | expect_op("("); | |
ace7c23f FCE |
3691 | pop->operand = parse_literal_string (); |
3692 | if (pop->operand->value == "") | |
3693 | throw parse_error (_("expected non-empty string")); | |
3689db05 SC |
3694 | expect_op(")"); |
3695 | return pop; | |
3696 | } | |
3697 | ||
3698 | ||
30263a73 | 3699 | |
81931eab JS |
3700 | void |
3701 | parser::parse_target_symbol_components (target_symbol* e) | |
3702 | { | |
5f36109e JS |
3703 | bool pprint = false; |
3704 | ||
3705 | // check for pretty-print in the form $foo$ | |
277c21bc | 3706 | string &base = e->name; |
5f36109e JS |
3707 | size_t pprint_pos = base.find_last_not_of('$'); |
3708 | if (0 < pprint_pos && pprint_pos < base.length() - 1) | |
3709 | { | |
3710 | string pprint_val = base.substr(pprint_pos + 1); | |
3711 | base.erase(pprint_pos + 1); | |
3712 | e->components.push_back (target_symbol::component(e->tok, pprint_val, true)); | |
3713 | pprint = true; | |
3714 | } | |
3715 | ||
3716 | while (!pprint) | |
81931eab | 3717 | { |
81931eab JS |
3718 | if (peek_op ("->")) |
3719 | { | |
c67847a0 JS |
3720 | const token* t = next(); |
3721 | string member; | |
3722 | expect_ident_or_keyword (member); | |
5f36109e JS |
3723 | |
3724 | // check for pretty-print in the form $foo->$ or $foo->bar$ | |
3725 | pprint_pos = member.find_last_not_of('$'); | |
3726 | string pprint_val; | |
3727 | if (pprint_pos == string::npos || pprint_pos < member.length() - 1) | |
3728 | { | |
3729 | pprint_val = member.substr(pprint_pos + 1); | |
3730 | member.erase(pprint_pos + 1); | |
3731 | pprint = true; | |
3732 | } | |
3733 | ||
3734 | if (!member.empty()) | |
3735 | e->components.push_back (target_symbol::component(t, member)); | |
3736 | if (pprint) | |
3737 | e->components.push_back (target_symbol::component(t, pprint_val, true)); | |
81931eab JS |
3738 | } |
3739 | else if (peek_op ("[")) | |
3740 | { | |
c67847a0 | 3741 | const token* t = next(); |
6fda2dff JS |
3742 | expression* index = parse_expression(); |
3743 | literal_number* ln = dynamic_cast<literal_number*>(index); | |
3744 | if (ln) | |
3745 | e->components.push_back (target_symbol::component(t, ln->value)); | |
3746 | else | |
3747 | e->components.push_back (target_symbol::component(t, index)); | |
81931eab | 3748 | expect_op ("]"); |
81931eab JS |
3749 | } |
3750 | else | |
3751 | break; | |
3752 | } | |
5f36109e JS |
3753 | |
3754 | if (!pprint) | |
3755 | { | |
3756 | // check for pretty-print in the form $foo $ | |
3757 | // i.e. as a separate token, esp. for $foo[i]$ and @cast(...)$ | |
3758 | const token* t = peek(); | |
3819d181 | 3759 | if (t != NULL && t->type == tok_identifier && |
5f36109e JS |
3760 | t->content.find_first_not_of('$') == string::npos) |
3761 | { | |
3762 | t = next(); | |
3763 | e->components.push_back (target_symbol::component(t, t->content, true)); | |
3764 | pprint = true; | |
3765 | } | |
3766 | } | |
3767 | ||
3768 | if (pprint && (peek_op ("->") || peek_op("["))) | |
ce0f6648 | 3769 | throw parse_error(_("-> and [ are not accepted for a pretty-printing variable")); |
81931eab JS |
3770 | } |
3771 | ||
73267b89 | 3772 | /* vim: set sw=2 ts=8 cino=>4,n-2,{2,^-2,t0,(0,u0,w1,M1 : */ |