]>
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 | { | |
1b1b4ceb | 318 | string align_parse_error (" "); |
2f1a1aea | 319 | |
16fc963f SM |
320 | const token *tok = pe.tok ? pe.tok : last_t; |
321 | ||
322 | // print either pe.what() or a deferred error from the lexer | |
323 | bool found_junk = false; | |
324 | if (tok && tok->type == tok_junk && tok->msg != "") | |
177a8ead | 325 | { |
16fc963f SM |
326 | found_junk = true; |
327 | cerr << _("parse error: ") << tok->msg << endl; | |
177a8ead | 328 | } |
2f1a1aea | 329 | else |
177a8ead | 330 | { |
16fc963f SM |
331 | cerr << _("parse error: ") << pe.what() << endl; |
332 | } | |
333 | ||
334 | // NB: It makes sense for lexer errors to always override parser | |
335 | // errors, since the original obvious scheme was for the lexer to | |
336 | // throw an exception before the token reached the parser. | |
337 | ||
338 | if (pe.tok || found_junk) | |
339 | { | |
340 | cerr << _("\tat: ") << *tok << endl; | |
341 | session.print_error_source (cerr, align_parse_error, tok); | |
342 | } | |
343 | else if (tok) // "expected" type error | |
344 | { | |
345 | cerr << _("\tsaw: ") << *tok << endl; | |
346 | session.print_error_source (cerr, align_parse_error, tok); | |
347 | } | |
348 | else | |
349 | { | |
350 | cerr << _("\tsaw: ") << input_name << " EOF" << endl; | |
177a8ead | 351 | } |
2f1a1aea | 352 | |
534aad8b SM |
353 | // print chained macro invocations |
354 | while (tok && tok->chain) { | |
355 | tok = tok->chain; | |
356 | cerr << _("\tin expansion of macro: ") << *tok << endl; | |
357 | session.print_error_source (cerr, align_parse_error, tok); | |
358 | } | |
359 | ||
2f1a1aea FCE |
360 | num_errors ++; |
361 | } | |
362 | ||
363 | ||
2f1a1aea | 364 | |
c434ec7e FCE |
365 | |
366 | template <typename OPERAND> | |
367 | bool eval_comparison (const OPERAND& lhs, const token* op, const OPERAND& rhs) | |
368 | { | |
369 | if (op->type == tok_operator && op->content == "<=") | |
370 | { return lhs <= rhs; } | |
371 | else if (op->type == tok_operator && op->content == ">=") | |
372 | { return lhs >= rhs; } | |
373 | else if (op->type == tok_operator && op->content == "<") | |
374 | { return lhs < rhs; } | |
375 | else if (op->type == tok_operator && op->content == ">") | |
376 | { return lhs > rhs; } | |
377 | else if (op->type == tok_operator && op->content == "==") | |
378 | { return lhs == rhs; } | |
379 | else if (op->type == tok_operator && op->content == "!=") | |
380 | { return lhs != rhs; } | |
381 | else | |
2677d2fb | 382 | throw parse_error (_("expected comparison operator"), op); |
c434ec7e FCE |
383 | } |
384 | ||
385 | ||
534aad8b SM |
386 | // Here, we perform on-the-fly preprocessing in two passes. |
387 | ||
388 | // First pass - macro declaration and expansion. | |
389 | // | |
390 | // The basic form of a declaration is @define SIGNATURE %( BODY %) | |
391 | // where SIGNATURE is of the form macro_name (a, b, c, ...) | |
392 | // and BODY can obtain the parameter contents as @a, @b, @c, .... | |
393 | // Note that parameterless macros can also be declared. | |
394 | // | |
3932c705 | 395 | // Macro definitions may not be nested. |
534aad8b SM |
396 | // A macro is available textually after it has been defined. |
397 | // | |
398 | // The basic form of a macro invocation | |
399 | // for a parameterless macro is @macro_name, | |
400 | // for a macro with parameters is @macro_name(param_1, param_2, ...). | |
401 | // | |
402 | // TODOXXX NB: this means that a parameterless macro @foo called as | |
403 | // @foo(a, b, c) leaves its 'parameters' alone, rather than consuming | |
404 | // them to result in a "too many parameters error". | |
405 | // | |
406 | // Invocations of unknown macros are left unexpanded, to allow | |
407 | // the continued use of constructs such as @cast, @var, etc. | |
408 | ||
fe410f52 | 409 | macrodecl::~macrodecl () |
534aad8b SM |
410 | { |
411 | delete tok; | |
412 | for (vector<const token*>::iterator it = body.begin(); | |
413 | it != body.end(); it++) | |
414 | delete *it; | |
415 | } | |
416 | ||
417 | parser::pp1_activation::~pp1_activation () | |
418 | { | |
419 | delete tok; | |
fe410f52 | 420 | if (curr_macro->is_closure()) return; // body is shared with an earlier declaration |
534aad8b SM |
421 | for (map<string, pp_macrodecl*>::iterator it = params.begin(); |
422 | it != params.end(); it++) | |
423 | delete it->second; | |
424 | } | |
425 | ||
426 | // Grab a token from the current input source (main file or macro body): | |
427 | const token* | |
428 | parser::next_pp1 () | |
429 | { | |
430 | if (pp1_state.empty()) | |
431 | return input.scan (); | |
432 | ||
433 | // otherwise, we're inside a macro | |
434 | pp1_activation* act = pp1_state.back(); | |
435 | unsigned& cursor = act->cursor; | |
436 | if (cursor < act->curr_macro->body.size()) | |
437 | { | |
438 | token* t = new token(*act->curr_macro->body[cursor]); | |
0002fc51 | 439 | t->chain = new token(*act->tok); // mark chained token |
534aad8b SM |
440 | cursor++; |
441 | return t; | |
442 | } | |
443 | else | |
444 | return 0; // reached end of macro body | |
445 | } | |
446 | ||
447 | const token* | |
448 | parser::scan_pp1 () | |
449 | { | |
450 | while (true) | |
451 | { | |
452 | const token* t = next_pp1 (); | |
453 | if (t == 0) // EOF or end of macro body | |
454 | { | |
455 | if (pp1_state.empty()) // actual EOF | |
456 | return 0; | |
457 | ||
458 | // Exit macro and loop around to look for the next token. | |
459 | pp1_activation* act = pp1_state.back(); | |
460 | pp1_state.pop_back(); delete act; | |
461 | continue; | |
462 | } | |
463 | ||
464 | // macro definition | |
465 | if (t->type == tok_operator && t->content == "@define") | |
466 | { | |
467 | if (!pp1_state.empty()) | |
468 | throw parse_error (_("'@define' forbidden inside macro body"), t); | |
469 | delete t; | |
470 | ||
471 | // handle macro definition | |
472 | // (1) consume macro signature | |
3932c705 | 473 | t = input.scan(); |
534aad8b SM |
474 | if (! (t && t->type == tok_identifier)) |
475 | throw parse_error (_("expected identifier"), t); | |
476 | string name = t->content; | |
477 | ||
478 | // check for redefinition of existing macro | |
479 | if (pp1_namespace.find(name) != pp1_namespace.end()) | |
480 | // TODOXXX use a slightly different chaining hack to also point to | |
481 | // pp1_namespace[name]->tok, the site of the original definition? | |
482 | throw parse_error (_F("attempt to redefine macro '@%s' in the same file", name.c_str ()), t); | |
483 | // TODOXXX this is only really necessary if we want to leave open the possibility of statically-scoped semantics in the future...? | |
484 | ||
1d94e4e5 | 485 | // XXX this cascades into further parse errors as the |
534aad8b SM |
486 | // parser tries to parse the remaining definition... |
487 | if (name == "define") | |
488 | throw parse_error (_("attempt to redefine '@define'"), t); | |
489 | if (input.atwords.count("@" + name)) | |
490 | session.print_warning (_F("macro redefines built-in operator '@%s'", name.c_str()), t); | |
491 | ||
fe410f52 SM |
492 | macrodecl* decl = (pp1_namespace[name] = new macrodecl); |
493 | decl->tok = t; | |
534aad8b SM |
494 | |
495 | // determine if the macro takes parameters | |
4ac28d7e | 496 | bool saw_params = false; |
3932c705 SM |
497 | t = input.scan(); |
498 | if (t && t->type == tok_operator && t->content == "(") | |
4ac28d7e SM |
499 | { |
500 | saw_params = true; | |
501 | do | |
502 | { | |
503 | delete t; | |
504 | ||
505 | t = input.scan (); | |
506 | if (! (t && t->type == tok_identifier)) | |
507 | throw parse_error(_("expected identifier"), t); | |
fe410f52 | 508 | decl->formal_args.push_back(t->content); |
4ac28d7e SM |
509 | delete t; |
510 | ||
511 | t = input.scan (); | |
512 | if (t && t->type == tok_operator && t->content == ",") | |
513 | { | |
514 | continue; | |
515 | } | |
516 | else if (t && t->type == tok_operator && t->content == ")") | |
517 | { | |
518 | delete t; | |
519 | t = input.scan(); | |
520 | break; | |
521 | } | |
522 | else | |
523 | { | |
524 | throw parse_error (_("expected ',' or ')'"), t); | |
525 | } | |
526 | } | |
527 | while (true); | |
528 | } | |
534aad8b SM |
529 | |
530 | // (2) identify & consume macro body | |
3932c705 | 531 | if (! (t && t->type == tok_operator && t->content == "%(")) |
4ac28d7e SM |
532 | { |
533 | if (saw_params) | |
534 | throw parse_error (_("expected '%('"), t); | |
535 | else | |
536 | throw parse_error (_("expected '%(' or '('"), t); | |
537 | } | |
3932c705 | 538 | delete t; |
534aad8b | 539 | |
3932c705 SM |
540 | t = slurp_pp1_body (decl->body); |
541 | if (!t) | |
542 | throw parse_error (_("incomplete macro definition - missing '%)'"), decl->tok); | |
543 | delete t; | |
534aad8b SM |
544 | |
545 | // Now loop around to look for a real token. | |
546 | continue; | |
547 | } | |
548 | ||
549 | // (potential) macro invocation | |
550 | if (t->type == tok_operator && t->content[0] == '@') | |
551 | { | |
552 | string name = t->content.substr(1); // strip initial '@' | |
553 | ||
554 | // check if name refers to a real parameter or macro | |
fe410f52 | 555 | macrodecl* decl; |
534aad8b SM |
556 | pp1_activation* act = pp1_state.empty() ? 0 : pp1_state.back(); |
557 | if (act && act->params.find(name) != act->params.end()) | |
558 | decl = act->params[name]; | |
fe410f52 SM |
559 | else if (!(act && act->curr_macro->context == ctx_library) |
560 | && pp1_namespace.find(name) != pp1_namespace.end()) | |
534aad8b | 561 | decl = pp1_namespace[name]; |
fe410f52 SM |
562 | else if (session.library_macros.find(name) |
563 | != session.library_macros.end()) | |
564 | decl = session.library_macros[name]; | |
534aad8b SM |
565 | else // this is an ordinary @operator |
566 | return t; | |
567 | ||
568 | // handle macro invocation | |
569 | pp1_activation *new_act = new pp1_activation(*t, decl); | |
fe410f52 | 570 | unsigned num_params = decl->formal_args.size(); |
534aad8b SM |
571 | |
572 | // (1a) restore parameter invocation closure | |
fe410f52 | 573 | if (num_params == 0 && decl->is_closure()) |
534aad8b SM |
574 | { |
575 | // NB: decl->parent_act is always safe since the | |
576 | // parameter decl (if any) comes from an activation | |
577 | // record which deeper in the stack than new_act. | |
578 | ||
579 | // decl is a macro parameter which must be evaluated in | |
580 | // the context of the original point of invocation: | |
fe410f52 | 581 | new_act->params = ((pp_macrodecl*)decl)->parent_act->params; |
534aad8b SM |
582 | goto expand; |
583 | } | |
584 | ||
585 | // (1b) consume macro parameters (if any) | |
586 | if (num_params == 0) | |
587 | goto expand; | |
588 | ||
589 | // for simplicity, we do not allow macro constructs here | |
590 | // -- if we did, we'd have to recursively call scan_pp1() | |
591 | t = next_pp1 (); | |
39566df2 | 592 | if (! (t && t->type == tok_operator && t->content == "(")) |
534aad8b SM |
593 | { |
594 | delete new_act; | |
52c2652f | 595 | throw parse_error (_NF |
534aad8b | 596 | ("expected '(' in invocation of macro '@%s'" |
f499dee5 | 597 | " taking %d parameter", |
534aad8b | 598 | "expected '(' in invocation of macro '@%s'" |
f499dee5 | 599 | " taking %d parameters", |
52c2652f | 600 | num_params, name.c_str(), num_params), t); |
534aad8b SM |
601 | } |
602 | ||
603 | // XXX perhaps parse/count the full number of params, | |
604 | // so we can say "expected x, found y params" on error? | |
605 | for (unsigned i = 0; i < num_params; i++) | |
606 | { | |
607 | delete t; | |
608 | ||
609 | // create parameter closure | |
fe410f52 | 610 | string param_name = decl->formal_args[i]; |
534aad8b | 611 | pp_macrodecl* p = (new_act->params[param_name] |
fe410f52 SM |
612 | = new pp_macrodecl); |
613 | p->tok = new token(*new_act->tok); | |
614 | p->parent_act = act; | |
534aad8b SM |
615 | // NB: *new_act->tok points to invocation, act is NULL at top level |
616 | ||
617 | t = slurp_pp1_param (p->body); | |
618 | ||
619 | // check correct usage of ',' or ')' | |
620 | if (t == 0) // hit unexpected EOF or end of macro | |
621 | { | |
622 | // XXX could we pop the stack and continue parsing | |
623 | // the invocation, allowing macros to construct new | |
624 | // invocations in piecemeal fashion?? | |
625 | const token* orig_t = new token(*new_act->tok); | |
626 | delete new_act; | |
627 | throw parse_error (_("could not find end of macro invocation"), orig_t); | |
628 | } | |
629 | if (t->type == tok_operator && t->content == ",") | |
630 | { | |
631 | if (i + 1 == num_params) | |
632 | { | |
633 | delete new_act; | |
634 | throw parse_error (_F("too many parameters for macro '@%s' (expected %d)", name.c_str(), num_params), t); | |
635 | } | |
636 | } | |
637 | else if (t->type == tok_operator && t->content == ")") | |
638 | { | |
639 | if (i + 1 != num_params) | |
640 | { | |
641 | delete new_act; | |
642 | throw parse_error (_F("too few parameters for macro '@%s' (expected %d)", name.c_str(), num_params), t); | |
643 | } | |
644 | } | |
645 | else | |
646 | { | |
647 | // XXX this is, incidentally, impossible | |
648 | delete new_act; | |
649 | throw parse_error(_("expected ',' or ')' after macro parameter"), t); | |
650 | } | |
651 | } | |
652 | ||
653 | delete t; | |
654 | ||
655 | // (2) set up macro expansion | |
656 | expand: | |
657 | pp1_state.push_back (new_act); | |
658 | ||
659 | // Now loop around to look for a real token. | |
660 | continue; | |
661 | } | |
662 | ||
663 | // Otherwise, we have an ordinary token. | |
664 | return t; | |
665 | } | |
666 | } | |
667 | ||
668 | // Consume a single macro invocation's parameters, heeding nested ( ) | |
669 | // brackets and stopping on an unbalanced ')' or an unbracketed ',' | |
670 | // (and returning the final separator token). | |
671 | const token* | |
672 | parser::slurp_pp1_param (vector<const token*>& param) | |
673 | { | |
674 | const token* t = 0; | |
675 | unsigned nesting = 0; | |
676 | do | |
677 | { | |
678 | t = next_pp1 (); | |
679 | ||
680 | if (!t) | |
681 | break; | |
682 | if (t->type == tok_operator && t->content == "(") | |
683 | ++nesting; | |
684 | else if (nesting && t->type == tok_operator && t->content == ")") | |
685 | --nesting; | |
686 | else if (!nesting && t->type == tok_operator | |
687 | && (t->content == ")" || t->content == ",")) | |
688 | break; | |
689 | param.push_back(t); | |
690 | } | |
691 | while (true); | |
692 | return t; // report ")" or "," or NULL | |
693 | } | |
694 | ||
695 | ||
696 | // Consume a macro declaration's body, heeding nested %( %) brackets. | |
697 | const token* | |
698 | parser::slurp_pp1_body (vector<const token*>& body) | |
699 | { | |
700 | const token* t = 0; | |
701 | unsigned nesting = 0; | |
702 | do | |
703 | { | |
704 | t = next_pp1 (); | |
705 | ||
706 | if (!t) | |
707 | break; | |
708 | if (t->type == tok_operator && t->content == "%(") | |
709 | ++nesting; | |
710 | else if (nesting && t->type == tok_operator && t->content == "%)") | |
711 | --nesting; | |
712 | else if (!nesting && t->type == tok_operator && t->content == "%)") | |
713 | break; | |
714 | body.push_back(t); | |
715 | } | |
716 | while (true); | |
717 | return t; // report final "%)" or NULL | |
718 | } | |
719 | ||
fe410f52 SM |
720 | // Used for parsing .stpm files. |
721 | stapfile* | |
722 | parser::parse_library_macros () | |
723 | { | |
724 | stapfile* f = new stapfile; | |
725 | input.set_current_file (f); | |
726 | ||
727 | try | |
728 | { | |
729 | const token* t = scan_pp1 (); | |
730 | ||
731 | // Currently we only take objection to macro invocations if they | |
732 | // produce a non-whitespace token after being expanded. | |
733 | ||
734 | // XXX should we prevent macro invocations even if they expand to empty?? | |
735 | ||
736 | if (t != 0) | |
737 | throw parse_error (_F("library macro file '%s' contains non-@define construct", input_name.c_str()), t); | |
738 | ||
739 | // We need to first check whether *any* of the macros are duplicates, | |
740 | // then commit to including the entire file in the global namespace | |
741 | // (or not). Yuck. | |
742 | for (map<string, macrodecl*>::iterator it = pp1_namespace.begin(); | |
743 | it != pp1_namespace.end(); it++) | |
744 | { | |
745 | string name = it->first; | |
746 | ||
747 | if (session.library_macros.find(name) != session.library_macros.end()) | |
748 | { | |
749 | // XXX ugly hack simulates chaining | |
750 | parse_error* er1 = new parse_error (_F("duplicate definition of library macro '%s'", name.c_str()), it->second->tok); | |
751 | parse_error* er2 = new parse_error (_("location of original definition was"), session.library_macros[name]->tok); | |
752 | print_error (*er1); | |
753 | print_error (*er2); | |
754 | delete er1; delete er2; | |
755 | ||
756 | delete f; | |
757 | return 0; | |
758 | } | |
759 | } | |
760 | ||
761 | } | |
762 | catch (const parse_error& pe) | |
763 | { | |
764 | print_error (pe); | |
765 | delete f; | |
766 | return 0; | |
767 | } | |
768 | ||
769 | // If no errors, include the entire file. Note how this is outside | |
770 | // of the try-catch block -- no errors possible. | |
771 | for (map<string, macrodecl*>::iterator it = pp1_namespace.begin(); | |
772 | it != pp1_namespace.end(); it++) | |
773 | { | |
774 | string name = it->first; | |
775 | ||
776 | session.library_macros[name] = it->second; | |
777 | session.library_macros[name]->context = ctx_library; | |
778 | // TODOXXX be sure declaration is retained and not deleted | |
779 | } | |
780 | ||
781 | return f; | |
782 | } | |
783 | ||
534aad8b SM |
784 | // Second pass - preprocessor conditional expansion. |
785 | // | |
177a8ead | 786 | // The basic form is %( CONDITION %? THEN-TOKENS %: ELSE-TOKENS %) |
44ce8ed5 FCE |
787 | // where CONDITION is: kernel_v[r] COMPARISON-OP "version-string" |
788 | // or: arch COMPARISON-OP "arch-string" | |
db135493 | 789 | // or: systemtap_v COMPARISON-OP "version-string" |
2e6dd9d0 | 790 | // or: systemtap_privilege COMPARISON-OP "privilege-string" |
561079c8 | 791 | // or: CONFIG_foo COMPARISON-OP "config-string" |
717a457b | 792 | // or: CONFIG_foo COMPARISON-OP number |
4227f98d | 793 | // or: CONFIG_foo COMPARISON-OP CONFIG_bar |
5811366a FCE |
794 | // or: "string1" COMPARISON-OP "string2" |
795 | // or: number1 COMPARISON-OP number2 | |
44ce8ed5 | 796 | // The %: ELSE-TOKENS part is optional. |
177a8ead FCE |
797 | // |
798 | // e.g. %( kernel_v > "2.5" %? "foo" %: "baz" %) | |
c434ec7e | 799 | // e.g. %( arch != "i?86" %? "foo" %: "baz" %) |
561079c8 | 800 | // e.g. %( CONFIG_foo %? "foo" %: "baz" %) |
177a8ead FCE |
801 | // |
802 | // Up to an entire %( ... %) expression is processed by a single call | |
803 | // to this function. Tokens included by any nested conditions are | |
804 | // enqueued in a private vector. | |
805 | ||
806 | bool eval_pp_conditional (systemtap_session& s, | |
807 | const token* l, const token* op, const token* r) | |
808 | { | |
44ce8ed5 | 809 | if (l->type == tok_identifier && (l->content == "kernel_v" || |
db135493 FCE |
810 | l->content == "kernel_vr" || |
811 | l->content == "systemtap_v")) | |
44ce8ed5 | 812 | { |
db135493 | 813 | if (! (r->type == tok_string)) |
2677d2fb | 814 | throw parse_error (_("expected string literal"), r); |
db135493 | 815 | |
44ce8ed5 | 816 | string target_kernel_vr = s.kernel_release; |
197a4d62 | 817 | string target_kernel_v = s.kernel_base_release; |
db135493 | 818 | string target; |
dff50e09 | 819 | |
db135493 FCE |
820 | if (l->content == "kernel_v") target = target_kernel_v; |
821 | else if (l->content == "kernel_vr") target = target_kernel_vr; | |
822 | else if (l->content == "systemtap_v") target = s.compatible; | |
823 | else assert (0); | |
7a468d68 | 824 | |
7a468d68 FCE |
825 | string query = r->content; |
826 | bool rhs_wildcard = (strpbrk (query.c_str(), "*?[") != 0); | |
827 | ||
44ce8ed5 FCE |
828 | // collect acceptable strverscmp results. |
829 | int rvc_ok1, rvc_ok2; | |
7a468d68 | 830 | bool wc_ok = false; |
44ce8ed5 FCE |
831 | if (op->type == tok_operator && op->content == "<=") |
832 | { rvc_ok1 = -1; rvc_ok2 = 0; } | |
833 | else if (op->type == tok_operator && op->content == ">=") | |
834 | { rvc_ok1 = 1; rvc_ok2 = 0; } | |
835 | else if (op->type == tok_operator && op->content == "<") | |
836 | { rvc_ok1 = -1; rvc_ok2 = -1; } | |
837 | else if (op->type == tok_operator && op->content == ">") | |
838 | { rvc_ok1 = 1; rvc_ok2 = 1; } | |
839 | else if (op->type == tok_operator && op->content == "==") | |
7a468d68 | 840 | { rvc_ok1 = 0; rvc_ok2 = 0; wc_ok = true; } |
44ce8ed5 | 841 | else if (op->type == tok_operator && op->content == "!=") |
7a468d68 | 842 | { rvc_ok1 = -1; rvc_ok2 = 1; wc_ok = true; } |
44ce8ed5 | 843 | else |
2677d2fb | 844 | throw parse_error (_("expected comparison operator"), op); |
7a468d68 FCE |
845 | |
846 | if ((!wc_ok) && rhs_wildcard) | |
2677d2fb | 847 | throw parse_error (_("wildcard not allowed with order comparison operators"), op); |
7a468d68 FCE |
848 | |
849 | if (rhs_wildcard) | |
850 | { | |
851 | int rvc_result = fnmatch (query.c_str(), target.c_str(), | |
852 | FNM_NOESCAPE); // spooky | |
853 | bool badness = (rvc_result == 0) ^ (op->content == "=="); | |
854 | return !badness; | |
855 | } | |
856 | else | |
857 | { | |
858 | int rvc_result = strverscmp (target.c_str(), query.c_str()); | |
859 | // normalize rvc_result | |
860 | if (rvc_result < 0) rvc_result = -1; | |
861 | if (rvc_result > 0) rvc_result = 1; | |
862 | return (rvc_result == rvc_ok1 || rvc_result == rvc_ok2); | |
863 | } | |
44ce8ed5 | 864 | } |
2e6dd9d0 SM |
865 | else if (l->type == tok_identifier && l->content == "systemtap_privilege") |
866 | { | |
867 | string target_privilege = | |
cba5b802 SM |
868 | /* XXX perhaps include a "guru" state */ |
869 | pr_contains(s.privilege, pr_stapdev) ? "stapdev" | |
870 | : pr_contains(s.privilege, pr_stapsys) ? "stapsys" | |
871 | : pr_contains(s.privilege, pr_stapusr) ? "stapusr" | |
2e6dd9d0 SM |
872 | : "none"; /* should be impossible -- s.privilege always one of above */ |
873 | assert(target_privilege != "none"); | |
874 | ||
875 | if (! (r->type == tok_string)) | |
876 | throw parse_error (_("expected string literal"), r); | |
877 | string query_privilege = r->content; | |
878 | ||
879 | bool nomatch = (target_privilege != query_privilege); | |
880 | ||
881 | bool result; | |
882 | if (op->type == tok_operator && op->content == "==") | |
883 | result = !nomatch; | |
884 | else if (op->type == tok_operator && op->content == "!=") | |
885 | result = nomatch; | |
886 | else | |
887 | throw parse_error (_("expected '==' or '!='"), op); | |
cba5b802 | 888 | /* XXX perhaps allow <= >= and similar comparisons */ |
2e6dd9d0 SM |
889 | |
890 | return result; | |
891 | } | |
44ce8ed5 FCE |
892 | else if (l->type == tok_identifier && l->content == "arch") |
893 | { | |
894 | string target_architecture = s.architecture; | |
895 | if (! (r->type == tok_string)) | |
2677d2fb | 896 | throw parse_error (_("expected string literal"), r); |
44ce8ed5 | 897 | string query_architecture = r->content; |
dff50e09 | 898 | |
7a468d68 FCE |
899 | int nomatch = fnmatch (query_architecture.c_str(), |
900 | target_architecture.c_str(), | |
901 | FNM_NOESCAPE); // still spooky | |
902 | ||
561079c8 FCE |
903 | bool result; |
904 | if (op->type == tok_operator && op->content == "==") | |
905 | result = !nomatch; | |
906 | else if (op->type == tok_operator && op->content == "!=") | |
907 | result = nomatch; | |
908 | else | |
2677d2fb | 909 | throw parse_error (_("expected '==' or '!='"), op); |
561079c8 FCE |
910 | |
911 | return result; | |
912 | } | |
60d98537 | 913 | else if (l->type == tok_identifier && startswith(l->content, "CONFIG_")) |
561079c8 | 914 | { |
717a457b MW |
915 | if (r->type == tok_string) |
916 | { | |
917 | string lhs = s.kernel_config[l->content]; // may be empty | |
918 | string rhs = r->content; | |
561079c8 | 919 | |
717a457b | 920 | int nomatch = fnmatch (rhs.c_str(), lhs.c_str(), FNM_NOESCAPE); // still spooky |
561079c8 | 921 | |
717a457b MW |
922 | bool result; |
923 | if (op->type == tok_operator && op->content == "==") | |
924 | result = !nomatch; | |
925 | else if (op->type == tok_operator && op->content == "!=") | |
926 | result = nomatch; | |
927 | else | |
2677d2fb | 928 | throw parse_error (_("expected '==' or '!='"), op); |
dff50e09 | 929 | |
717a457b MW |
930 | return result; |
931 | } | |
932 | else if (r->type == tok_number) | |
933 | { | |
934 | const char* startp = s.kernel_config[l->content].c_str (); | |
935 | char* endp = (char*) startp; | |
936 | errno = 0; | |
937 | int64_t lhs = (int64_t) strtoll (startp, & endp, 0); | |
938 | if (errno == ERANGE || errno == EINVAL || *endp != '\0') | |
939 | throw parse_error ("Config option value not a number", l); | |
940 | ||
941 | int64_t rhs = lex_cast<int64_t>(r->content); | |
942 | return eval_comparison (lhs, op, rhs); | |
943 | } | |
4227f98d | 944 | else if (r->type == tok_identifier |
60d98537 | 945 | && startswith(r->content, "CONFIG_")) |
4227f98d MW |
946 | { |
947 | // First try to convert both to numbers, | |
948 | // otherwise threat both as strings. | |
949 | const char* startp = s.kernel_config[l->content].c_str (); | |
950 | char* endp = (char*) startp; | |
951 | errno = 0; | |
952 | int64_t val = (int64_t) strtoll (startp, & endp, 0); | |
953 | if (errno != ERANGE && errno != EINVAL && *endp == '\0') | |
954 | { | |
955 | int64_t lhs = val; | |
956 | startp = s.kernel_config[r->content].c_str (); | |
957 | endp = (char*) startp; | |
958 | errno = 0; | |
959 | int64_t rhs = (int64_t) strtoll (startp, & endp, 0); | |
960 | if (errno != ERANGE && errno != EINVAL && *endp == '\0') | |
961 | return eval_comparison (lhs, op, rhs); | |
962 | } | |
963 | ||
964 | string lhs = s.kernel_config[l->content]; | |
965 | string rhs = s.kernel_config[r->content]; | |
966 | return eval_comparison (lhs, op, rhs); | |
967 | } | |
717a457b | 968 | else |
ce0f6648 | 969 | throw parse_error (_("expected string, number literal or other CONFIG_... as right side operand"), r); |
dff50e09 | 970 | } |
c434ec7e | 971 | else if (l->type == tok_string && r->type == tok_string) |
5811366a | 972 | { |
c434ec7e FCE |
973 | string lhs = l->content; |
974 | string rhs = r->content; | |
975 | return eval_comparison (lhs, op, rhs); | |
976 | // NB: no wildcarding option here | |
977 | } | |
978 | else if (l->type == tok_number && r->type == tok_number) | |
979 | { | |
980 | int64_t lhs = lex_cast<int64_t>(l->content); | |
981 | int64_t rhs = lex_cast<int64_t>(r->content); | |
982 | return eval_comparison (lhs, op, rhs); | |
7a468d68 | 983 | // NB: no wildcarding option here |
5811366a FCE |
984 | } |
985 | else if (l->type == tok_string && r->type == tok_number | |
986 | && op->type == tok_operator) | |
2677d2fb | 987 | throw parse_error (_("expected string literal as right value"), r); |
5811366a FCE |
988 | else if (l->type == tok_number && r->type == tok_string |
989 | && op->type == tok_operator) | |
2677d2fb | 990 | throw parse_error (_("expected number literal as right value"), r); |
c434ec7e | 991 | |
177a8ead | 992 | else |
2677d2fb LB |
993 | throw parse_error (_("expected 'arch' or 'kernel_v' or 'kernel_vr' or 'CONFIG_...'\n" |
994 | " or comparison between strings or integers"), l); | |
177a8ead FCE |
995 | } |
996 | ||
997 | ||
5811366a | 998 | // Only tokens corresponding to the TRUE statement must be expanded |
177a8ead | 999 | const token* |
b5477cd9 | 1000 | parser::scan_pp () |
177a8ead FCE |
1001 | { |
1002 | while (true) | |
1003 | { | |
e92f2566 JS |
1004 | pp_state_t pp = PP_NONE; |
1005 | if (!pp_state.empty()) | |
1006 | pp = pp_state.back().second; | |
1007 | ||
1008 | const token* t = 0; | |
1009 | if (pp == PP_SKIP_THEN || pp == PP_SKIP_ELSE) | |
1010 | t = skip_pp (); | |
1011 | else | |
534aad8b | 1012 | t = scan_pp1 (); |
e92f2566 JS |
1013 | |
1014 | if (t == 0) // EOF | |
177a8ead | 1015 | { |
e92f2566 JS |
1016 | if (pp != PP_NONE) |
1017 | { | |
1018 | t = pp_state.back().first; | |
1019 | pp_state.pop_back(); // so skip_some doesn't keep trying to close this | |
ce0f6648 | 1020 | //TRANSLATORS: 'conditional' meaning 'conditional preprocessing' |
2677d2fb | 1021 | throw parse_error (_("incomplete conditional at end of file"), t); |
e92f2566 | 1022 | } |
177a8ead FCE |
1023 | return t; |
1024 | } | |
1025 | ||
e92f2566 JS |
1026 | // misplaced preprocessor "then" |
1027 | if (t->type == tok_operator && t->content == "%?") | |
2677d2fb | 1028 | throw parse_error (_("incomplete conditional - missing '%('"), t); |
e92f2566 JS |
1029 | |
1030 | // preprocessor "else" | |
1031 | if (t->type == tok_operator && t->content == "%:") | |
1032 | { | |
1033 | if (pp == PP_NONE) | |
2677d2fb | 1034 | throw parse_error (_("incomplete conditional - missing '%('"), t); |
e92f2566 | 1035 | if (pp == PP_KEEP_ELSE || pp == PP_SKIP_ELSE) |
2677d2fb | 1036 | throw parse_error (_("invalid conditional - duplicate '%:'"), t); |
1d94e4e5 SM |
1037 | // XXX: here and elsewhere, error cascades might be avoided |
1038 | // by dropping tokens until we reach the closing %) | |
e92f2566 JS |
1039 | |
1040 | pp_state.back().second = (pp == PP_KEEP_THEN) ? | |
1041 | PP_SKIP_ELSE : PP_KEEP_ELSE; | |
1042 | delete t; | |
1043 | continue; | |
1044 | } | |
1045 | ||
1046 | // preprocessor close | |
1047 | if (t->type == tok_operator && t->content == "%)") | |
1048 | { | |
1049 | if (pp == PP_NONE) | |
2677d2fb | 1050 | throw parse_error (_("incomplete conditional - missing '%('"), t); |
e92f2566 | 1051 | delete pp_state.back().first; |
a07a2c28 | 1052 | delete t; //this is the closing bracket |
e92f2566 JS |
1053 | pp_state.pop_back(); |
1054 | continue; | |
1055 | } | |
dff50e09 | 1056 | |
177a8ead FCE |
1057 | if (! (t->type == tok_operator && t->content == "%(")) // ordinary token |
1058 | return t; | |
1059 | ||
1060 | // We have a %( - it's time to throw a preprocessing party! | |
1061 | ||
2d7881bf PP |
1062 | bool result = false; |
1063 | bool and_result = true; | |
1064 | const token *n = NULL; | |
1065 | do { | |
1066 | const token *l, *op, *r; | |
534aad8b SM |
1067 | l = scan_pp1 (); |
1068 | op = scan_pp1 (); | |
1069 | r = scan_pp1 (); | |
2d7881bf | 1070 | if (l == 0 || op == 0 || r == 0) |
2677d2fb | 1071 | throw parse_error (_("incomplete condition after '%('"), t); |
2d7881bf PP |
1072 | // NB: consider generalizing to consume all tokens until %?, and |
1073 | // passing that as a vector to an evaluator. | |
1074 | ||
1075 | // Do not evaluate the condition if we haven't expanded everything. | |
1076 | // This may occur when having several recursive conditionals. | |
1077 | and_result &= eval_pp_conditional (session, l, op, r); | |
a07a2c28 LB |
1078 | if(l->content=="systemtap_v") |
1079 | systemtap_v_seen=r; | |
1080 | ||
1081 | else | |
1082 | delete r; | |
1083 | ||
2d7881bf PP |
1084 | delete l; |
1085 | delete op; | |
2d7881bf PP |
1086 | delete n; |
1087 | ||
534aad8b | 1088 | n = scan_pp1 (); |
2d7881bf PP |
1089 | if (n && n->type == tok_operator && n->content == "&&") |
1090 | continue; | |
1091 | result |= and_result; | |
1092 | and_result = true; | |
1093 | if (! (n && n->type == tok_operator && n->content == "||")) | |
1094 | break; | |
1095 | } while (true); | |
3f847830 FCE |
1096 | |
1097 | /* | |
1098 | clog << "PP eval (" << *t << ") == " << result << endl; | |
1099 | */ | |
1100 | ||
e92f2566 | 1101 | const token *m = n; |
177a8ead | 1102 | if (! (m && m->type == tok_operator && m->content == "%?")) |
2677d2fb | 1103 | throw parse_error (_("expected '%?' marker for conditional"), t); |
70c743d8 | 1104 | delete m; // "%?" |
177a8ead | 1105 | |
e92f2566 JS |
1106 | pp = result ? PP_KEEP_THEN : PP_SKIP_THEN; |
1107 | pp_state.push_back (make_pair (t, pp)); | |
3f847830 | 1108 | |
e92f2566 JS |
1109 | // Now loop around to look for a real token. |
1110 | } | |
1111 | } | |
3f847830 | 1112 | |
3f847830 | 1113 | |
e92f2566 JS |
1114 | // Skip over tokens and any errors, heeding |
1115 | // only nested preprocessor starts and ends. | |
1116 | const token* | |
1117 | parser::skip_pp () | |
1118 | { | |
1119 | const token* t = 0; | |
1120 | unsigned nesting = 0; | |
1121 | do | |
1122 | { | |
1123 | try | |
1124 | { | |
534aad8b | 1125 | t = scan_pp1 (); |
177a8ead | 1126 | } |
e92f2566 | 1127 | catch (const parse_error &e) |
70c743d8 | 1128 | { |
e92f2566 | 1129 | continue; |
70c743d8 | 1130 | } |
e92f2566 JS |
1131 | if (!t) |
1132 | break; | |
1133 | if (t->type == tok_operator && t->content == "%(") | |
1134 | ++nesting; | |
1135 | else if (nesting && t->type == tok_operator && t->content == "%)") | |
1136 | --nesting; | |
1137 | else if (!nesting && t->type == tok_operator && | |
1138 | (t->content == "%:" || t->content == "%?" || t->content == "%)")) | |
1139 | break; | |
1140 | delete t; | |
177a8ead | 1141 | } |
e92f2566 JS |
1142 | while (true); |
1143 | return t; | |
177a8ead FCE |
1144 | } |
1145 | ||
1146 | ||
2f1a1aea | 1147 | const token* |
b5477cd9 | 1148 | parser::next () |
2f1a1aea FCE |
1149 | { |
1150 | if (! next_t) | |
b5477cd9 | 1151 | next_t = scan_pp (); |
2f1a1aea | 1152 | if (! next_t) |
2677d2fb | 1153 | throw parse_error (_("unexpected end-of-file")); |
2f1a1aea | 1154 | |
2f1a1aea FCE |
1155 | last_t = next_t; |
1156 | // advance by zeroing next_t | |
1157 | next_t = 0; | |
1158 | return last_t; | |
1159 | } | |
1160 | ||
1161 | ||
1162 | const token* | |
b5477cd9 | 1163 | parser::peek () |
2f1a1aea FCE |
1164 | { |
1165 | if (! next_t) | |
b5477cd9 | 1166 | next_t = scan_pp (); |
2f1a1aea FCE |
1167 | |
1168 | // don't advance by zeroing next_t | |
1169 | last_t = next_t; | |
1170 | return next_t; | |
1171 | } | |
1172 | ||
1173 | ||
731a5359 MW |
1174 | void |
1175 | parser::swallow () | |
1176 | { | |
1177 | // can only swallow something last peeked or nexted token. | |
1178 | assert (last_t != 0); | |
1179 | delete last_t; | |
1180 | // advance by zeroing next_t | |
1181 | last_t = next_t = 0; | |
1182 | } | |
1183 | ||
1184 | ||
d7f3e0c5 GH |
1185 | static inline bool |
1186 | tok_is(token const * t, token_type tt, string const & expected) | |
1187 | { | |
1188 | return t && t->type == tt && t->content == expected; | |
1189 | } | |
1190 | ||
1191 | ||
731a5359 | 1192 | void |
d7f3e0c5 GH |
1193 | parser::expect_known (token_type tt, string const & expected) |
1194 | { | |
1195 | const token *t = next(); | |
57b73400 | 1196 | if (! (t && t->type == tt && t->content == expected)) |
ce0f6648 | 1197 | throw parse_error (_F("expected '%s'", expected.c_str())); |
731a5359 | 1198 | swallow (); // We are done with it, content was copied. |
d7f3e0c5 GH |
1199 | } |
1200 | ||
1201 | ||
731a5359 | 1202 | void |
d7f3e0c5 GH |
1203 | parser::expect_unknown (token_type tt, string & target) |
1204 | { | |
1205 | const token *t = next(); | |
1206 | if (!(t && t->type == tt)) | |
2677d2fb | 1207 | throw parse_error (_("expected ") + tt2str(tt)); |
d7f3e0c5 | 1208 | target = t->content; |
731a5359 | 1209 | swallow (); // We are done with it, content was copied. |
d7f3e0c5 GH |
1210 | } |
1211 | ||
1212 | ||
731a5359 | 1213 | void |
493ee224 DS |
1214 | parser::expect_unknown2 (token_type tt1, token_type tt2, string & target) |
1215 | { | |
1216 | const token *t = next(); | |
1217 | if (!(t && (t->type == tt1 || t->type == tt2))) | |
dd90d565 | 1218 | throw parse_error (_F("expected %s or %s", tt2str(tt1).c_str(), tt2str(tt2).c_str())); |
493ee224 | 1219 | target = t->content; |
731a5359 | 1220 | swallow (); // We are done with it, content was copied. |
493ee224 DS |
1221 | } |
1222 | ||
1223 | ||
731a5359 | 1224 | void |
d7f3e0c5 GH |
1225 | parser::expect_op (std::string const & expected) |
1226 | { | |
731a5359 | 1227 | expect_known (tok_operator, expected); |
d7f3e0c5 GH |
1228 | } |
1229 | ||
1230 | ||
731a5359 | 1231 | void |
d7f3e0c5 GH |
1232 | parser::expect_kw (std::string const & expected) |
1233 | { | |
731a5359 | 1234 | expect_known (tok_keyword, expected); |
d7f3e0c5 GH |
1235 | } |
1236 | ||
dff50e09 | 1237 | const token* |
731a5359 MW |
1238 | parser::expect_kw_token (std::string const & expected) |
1239 | { | |
1240 | const token *t = next(); | |
1241 | if (! (t && t->type == tok_keyword && t->content == expected)) | |
1242 | throw parse_error (_F("expected '%s'", expected.c_str())); | |
1243 | return t; | |
1244 | } | |
1245 | ||
1246 | void | |
e38723d2 | 1247 | parser::expect_number (int64_t & value) |
57b73400 | 1248 | { |
e38723d2 MH |
1249 | bool neg = false; |
1250 | const token *t = next(); | |
1251 | if (t->type == tok_operator && t->content == "-") | |
1252 | { | |
1253 | neg = true; | |
731a5359 | 1254 | swallow (); |
e38723d2 MH |
1255 | t = next (); |
1256 | } | |
1257 | if (!(t && t->type == tok_number)) | |
2677d2fb | 1258 | throw parse_error (_("expected number")); |
e38723d2 MH |
1259 | |
1260 | const char* startp = t->content.c_str (); | |
1261 | char* endp = (char*) startp; | |
1262 | ||
1263 | // NB: we allow controlled overflow from LLONG_MIN .. ULLONG_MAX | |
1264 | // Actually, this allows all the way from -ULLONG_MAX to ULLONG_MAX, | |
1265 | // since the lexer only gives us positive digit strings, but we'll | |
1266 | // limit it to LLONG_MIN when a '-' operator is fed into the literal. | |
1267 | errno = 0; | |
1268 | value = (int64_t) strtoull (startp, & endp, 0); | |
1269 | if (errno == ERANGE || errno == EINVAL || *endp != '\0' | |
1270 | || (neg && (unsigned long long) value > 9223372036854775808ULL) | |
1271 | || (unsigned long long) value > 18446744073709551615ULL | |
1272 | || value < -9223372036854775807LL-1) | |
2677d2fb | 1273 | throw parse_error (_("number invalid or out of range")); |
dff50e09 | 1274 | |
e38723d2 MH |
1275 | if (neg) |
1276 | value = -value; | |
1277 | ||
731a5359 | 1278 | swallow (); // We are done with it, content was parsed and copied into value. |
57b73400 GH |
1279 | } |
1280 | ||
d7f3e0c5 | 1281 | |
dff50e09 | 1282 | const token* |
50cc7cd5 | 1283 | parser::expect_ident_or_atword (std::string & target) |
d7f3e0c5 | 1284 | { |
06219d6f SM |
1285 | const token *t = next(); |
1286 | ||
1287 | // accept identifiers and operators beginning in '@': | |
1288 | if (!t || (t->type != tok_identifier | |
1289 | && (t->type != tok_operator || t->content[0] != '@'))) | |
1290 | // XXX currently this is only called from parse_hist_op_or_bare_name(), | |
1291 | // so the message is accurate, but keep an eye out in the future: | |
dd90d565 | 1292 | throw parse_error (_F("expected %s or statistical operation", tt2str(tok_identifier).c_str())); |
06219d6f SM |
1293 | |
1294 | target = t->content; | |
1295 | return t; | |
d7f3e0c5 GH |
1296 | } |
1297 | ||
1298 | ||
731a5359 | 1299 | void |
493ee224 DS |
1300 | parser::expect_ident_or_keyword (std::string & target) |
1301 | { | |
731a5359 | 1302 | expect_unknown2 (tok_identifier, tok_keyword, target); |
493ee224 DS |
1303 | } |
1304 | ||
1305 | ||
dff50e09 | 1306 | bool |
d7f3e0c5 GH |
1307 | parser::peek_op (std::string const & op) |
1308 | { | |
1309 | return tok_is (peek(), tok_operator, op); | |
1310 | } | |
1311 | ||
1312 | ||
dff50e09 | 1313 | bool |
d7f3e0c5 GH |
1314 | parser::peek_kw (std::string const & kw) |
1315 | { | |
1316 | return tok_is (peek(), tok_identifier, kw); | |
1317 | } | |
1318 | ||
1319 | ||
1320 | ||
66c7d4c1 | 1321 | lexer::lexer (istream& input, const string& in, systemtap_session& s): |
03ba36d9 SM |
1322 | ate_comment(false), ate_whitespace(false), saw_tokens(false), |
1323 | input_name (in), input_pointer (0), input_end (0), cursor_suspend_count(0), | |
1324 | cursor_suspend_line (1), cursor_suspend_column (1), cursor_line (1), | |
1325 | cursor_column (1), session(s), current_file (0) | |
eacb10ce | 1326 | { |
66c7d4c1 | 1327 | getline(input, input_contents, '\0'); |
2203b032 | 1328 | |
66c7d4c1 JS |
1329 | input_pointer = input_contents.data(); |
1330 | input_end = input_contents.data() + input_contents.size(); | |
1331 | ||
1332 | if (keywords.empty()) | |
1333 | { | |
3a7ec735 FCE |
1334 | // NB: adding new keywords is highly disruptive to the language, |
1335 | // in particular to existing scripts that could be suddenly | |
1336 | // broken. If done at all, it has to be s.compatible-sensitive, | |
1337 | // and broadly advertised. | |
66c7d4c1 JS |
1338 | keywords.insert("probe"); |
1339 | keywords.insert("global"); | |
1340 | keywords.insert("function"); | |
1341 | keywords.insert("if"); | |
1342 | keywords.insert("else"); | |
1343 | keywords.insert("for"); | |
1344 | keywords.insert("foreach"); | |
1345 | keywords.insert("in"); | |
1346 | keywords.insert("limit"); | |
1347 | keywords.insert("return"); | |
1348 | keywords.insert("delete"); | |
1349 | keywords.insert("while"); | |
1350 | keywords.insert("break"); | |
1351 | keywords.insert("continue"); | |
1352 | keywords.insert("next"); | |
1353 | keywords.insert("string"); | |
1354 | keywords.insert("long"); | |
f4fe2e93 FCE |
1355 | keywords.insert("try"); |
1356 | keywords.insert("catch"); | |
66c7d4c1 | 1357 | } |
2524d1fd SM |
1358 | |
1359 | if (atwords.empty()) | |
1360 | { | |
1361 | // NB: adding new @words is mildly disruptive to existing | |
1362 | // scripts that define macros with the same name, but not | |
1363 | // really. The user will merely receive a warning that they are | |
1364 | // redefining an existing operator. | |
1365 | atwords.insert("@cast"); | |
1366 | atwords.insert("@defined"); | |
1367 | atwords.insert("@entry"); | |
1368 | atwords.insert("@var"); | |
1369 | atwords.insert("@avg"); | |
1370 | atwords.insert("@count"); | |
1371 | atwords.insert("@sum"); | |
1372 | atwords.insert("@min"); | |
1373 | atwords.insert("@max"); | |
1374 | atwords.insert("@hist_linear"); | |
1375 | atwords.insert("@hist_log"); | |
1376 | } | |
eacb10ce | 1377 | } |
2f1a1aea | 1378 | |
66c7d4c1 | 1379 | set<string> lexer::keywords; |
2524d1fd | 1380 | set<string> lexer::atwords; |
66c7d4c1 | 1381 | |
1b1b4ceb RA |
1382 | void |
1383 | lexer::set_current_file (stapfile* f) | |
1384 | { | |
1385 | current_file = f; | |
2203b032 JS |
1386 | if (f) |
1387 | { | |
1388 | f->file_contents = input_contents; | |
1389 | f->name = input_name; | |
1390 | } | |
1b1b4ceb | 1391 | } |
bb2e3076 FCE |
1392 | |
1393 | int | |
1394 | lexer::input_peek (unsigned n) | |
1395 | { | |
66c7d4c1 JS |
1396 | if (input_pointer + n >= input_end) |
1397 | return -1; // EOF | |
1398 | return (unsigned char)*(input_pointer + n); | |
bb2e3076 FCE |
1399 | } |
1400 | ||
1401 | ||
dff50e09 | 1402 | int |
2f1a1aea FCE |
1403 | lexer::input_get () |
1404 | { | |
66c7d4c1 | 1405 | int c = input_peek(); |
bb2e3076 FCE |
1406 | if (c < 0) return c; // EOF |
1407 | ||
66c7d4c1 JS |
1408 | ++input_pointer; |
1409 | ||
3f99432c | 1410 | if (cursor_suspend_count) |
9300f661 JS |
1411 | { |
1412 | // Track effect of input_put: preserve previous cursor/line_column | |
1413 | // until all of its characters are consumed. | |
1414 | if (--cursor_suspend_count == 0) | |
1415 | { | |
1416 | cursor_line = cursor_suspend_line; | |
1417 | cursor_column = cursor_suspend_column; | |
1418 | } | |
1419 | } | |
3f99432c | 1420 | else |
2f1a1aea | 1421 | { |
3f99432c FCE |
1422 | // update source cursor |
1423 | if (c == '\n') | |
1424 | { | |
1425 | cursor_line ++; | |
1426 | cursor_column = 1; | |
1427 | } | |
1428 | else | |
1429 | cursor_column ++; | |
2f1a1aea | 1430 | } |
2f1a1aea | 1431 | |
eacb10ce | 1432 | // clog << "[" << (char)c << "]"; |
2f1a1aea FCE |
1433 | return c; |
1434 | } | |
1435 | ||
1436 | ||
3f99432c | 1437 | void |
9300f661 | 1438 | lexer::input_put (const string& chars, const token* t) |
3f99432c | 1439 | { |
66c7d4c1 JS |
1440 | size_t pos = input_pointer - input_contents.data(); |
1441 | // clog << "[put:" << chars << " @" << pos << "]"; | |
1442 | input_contents.insert (pos, chars); | |
eacb10ce | 1443 | cursor_suspend_count += chars.size(); |
9300f661 JS |
1444 | cursor_suspend_line = cursor_line; |
1445 | cursor_suspend_column = cursor_column; | |
1446 | cursor_line = t->location.line; | |
1447 | cursor_column = t->location.column; | |
66c7d4c1 JS |
1448 | input_pointer = input_contents.data() + pos; |
1449 | input_end = input_contents.data() + input_contents.size(); | |
3f99432c FCE |
1450 | } |
1451 | ||
1452 | ||
2f1a1aea | 1453 | token* |
b5477cd9 | 1454 | lexer::scan () |
2f1a1aea | 1455 | { |
fee28e5c | 1456 | ate_comment = false; // reset for each new token |
b5477cd9 | 1457 | ate_whitespace = false; // reset for each new token |
534aad8b SM |
1458 | |
1459 | // XXX be very sure to restore old_saw_tokens if we return without a token: | |
1460 | bool old_saw_tokens = saw_tokens; | |
1461 | saw_tokens = true; | |
1462 | ||
2f1a1aea | 1463 | token* n = new token; |
2203b032 | 1464 | n->location.file = current_file; |
534aad8b | 1465 | n->chain = NULL; // important safety dance |
2f1a1aea | 1466 | |
9300f661 JS |
1467 | skip: |
1468 | bool suspended = (cursor_suspend_count > 0); | |
2f1a1aea FCE |
1469 | n->location.line = cursor_line; |
1470 | n->location.column = cursor_column; | |
1471 | ||
1472 | int c = input_get(); | |
3f99432c | 1473 | // clog << "{" << (char)c << (char)c2 << "}"; |
2f1a1aea FCE |
1474 | if (c < 0) |
1475 | { | |
1476 | delete n; | |
534aad8b | 1477 | saw_tokens = old_saw_tokens; |
2f1a1aea FCE |
1478 | return 0; |
1479 | } | |
1480 | ||
1481 | if (isspace (c)) | |
b5477cd9 SM |
1482 | { |
1483 | ate_whitespace = true; | |
1484 | goto skip; | |
1485 | } | |
2f1a1aea | 1486 | |
66c7d4c1 JS |
1487 | int c2 = input_peek (); |
1488 | ||
3f99432c FCE |
1489 | // Paste command line arguments as character streams into |
1490 | // the beginning of a token. $1..$999 go through as raw | |
1491 | // characters; @1..@999 are quoted/escaped as strings. | |
1492 | // $# and @# expand to the number of arguments, similarly | |
1493 | // raw or quoted. | |
9300f661 | 1494 | if ((c == '$' || c == '@') && (c2 == '#')) |
3f99432c | 1495 | { |
9300f661 JS |
1496 | n->content.push_back (c); |
1497 | n->content.push_back (c2); | |
3f99432c | 1498 | input_get(); // swallow '#' |
9300f661 | 1499 | if (suspended) |
16fc963f SM |
1500 | { |
1501 | n->make_junk(_("invalid nested substitution of command line arguments")); | |
1502 | return n; | |
1503 | } | |
9300f661 JS |
1504 | size_t num_args = session.args.size (); |
1505 | input_put ((c == '$') ? lex_cast (num_args) : lex_cast_qstring (num_args), n); | |
1506 | n->content.clear(); | |
1507 | goto skip; | |
3f99432c | 1508 | } |
9300f661 | 1509 | else if ((c == '$' || c == '@') && (isdigit (c2))) |
3f99432c | 1510 | { |
9300f661 | 1511 | n->content.push_back (c); |
3f99432c FCE |
1512 | unsigned idx = 0; |
1513 | do | |
1514 | { | |
1515 | input_get (); | |
1516 | idx = (idx * 10) + (c2 - '0'); | |
9300f661 | 1517 | n->content.push_back (c2); |
3f99432c FCE |
1518 | c2 = input_peek (); |
1519 | } while (c2 > 0 && | |
dff50e09 | 1520 | isdigit (c2) && |
3f99432c | 1521 | idx <= session.args.size()); // prevent overflow |
16fc963f SM |
1522 | if (suspended) |
1523 | { | |
1524 | n->make_junk(_("invalid nested substitution of command line arguments")); | |
1525 | return n; | |
1526 | } | |
3f99432c FCE |
1527 | if (idx == 0 || |
1528 | idx-1 >= session.args.size()) | |
16fc963f SM |
1529 | { |
1530 | n->make_junk(_F("command line argument index %lu out of range [1-%lu]", | |
1531 | (unsigned long) idx, (unsigned long) session.args.size())); | |
1532 | return n; | |
1533 | } | |
9300f661 JS |
1534 | const string& arg = session.args[idx-1]; |
1535 | input_put ((c == '$') ? arg : lex_cast_qstring (arg), n); | |
1536 | n->content.clear(); | |
1537 | goto skip; | |
3f99432c FCE |
1538 | } |
1539 | ||
b5477cd9 | 1540 | else if (isalpha (c) || c == '$' || c == '@' || c == '_') |
2f1a1aea FCE |
1541 | { |
1542 | n->type = tok_identifier; | |
1543 | n->content = (char) c; | |
b5477cd9 | 1544 | while (isalnum (c2) || c2 == '_' || c2 == '$') |
2f1a1aea | 1545 | { |
3f99432c FCE |
1546 | input_get (); |
1547 | n->content.push_back (c2); | |
1548 | c2 = input_peek (); | |
6e213f58 | 1549 | } |
213bee8f | 1550 | |
66c7d4c1 | 1551 | if (keywords.count(n->content)) |
3f99432c | 1552 | n->type = tok_keyword; |
06219d6f | 1553 | else if (n->content[0] == '@') |
dd90d565 | 1554 | // makes it easier to detect illegal use of @words: |
06219d6f | 1555 | n->type = tok_operator; |
dff50e09 | 1556 | |
2f1a1aea FCE |
1557 | return n; |
1558 | } | |
1559 | ||
3a20432b | 1560 | else if (isdigit (c)) // positive literal |
2f1a1aea | 1561 | { |
2f1a1aea | 1562 | n->type = tok_number; |
9c0c0e46 FCE |
1563 | n->content = (char) c; |
1564 | ||
66c7d4c1 | 1565 | while (isalnum (c2)) |
2f1a1aea | 1566 | { |
9c0c0e46 FCE |
1567 | // NB: isalnum is very permissive. We rely on strtol, called in |
1568 | // parser::parse_literal below, to confirm that the number string | |
1569 | // is correctly formatted and in range. | |
1570 | ||
66c7d4c1 JS |
1571 | input_get (); |
1572 | n->content.push_back (c2); | |
1573 | c2 = input_peek (); | |
2f1a1aea FCE |
1574 | } |
1575 | return n; | |
1576 | } | |
1577 | ||
1578 | else if (c == '\"') | |
1579 | { | |
1580 | n->type = tok_string; | |
1581 | while (1) | |
1582 | { | |
1583 | c = input_get (); | |
1584 | ||
3f99432c | 1585 | if (c < 0 || c == '\n') |
2f1a1aea | 1586 | { |
16fc963f SM |
1587 | n->make_junk(_("Could not find matching closing quote")); |
1588 | return n; | |
2f1a1aea FCE |
1589 | } |
1590 | if (c == '\"') // closing double-quotes | |
1591 | break; | |
3f99432c | 1592 | else if (c == '\\') // see also input_put |
dff50e09 | 1593 | { |
7d46afb8 GH |
1594 | c = input_get (); |
1595 | switch (c) | |
1596 | { | |
1597 | case 'a': | |
1598 | case 'b': | |
1599 | case 't': | |
1600 | case 'n': | |
1601 | case 'v': | |
1602 | case 'f': | |
1603 | case 'r': | |
f03954fd | 1604 | case '0' ... '7': // NB: need only match the first digit |
7d46afb8 | 1605 | case '\\': |
7d46afb8 | 1606 | // Pass these escapes through to the string value |
dff50e09 | 1607 | // being parsed; it will be emitted into a C literal. |
7d46afb8 GH |
1608 | |
1609 | n->content.push_back('\\'); | |
1610 | ||
3f99432c | 1611 | // fall through |
7d46afb8 | 1612 | default: |
7d46afb8 GH |
1613 | n->content.push_back(c); |
1614 | break; | |
1615 | } | |
2f1a1aea FCE |
1616 | } |
1617 | else | |
1618 | n->content.push_back(c); | |
1619 | } | |
1620 | return n; | |
1621 | } | |
1622 | ||
1623 | else if (ispunct (c)) | |
1624 | { | |
bb2e3076 | 1625 | int c3 = input_peek (1); |
2f1a1aea | 1626 | |
3a20432b FCE |
1627 | // NB: if we were to recognize negative numeric literals here, |
1628 | // we'd introduce another grammar ambiguity: | |
1629 | // 1-1 would be parsed as tok_number(1) and tok_number(-1) | |
1630 | // instead of tok_number(1) tok_operator('-') tok_number(1) | |
1631 | ||
66c7d4c1 | 1632 | if (c == '#') // shell comment |
2f1a1aea FCE |
1633 | { |
1634 | unsigned this_line = cursor_line; | |
bb2e3076 FCE |
1635 | do { c = input_get (); } |
1636 | while (c >= 0 && cursor_line == this_line); | |
fee28e5c | 1637 | ate_comment = true; |
b5477cd9 | 1638 | ate_whitespace = true; |
2f1a1aea FCE |
1639 | goto skip; |
1640 | } | |
66c7d4c1 | 1641 | else if ((c == '/' && c2 == '/')) // C++ comment |
63a7c90e FCE |
1642 | { |
1643 | unsigned this_line = cursor_line; | |
bb2e3076 FCE |
1644 | do { c = input_get (); } |
1645 | while (c >= 0 && cursor_line == this_line); | |
fee28e5c | 1646 | ate_comment = true; |
b5477cd9 | 1647 | ate_whitespace = true; |
63a7c90e FCE |
1648 | goto skip; |
1649 | } | |
1650 | else if (c == '/' && c2 == '*') // C comment | |
1651 | { | |
66c7d4c1 JS |
1652 | (void) input_get (); // swallow '*' already in c2 |
1653 | c = input_get (); | |
63a7c90e | 1654 | c2 = input_get (); |
bb2e3076 | 1655 | while (c2 >= 0) |
63a7c90e | 1656 | { |
66c7d4c1 JS |
1657 | if (c == '*' && c2 == '/') |
1658 | break; | |
63a7c90e FCE |
1659 | c = c2; |
1660 | c2 = input_get (); | |
63a7c90e | 1661 | } |
fee28e5c | 1662 | ate_comment = true; |
b5477cd9 | 1663 | ate_whitespace = true; |
bb2e3076 | 1664 | goto skip; |
63a7c90e | 1665 | } |
54dfabe9 FCE |
1666 | else if (c == '%' && c2 == '{') // embedded code |
1667 | { | |
1668 | n->type = tok_embedded; | |
1669 | (void) input_get (); // swallow '{' already in c2 | |
66c7d4c1 JS |
1670 | c = input_get (); |
1671 | c2 = input_get (); | |
1672 | while (c2 >= 0) | |
54dfabe9 | 1673 | { |
66c7d4c1 JS |
1674 | if (c == '%' && c2 == '}') |
1675 | return n; | |
54dfabe9 | 1676 | n->content += c; |
66c7d4c1 JS |
1677 | c = c2; |
1678 | c2 = input_get (); | |
54dfabe9 | 1679 | } |
72cdb9cd | 1680 | |
16fc963f SM |
1681 | n->make_junk(_("Could not find matching '%}' to close embedded function block")); |
1682 | return n; | |
54dfabe9 | 1683 | } |
2f1a1aea | 1684 | |
bb2e3076 FCE |
1685 | // We're committed to recognizing at least the first character |
1686 | // as an operator. | |
2f1a1aea | 1687 | n->type = tok_operator; |
66c7d4c1 | 1688 | n->content = c; |
2f1a1aea | 1689 | |
bb2e3076 | 1690 | // match all valid operators, in decreasing size order |
66c7d4c1 JS |
1691 | if ((c == '<' && c2 == '<' && c3 == '<') || |
1692 | (c == '<' && c2 == '<' && c3 == '=') || | |
1693 | (c == '>' && c2 == '>' && c3 == '=')) | |
82919855 | 1694 | { |
66c7d4c1 JS |
1695 | n->content += c2; |
1696 | n->content += c3; | |
bb2e3076 FCE |
1697 | input_get (); input_get (); // swallow other two characters |
1698 | } | |
66c7d4c1 JS |
1699 | else if ((c == '=' && c2 == '=') || |
1700 | (c == '!' && c2 == '=') || | |
1701 | (c == '<' && c2 == '=') || | |
1702 | (c == '>' && c2 == '=') || | |
93daaca8 SM |
1703 | (c == '=' && c2 == '~') || |
1704 | (c == '!' && c2 == '~') || | |
66c7d4c1 JS |
1705 | (c == '+' && c2 == '=') || |
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 == '>') || | |
177a8ead | 1721 | // preprocessor tokens |
66c7d4c1 JS |
1722 | (c == '%' && c2 == '(') || |
1723 | (c == '%' && c2 == '?') || | |
1724 | (c == '%' && c2 == ':') || | |
1725 | (c == '%' && c2 == ')')) | |
bb2e3076 | 1726 | { |
66c7d4c1 | 1727 | n->content += c2; |
bb2e3076 | 1728 | input_get (); // swallow other character |
dff50e09 | 1729 | } |
2f1a1aea FCE |
1730 | |
1731 | return n; | |
1732 | } | |
1733 | ||
1734 | else | |
1735 | { | |
1736 | n->type = tok_junk; | |
e3795795 FCE |
1737 | ostringstream s; |
1738 | s << "\\x" << hex << setw(2) << setfill('0') << c; | |
1739 | n->content = s.str(); | |
16fc963f | 1740 | n->msg = ""; // signal parser to emit "expected X, found junk" type error |
2f1a1aea FCE |
1741 | return n; |
1742 | } | |
1743 | } | |
1744 | ||
16fc963f SM |
1745 | // ------------------------------------------------------------------------ |
1746 | ||
1747 | void | |
1748 | token::make_junk (const string new_msg) | |
1749 | { | |
1750 | type = tok_junk; | |
1751 | msg = new_msg; | |
1752 | } | |
2f1a1aea FCE |
1753 | |
1754 | // ------------------------------------------------------------------------ | |
1755 | ||
1756 | stapfile* | |
1757 | parser::parse () | |
1758 | { | |
1759 | stapfile* f = new stapfile; | |
1b1b4ceb | 1760 | input.set_current_file (f); |
56099f08 FCE |
1761 | |
1762 | bool empty = true; | |
1763 | ||
2f1a1aea FCE |
1764 | while (1) |
1765 | { | |
1766 | try | |
1767 | { | |
a07a2c28 | 1768 | systemtap_v_seen = 0; |
2f1a1aea | 1769 | const token* t = peek (); |
534aad8b | 1770 | if (! t) // nice clean EOF, modulo any preprocessing that occurred |
2f1a1aea FCE |
1771 | break; |
1772 | ||
56099f08 | 1773 | empty = false; |
6e213f58 DS |
1774 | if (t->type == tok_keyword && t->content == "probe") |
1775 | { | |
1776 | context = con_probe; | |
1777 | parse_probe (f->probes, f->aliases); | |
1778 | } | |
1779 | else if (t->type == tok_keyword && t->content == "global") | |
1780 | { | |
1781 | context = con_global; | |
4b5f3e45 | 1782 | parse_global (f->globals, f->probes); |
6e213f58 DS |
1783 | } |
1784 | else if (t->type == tok_keyword && t->content == "function") | |
1785 | { | |
1786 | context = con_function; | |
1787 | parse_functiondecl (f->functions); | |
1788 | } | |
54dfabe9 | 1789 | else if (t->type == tok_embedded) |
6e213f58 DS |
1790 | { |
1791 | context = con_embedded; | |
1792 | f->embeds.push_back (parse_embeddedcode ()); | |
1793 | } | |
2f1a1aea | 1794 | else |
6e213f58 DS |
1795 | { |
1796 | context = con_unknown; | |
2677d2fb | 1797 | throw parse_error (_("expected 'probe', 'global', 'function', or '%{'")); |
6e213f58 | 1798 | } |
2f1a1aea FCE |
1799 | } |
1800 | catch (parse_error& pe) | |
1801 | { | |
1802 | print_error (pe); | |
16fc963f SM |
1803 | |
1804 | // XXX: do we want tok_junk to be able to force skip_some behaviour? | |
cd7116b8 | 1805 | if (pe.skip_some) // for recovery |
46954f1d FCE |
1806 | // Quietly swallow all tokens until the next keyword we can start parsing from. |
1807 | while (1) | |
1808 | try | |
1809 | { | |
cd7116b8 FCE |
1810 | { |
1811 | const token* t = peek (); | |
1812 | if (! t) | |
1813 | break; | |
46954f1d FCE |
1814 | if (t->type == tok_keyword && t->content == "probe") break; |
1815 | else if (t->type == tok_keyword && t->content == "global") break; | |
1816 | else if (t->type == tok_keyword && t->content == "function") break; | |
1817 | else if (t->type == tok_embedded) break; | |
731a5359 | 1818 | swallow (); // swallow it |
cd7116b8 | 1819 | } |
46954f1d FCE |
1820 | } |
1821 | catch (parse_error& pe2) | |
1822 | { | |
1823 | // parse error during recovery ... ugh | |
1824 | print_error (pe2); | |
1825 | } | |
177a8ead | 1826 | } |
2f1a1aea FCE |
1827 | } |
1828 | ||
56099f08 FCE |
1829 | if (empty) |
1830 | { | |
534aad8b SM |
1831 | // vary message depending on whether file was *actually* empty: |
1832 | cerr << (input.saw_tokens | |
1833 | ? _F("Input file '%s' is empty after preprocessing.", input_name.c_str()) | |
4cd32d8c | 1834 | : _F("Input file '%s' is empty.", input_name.c_str())) |
534aad8b | 1835 | << endl; |
56099f08 | 1836 | delete f; |
2203b032 | 1837 | f = 0; |
56099f08 FCE |
1838 | } |
1839 | else if (num_errors > 0) | |
2f1a1aea | 1840 | { |
52c2652f | 1841 | cerr << _NF("%d parse error.", "%d parse errors.", num_errors, num_errors) << endl; |
2f1a1aea | 1842 | delete f; |
2203b032 | 1843 | f = 0; |
2f1a1aea | 1844 | } |
dff50e09 | 1845 | |
2203b032 | 1846 | input.set_current_file(0); |
2f1a1aea FCE |
1847 | return f; |
1848 | } | |
1849 | ||
1850 | ||
20c6c071 | 1851 | void |
54dfabe9 FCE |
1852 | parser::parse_probe (std::vector<probe *> & probe_ret, |
1853 | std::vector<probe_alias *> & alias_ret) | |
2f1a1aea | 1854 | { |
82919855 | 1855 | const token* t0 = next (); |
6e213f58 | 1856 | if (! (t0->type == tok_keyword && t0->content == "probe")) |
2677d2fb | 1857 | throw parse_error (_("expected 'probe'")); |
82919855 | 1858 | |
20c6c071 GH |
1859 | vector<probe_point *> aliases; |
1860 | vector<probe_point *> locations; | |
1861 | ||
1862 | bool equals_ok = true; | |
82919855 | 1863 | |
97266278 LG |
1864 | int epilogue_alias = 0; |
1865 | ||
2f1a1aea FCE |
1866 | while (1) |
1867 | { | |
b4ceace2 | 1868 | probe_point * pp = parse_probe_point (); |
dff50e09 | 1869 | |
b4ceace2 | 1870 | const token* t = peek (); |
dff50e09 | 1871 | if (equals_ok && t |
b4ceace2 FCE |
1872 | && t->type == tok_operator && t->content == "=") |
1873 | { | |
1ad820e3 | 1874 | if (pp->optional || pp->sufficient) |
2677d2fb | 1875 | throw parse_error (_("probe point alias name cannot be optional nor sufficient"), pp->components.front()->tok); |
b4ceace2 | 1876 | aliases.push_back(pp); |
731a5359 | 1877 | swallow (); |
b4ceace2 FCE |
1878 | continue; |
1879 | } | |
dff50e09 | 1880 | else if (equals_ok && t |
97266278 LG |
1881 | && t->type == tok_operator && t->content == "+=") |
1882 | { | |
1ad820e3 | 1883 | if (pp->optional || pp->sufficient) |
2677d2fb | 1884 | throw parse_error (_("probe point alias name cannot be optional nor sufficient"), pp->components.front()->tok); |
97266278 LG |
1885 | aliases.push_back(pp); |
1886 | epilogue_alias = 1; | |
731a5359 | 1887 | swallow (); |
97266278 LG |
1888 | continue; |
1889 | } | |
b4ceace2 FCE |
1890 | else if (t && t->type == tok_operator && t->content == ",") |
1891 | { | |
1892 | locations.push_back(pp); | |
1893 | equals_ok = false; | |
731a5359 | 1894 | swallow (); |
b4ceace2 FCE |
1895 | continue; |
1896 | } | |
1897 | else if (t && t->type == tok_operator && t->content == "{") | |
1898 | { | |
1899 | locations.push_back(pp); | |
1900 | break; | |
1901 | } | |
2f1a1aea | 1902 | else |
2677d2fb | 1903 | throw parse_error (_("expected probe point specifier")); |
2f1a1aea | 1904 | } |
20c6c071 | 1905 | |
20c6c071 GH |
1906 | if (aliases.empty()) |
1907 | { | |
54dfabe9 FCE |
1908 | probe* p = new probe; |
1909 | p->tok = t0; | |
1910 | p->locations = locations; | |
1911 | p->body = parse_stmt_block (); | |
37ebca01 | 1912 | p->privileged = privileged; |
a07a2c28 | 1913 | p->systemtap_v_conditional = systemtap_v_seen; |
54dfabe9 | 1914 | probe_ret.push_back (p); |
20c6c071 GH |
1915 | } |
1916 | else | |
1917 | { | |
54dfabe9 | 1918 | probe_alias* p = new probe_alias (aliases); |
97266278 LG |
1919 | if(epilogue_alias) |
1920 | p->epilogue_style = true; | |
1921 | else | |
1922 | p->epilogue_style = false; | |
54dfabe9 FCE |
1923 | p->tok = t0; |
1924 | p->locations = locations; | |
1925 | p->body = parse_stmt_block (); | |
37ebca01 | 1926 | p->privileged = privileged; |
a07a2c28 | 1927 | p->systemtap_v_conditional = systemtap_v_seen; |
54dfabe9 | 1928 | alias_ret.push_back (p); |
20c6c071 | 1929 | } |
54dfabe9 | 1930 | } |
20c6c071 | 1931 | |
54dfabe9 FCE |
1932 | |
1933 | embeddedcode* | |
1934 | parser::parse_embeddedcode () | |
1935 | { | |
1936 | embeddedcode* e = new embeddedcode; | |
1937 | const token* t = next (); | |
1938 | if (t->type != tok_embedded) | |
2677d2fb | 1939 | throw parse_error (_("expected '%{'")); |
24cb178f FCE |
1940 | |
1941 | if (! privileged) | |
efb02738 | 1942 | throw parse_error (_("embedded code in unprivileged script; need stap -g"), |
cd7116b8 | 1943 | false /* don't skip tokens for parse resumption */); |
54dfabe9 FCE |
1944 | |
1945 | e->tok = t; | |
1946 | e->code = t->content; | |
1947 | return e; | |
2f1a1aea FCE |
1948 | } |
1949 | ||
1950 | ||
1951 | block* | |
56099f08 | 1952 | parser::parse_stmt_block () |
2f1a1aea FCE |
1953 | { |
1954 | block* pb = new block; | |
1955 | ||
56099f08 FCE |
1956 | const token* t = next (); |
1957 | if (! (t->type == tok_operator && t->content == "{")) | |
2677d2fb | 1958 | throw parse_error (_("expected '{'")); |
56099f08 FCE |
1959 | |
1960 | pb->tok = t; | |
2b066ec1 | 1961 | |
2f1a1aea FCE |
1962 | while (1) |
1963 | { | |
46954f1d FCE |
1964 | t = peek (); |
1965 | if (t && t->type == tok_operator && t->content == "}") | |
1966 | { | |
731a5359 | 1967 | swallow (); |
46954f1d FCE |
1968 | break; |
1969 | } | |
1970 | pb->statements.push_back (parse_statement ()); | |
2f1a1aea FCE |
1971 | } |
1972 | ||
1973 | return pb; | |
1974 | } | |
1975 | ||
1976 | ||
f4fe2e93 FCE |
1977 | try_block* |
1978 | parser::parse_try_block () | |
1979 | { | |
1980 | try_block* pb = new try_block; | |
1981 | ||
731a5359 | 1982 | pb->tok = expect_kw_token ("try"); |
f4fe2e93 FCE |
1983 | pb->try_block = parse_stmt_block(); |
1984 | expect_kw ("catch"); | |
1985 | ||
1986 | const token* t = peek (); | |
3819d181 | 1987 | if (t != NULL && t->type == tok_operator && t->content == "(") |
f4fe2e93 | 1988 | { |
731a5359 | 1989 | swallow (); // swallow the '(' |
f4fe2e93 FCE |
1990 | |
1991 | t = next(); | |
1992 | if (! (t->type == tok_identifier)) | |
2677d2fb | 1993 | throw parse_error (_("expected identifier")); |
f4fe2e93 FCE |
1994 | symbol* sym = new symbol; |
1995 | sym->tok = t; | |
1996 | sym->name = t->content; | |
1997 | pb->catch_error_var = sym; | |
1998 | ||
1999 | expect_op (")"); | |
2000 | } | |
2001 | else | |
2002 | pb->catch_error_var = 0; | |
2003 | ||
2004 | pb->catch_block = parse_stmt_block(); | |
2005 | ||
2006 | return pb; | |
2007 | } | |
2008 | ||
2009 | ||
2010 | ||
2f1a1aea FCE |
2011 | statement* |
2012 | parser::parse_statement () | |
2013 | { | |
40b71c47 | 2014 | statement *ret; |
2f1a1aea FCE |
2015 | const token* t = peek (); |
2016 | if (t && t->type == tok_operator && t->content == ";") | |
f946b10f | 2017 | return new null_statement (next ()); |
dff50e09 | 2018 | else if (t && t->type == tok_operator && t->content == "{") |
40b71c47 | 2019 | return parse_stmt_block (); // Don't squash semicolons. |
f4fe2e93 FCE |
2020 | else if (t && t->type == tok_keyword && t->content == "try") |
2021 | return parse_try_block (); // Don't squash semicolons. | |
6e213f58 | 2022 | else if (t && t->type == tok_keyword && t->content == "if") |
40b71c47 | 2023 | return parse_if_statement (); // Don't squash semicolons. |
6e213f58 | 2024 | else if (t && t->type == tok_keyword && t->content == "for") |
40b71c47 | 2025 | return parse_for_loop (); // Don't squash semicolons. |
6e213f58 | 2026 | else if (t && t->type == tok_keyword && t->content == "foreach") |
40b71c47 MW |
2027 | return parse_foreach_loop (); // Don't squash semicolons. |
2028 | else if (t && t->type == tok_keyword && t->content == "while") | |
2029 | return parse_while_loop (); // Don't squash semicolons. | |
6e213f58 | 2030 | else if (t && t->type == tok_keyword && t->content == "return") |
40b71c47 | 2031 | ret = parse_return_statement (); |
6e213f58 | 2032 | else if (t && t->type == tok_keyword && t->content == "delete") |
40b71c47 | 2033 | ret = parse_delete_statement (); |
6e213f58 | 2034 | else if (t && t->type == tok_keyword && t->content == "break") |
40b71c47 | 2035 | ret = parse_break_statement (); |
6e213f58 | 2036 | else if (t && t->type == tok_keyword && t->content == "continue") |
40b71c47 | 2037 | ret = parse_continue_statement (); |
6e213f58 | 2038 | else if (t && t->type == tok_keyword && t->content == "next") |
40b71c47 | 2039 | ret = parse_next_statement (); |
2f1a1aea FCE |
2040 | else if (t && (t->type == tok_operator || // expressions are flexible |
2041 | t->type == tok_identifier || | |
2042 | t->type == tok_number || | |
7d902887 FCE |
2043 | t->type == tok_string || |
2044 | t->type == tok_embedded )) | |
40b71c47 | 2045 | ret = parse_expr_statement (); |
54dfabe9 | 2046 | // XXX: consider generally accepting tok_embedded here too |
2f1a1aea | 2047 | else |
2677d2fb | 2048 | throw parse_error (_("expected statement")); |
40b71c47 MW |
2049 | |
2050 | // Squash "empty" trailing colons after any "non-block-like" statement. | |
2051 | t = peek (); | |
2052 | if (t && t->type == tok_operator && t->content == ";") | |
2053 | { | |
731a5359 | 2054 | swallow (); // Silently eat trailing ; after statement |
40b71c47 MW |
2055 | } |
2056 | ||
2057 | return ret; | |
2f1a1aea FCE |
2058 | } |
2059 | ||
2060 | ||
56099f08 | 2061 | void |
78f6bba6 | 2062 | parser::parse_global (vector <vardecl*>& globals, vector<probe*>&) |
2f1a1aea | 2063 | { |
82919855 | 2064 | const token* t0 = next (); |
6e213f58 | 2065 | if (! (t0->type == tok_keyword && t0->content == "global")) |
2677d2fb | 2066 | throw parse_error (_("expected 'global'")); |
731a5359 | 2067 | swallow (); |
82919855 | 2068 | |
56099f08 FCE |
2069 | while (1) |
2070 | { | |
2071 | const token* t = next (); | |
2072 | if (! (t->type == tok_identifier)) | |
2677d2fb | 2073 | throw parse_error (_("expected identifier")); |
56099f08 | 2074 | |
2b066ec1 FCE |
2075 | for (unsigned i=0; i<globals.size(); i++) |
2076 | if (globals[i]->name == t->content) | |
2677d2fb | 2077 | throw parse_error (_("duplicate global name")); |
dff50e09 | 2078 | |
24cb178f FCE |
2079 | vardecl* d = new vardecl; |
2080 | d->name = t->content; | |
2081 | d->tok = t; | |
a07a2c28 | 2082 | d->systemtap_v_conditional = systemtap_v_seen; |
24cb178f | 2083 | globals.push_back (d); |
56099f08 | 2084 | |
82919855 | 2085 | t = peek (); |
ef474d24 | 2086 | |
74e6cc92 CM |
2087 | if(t && t->type == tok_operator && t->content == "%") //wrapping |
2088 | { | |
2089 | d->wrap = true; | |
731a5359 | 2090 | swallow (); |
74e6cc92 CM |
2091 | t = peek(); |
2092 | } | |
2093 | ||
ef474d24 JS |
2094 | if (t && t->type == tok_operator && t->content == "[") // array size |
2095 | { | |
2096 | int64_t size; | |
731a5359 | 2097 | swallow (); |
ef474d24 JS |
2098 | expect_number(size); |
2099 | if (size <= 0 || size > 1000000) // arbitrary max | |
2677d2fb | 2100 | throw parse_error(_("array size out of range")); |
ef474d24 JS |
2101 | d->maxsize = (int)size; |
2102 | expect_known(tok_operator, "]"); | |
2103 | t = peek (); | |
2104 | } | |
2105 | ||
4b5f3e45 | 2106 | if (t && t->type == tok_operator && t->content == "=") // initialization |
ef474d24 JS |
2107 | { |
2108 | if (!d->compatible_arity(0)) | |
2677d2fb | 2109 | throw parse_error(_("only scalar globals can be initialized")); |
58701b78 | 2110 | d->set_arity(0, t); |
731a5359 | 2111 | next (); // Don't swallow, set_arity() used the peeked token. |
ef474d24 JS |
2112 | d->init = parse_literal (); |
2113 | d->type = d->init->type; | |
2114 | t = peek (); | |
2115 | } | |
4b5f3e45 | 2116 | |
c3799d72 | 2117 | if (t && t->type == tok_operator && t->content == ";") // termination |
950da622 | 2118 | { |
731a5359 | 2119 | swallow (); |
950da622 MW |
2120 | break; |
2121 | } | |
c3799d72 | 2122 | |
4b5f3e45 | 2123 | if (t && t->type == tok_operator && t->content == ",") // next global |
82919855 | 2124 | { |
731a5359 | 2125 | swallow (); |
82919855 FCE |
2126 | continue; |
2127 | } | |
56099f08 | 2128 | else |
82919855 | 2129 | break; |
56099f08 FCE |
2130 | } |
2131 | } | |
2132 | ||
2133 | ||
24cb178f FCE |
2134 | void |
2135 | parser::parse_functiondecl (std::vector<functiondecl*>& functions) | |
56099f08 | 2136 | { |
82919855 | 2137 | const token* t = next (); |
6e213f58 | 2138 | if (! (t->type == tok_keyword && t->content == "function")) |
2677d2fb | 2139 | throw parse_error (_("expected 'function'")); |
731a5359 | 2140 | swallow (); |
56099f08 | 2141 | |
82919855 | 2142 | t = next (); |
6e213f58 DS |
2143 | if (! (t->type == tok_identifier) |
2144 | && ! (t->type == tok_keyword | |
2145 | && (t->content == "string" || t->content == "long"))) | |
2677d2fb | 2146 | throw parse_error (_("expected identifier")); |
24cb178f FCE |
2147 | |
2148 | for (unsigned i=0; i<functions.size(); i++) | |
2149 | if (functions[i]->name == t->content) | |
2677d2fb | 2150 | throw parse_error (_("duplicate function name")); |
24cb178f FCE |
2151 | |
2152 | functiondecl *fd = new functiondecl (); | |
56099f08 FCE |
2153 | fd->name = t->content; |
2154 | fd->tok = t; | |
2155 | ||
2156 | t = next (); | |
6a505121 FCE |
2157 | if (t->type == tok_operator && t->content == ":") |
2158 | { | |
731a5359 | 2159 | swallow (); |
6a505121 | 2160 | t = next (); |
6e213f58 | 2161 | if (t->type == tok_keyword && t->content == "string") |
6a505121 | 2162 | fd->type = pe_string; |
6e213f58 | 2163 | else if (t->type == tok_keyword && t->content == "long") |
6a505121 | 2164 | fd->type = pe_long; |
2677d2fb | 2165 | else throw parse_error (_("expected 'string' or 'long'")); |
731a5359 | 2166 | swallow (); |
6a505121 FCE |
2167 | |
2168 | t = next (); | |
2169 | } | |
2170 | ||
56099f08 | 2171 | if (! (t->type == tok_operator && t->content == "(")) |
2677d2fb | 2172 | throw parse_error (_("expected '('")); |
731a5359 | 2173 | swallow (); |
56099f08 FCE |
2174 | |
2175 | while (1) | |
2176 | { | |
2177 | t = next (); | |
2178 | ||
100a540e | 2179 | // permit zero-argument functions |
56099f08 | 2180 | if (t->type == tok_operator && t->content == ")") |
731a5359 MW |
2181 | { |
2182 | swallow (); | |
2183 | break; | |
2184 | } | |
56099f08 | 2185 | else if (! (t->type == tok_identifier)) |
2677d2fb | 2186 | throw parse_error (_("expected identifier")); |
56099f08 FCE |
2187 | vardecl* vd = new vardecl; |
2188 | vd->name = t->content; | |
2189 | vd->tok = t; | |
2190 | fd->formal_args.push_back (vd); | |
a07a2c28 | 2191 | fd->systemtap_v_conditional = systemtap_v_seen; |
56099f08 FCE |
2192 | |
2193 | t = next (); | |
6a505121 FCE |
2194 | if (t->type == tok_operator && t->content == ":") |
2195 | { | |
731a5359 | 2196 | swallow (); |
6a505121 | 2197 | t = next (); |
6e213f58 | 2198 | if (t->type == tok_keyword && t->content == "string") |
6a505121 | 2199 | vd->type = pe_string; |
6e213f58 | 2200 | else if (t->type == tok_keyword && t->content == "long") |
6a505121 | 2201 | vd->type = pe_long; |
2677d2fb | 2202 | else throw parse_error (_("expected 'string' or 'long'")); |
731a5359 | 2203 | swallow (); |
6a505121 FCE |
2204 | t = next (); |
2205 | } | |
56099f08 | 2206 | if (t->type == tok_operator && t->content == ")") |
731a5359 MW |
2207 | { |
2208 | swallow (); | |
2209 | break; | |
2210 | } | |
56099f08 | 2211 | if (t->type == tok_operator && t->content == ",") |
731a5359 MW |
2212 | { |
2213 | swallow (); | |
2214 | continue; | |
2215 | } | |
56099f08 | 2216 | else |
2677d2fb | 2217 | throw parse_error (_("expected ',' or ')'")); |
56099f08 FCE |
2218 | } |
2219 | ||
54dfabe9 FCE |
2220 | t = peek (); |
2221 | if (t && t->type == tok_embedded) | |
2222 | fd->body = parse_embeddedcode (); | |
2223 | else | |
2224 | fd->body = parse_stmt_block (); | |
24cb178f FCE |
2225 | |
2226 | functions.push_back (fd); | |
2f1a1aea FCE |
2227 | } |
2228 | ||
2229 | ||
9c0c0e46 FCE |
2230 | probe_point* |
2231 | parser::parse_probe_point () | |
2f1a1aea | 2232 | { |
9c0c0e46 | 2233 | probe_point* pl = new probe_point; |
2f1a1aea | 2234 | |
9c0c0e46 | 2235 | while (1) |
2f1a1aea | 2236 | { |
b5477cd9 | 2237 | const token* t = next (); |
6e213f58 DS |
2238 | if (! (t->type == tok_identifier |
2239 | // we must allow ".return" and ".function", which are keywords | |
b5477cd9 SM |
2240 | || t->type == tok_keyword |
2241 | // we must allow "*", due to being an operator | |
2242 | || (t->type == tok_operator && t->content == "*"))) | |
2677d2fb | 2243 | throw parse_error (_("expected identifier or '*'")); |
9c0c0e46 | 2244 | |
b5477cd9 SM |
2245 | // loop which reconstitutes an identifier with wildcards |
2246 | string content = t->content; | |
2247 | while (1) | |
2248 | { | |
2249 | const token* u = peek(); | |
3819d181 MW |
2250 | if (u == NULL) |
2251 | break; | |
b5477cd9 SM |
2252 | // ensure pieces of the identifier are adjacent: |
2253 | if (input.ate_whitespace) | |
2254 | break; | |
2255 | // ensure pieces of the identifier are valid: | |
2256 | if (! (u->type == tok_identifier | |
2257 | // we must allow arbitrary keywords with a wildcard | |
2258 | || u->type == tok_keyword | |
2259 | // we must allow "*", due to being an operator | |
2260 | || (u->type == tok_operator && u->content == "*"))) | |
2261 | break; | |
2262 | ||
2263 | // append u to t | |
2264 | content = content + u->content; | |
2265 | ||
2266 | // consume u | |
731a5359 | 2267 | swallow (); |
b5477cd9 | 2268 | } |
534aad8b SM |
2269 | // get around const-ness of t: |
2270 | token* new_t = new token(*t); | |
b5477cd9 SM |
2271 | new_t->content = content; |
2272 | delete t; t = new_t; | |
9c0c0e46 FCE |
2273 | |
2274 | probe_point::component* c = new probe_point::component; | |
2275 | c->functor = t->content; | |
f1a0157a | 2276 | c->tok = t; |
9c0c0e46 | 2277 | pl->components.push_back (c); |
6e3347a9 | 2278 | // NB we may add c->arg soon |
9c0c0e46 FCE |
2279 | |
2280 | t = peek (); | |
a477f3f1 | 2281 | |
6e3347a9 | 2282 | // consume optional parameter |
9c0c0e46 FCE |
2283 | if (t && t->type == tok_operator && t->content == "(") |
2284 | { | |
731a5359 | 2285 | swallow (); // consume "(" |
9c0c0e46 FCE |
2286 | c->arg = parse_literal (); |
2287 | ||
2288 | t = next (); | |
2289 | if (! (t->type == tok_operator && t->content == ")")) | |
2677d2fb | 2290 | throw parse_error (_("expected ')'")); |
731a5359 | 2291 | swallow (); |
9c0c0e46 FCE |
2292 | |
2293 | t = peek (); | |
9c0c0e46 | 2294 | } |
9c0c0e46 FCE |
2295 | |
2296 | if (t && t->type == tok_operator && t->content == ".") | |
6e3347a9 | 2297 | { |
731a5359 | 2298 | swallow (); |
6e3347a9 FCE |
2299 | continue; |
2300 | } | |
2301 | ||
f1a0157a | 2302 | // We only fall through here at the end of a probe point (past |
6e3347a9 FCE |
2303 | // all the dotted/parametrized components). |
2304 | ||
d898100a FCE |
2305 | if (t && t->type == tok_operator && |
2306 | (t->content == "?" || t->content == "!")) | |
6e3347a9 FCE |
2307 | { |
2308 | pl->optional = true; | |
d898100a FCE |
2309 | if (t->content == "!") pl->sufficient = true; |
2310 | // NB: sufficient implies optional | |
731a5359 | 2311 | swallow (); |
6e3347a9 FCE |
2312 | t = peek (); |
2313 | // fall through | |
cbbe8080 MH |
2314 | } |
2315 | ||
2316 | if (t && t->type == tok_keyword && t->content == "if") | |
2317 | { | |
731a5359 | 2318 | swallow (); |
cbbe8080 | 2319 | t = peek (); |
00917a8a | 2320 | if (!(t && t->type == tok_operator && t->content == "(")) |
2677d2fb | 2321 | throw parse_error (_("expected '('")); |
731a5359 | 2322 | swallow (); |
cbbe8080 MH |
2323 | |
2324 | pl->condition = parse_expression (); | |
2325 | ||
2326 | t = peek (); | |
00917a8a | 2327 | if (!(t && t->type == tok_operator && t->content == ")")) |
2677d2fb | 2328 | throw parse_error (_("expected ')'")); |
731a5359 | 2329 | swallow (); |
cbbe8080 MH |
2330 | t = peek (); |
2331 | // fall through | |
6e3347a9 FCE |
2332 | } |
2333 | ||
dff50e09 | 2334 | if (t && t->type == tok_operator |
6e3347a9 FCE |
2335 | && (t->content == "{" || t->content == "," || |
2336 | t->content == "=" || t->content == "+=" )) | |
2337 | break; | |
dff50e09 | 2338 | |
2677d2fb | 2339 | throw parse_error (_("expected one of '. , ( ? ! { = +='")); |
2f1a1aea FCE |
2340 | } |
2341 | ||
2342 | return pl; | |
2343 | } | |
2344 | ||
2345 | ||
d24f1ff4 SM |
2346 | literal_string* |
2347 | parser::consume_string_literals(const token *t) | |
2348 | { | |
2349 | literal_string *ls = new literal_string (t->content); | |
2350 | ||
2351 | // PR11208: check if the next token is also a string literal; | |
2352 | // auto-concatenate it. This is complicated to the extent that we | |
2353 | // need to skip intermediate whitespace. | |
2354 | // | |
2355 | // NB for versions prior to 2.0: but don't skip over intervening comments | |
2356 | const token *n = peek(); | |
2357 | while (n != NULL && n->type == tok_string | |
2358 | && ! (strverscmp(session.compatible.c_str(), "2.0") < 0 | |
2359 | && input.ate_comment)) | |
2360 | { | |
2361 | ls->value.append(next()->content); // consume and append the token | |
2362 | n = peek(); | |
2363 | } | |
2364 | return ls; | |
2365 | } | |
2366 | ||
2367 | ||
2368 | // Parse a string literal and perform backslash escaping on the contents: | |
2369 | literal_string* | |
2370 | parser::parse_literal_string () | |
2371 | { | |
2372 | const token* t = next (); | |
2373 | literal_string* l; | |
2374 | if (t->type == tok_string) | |
2375 | l = consume_string_literals (t); | |
2376 | else | |
6a420ae9 | 2377 | throw parse_error (_("expected literal string")); |
d24f1ff4 SM |
2378 | |
2379 | l->tok = t; | |
2380 | return l; | |
2381 | } | |
2382 | ||
2383 | ||
2f1a1aea FCE |
2384 | literal* |
2385 | parser::parse_literal () | |
2386 | { | |
2387 | const token* t = next (); | |
56099f08 | 2388 | literal* l; |
2f1a1aea | 2389 | if (t->type == tok_string) |
c5be7511 | 2390 | { |
d24f1ff4 | 2391 | l = consume_string_literals (t); |
c5be7511 | 2392 | } |
16e8f21f | 2393 | else |
9c0c0e46 | 2394 | { |
16e8f21f JS |
2395 | bool neg = false; |
2396 | if (t->type == tok_operator && t->content == "-") | |
2397 | { | |
2398 | neg = true; | |
731a5359 | 2399 | swallow (); |
16e8f21f JS |
2400 | t = next (); |
2401 | } | |
2402 | ||
2403 | if (t->type == tok_number) | |
2404 | { | |
2405 | const char* startp = t->content.c_str (); | |
2406 | char* endp = (char*) startp; | |
2407 | ||
2408 | // NB: we allow controlled overflow from LLONG_MIN .. ULLONG_MAX | |
2409 | // Actually, this allows all the way from -ULLONG_MAX to ULLONG_MAX, | |
79e6d33f JS |
2410 | // since the lexer only gives us positive digit strings, but we'll |
2411 | // limit it to LLONG_MIN when a '-' operator is fed into the literal. | |
16e8f21f JS |
2412 | errno = 0; |
2413 | long long value = (long long) strtoull (startp, & endp, 0); | |
16e8f21f | 2414 | if (errno == ERANGE || errno == EINVAL || *endp != '\0' |
79e6d33f | 2415 | || (neg && (unsigned long long) value > 9223372036854775808ULL) |
16e8f21f JS |
2416 | || (unsigned long long) value > 18446744073709551615ULL |
2417 | || value < -9223372036854775807LL-1) | |
2677d2fb | 2418 | throw parse_error (_("number invalid or out of range")); |
16e8f21f | 2419 | |
79e6d33f JS |
2420 | if (neg) |
2421 | value = -value; | |
2422 | ||
16e8f21f JS |
2423 | l = new literal_number (value); |
2424 | } | |
2425 | else | |
2677d2fb | 2426 | throw parse_error (_("expected literal string or number")); |
9c0c0e46 | 2427 | } |
56099f08 FCE |
2428 | |
2429 | l->tok = t; | |
2430 | return l; | |
2f1a1aea FCE |
2431 | } |
2432 | ||
2433 | ||
2434 | if_statement* | |
2435 | parser::parse_if_statement () | |
2436 | { | |
2437 | const token* t = next (); | |
6e213f58 | 2438 | if (! (t->type == tok_keyword && t->content == "if")) |
2677d2fb | 2439 | throw parse_error (_("expected 'if'")); |
56099f08 FCE |
2440 | if_statement* s = new if_statement; |
2441 | s->tok = t; | |
2442 | ||
2443 | t = next (); | |
2f1a1aea | 2444 | if (! (t->type == tok_operator && t->content == "(")) |
2677d2fb | 2445 | throw parse_error (_("expected '('")); |
731a5359 | 2446 | swallow (); |
2f1a1aea | 2447 | |
2f1a1aea FCE |
2448 | s->condition = parse_expression (); |
2449 | ||
2450 | t = next (); | |
2451 | if (! (t->type == tok_operator && t->content == ")")) | |
2677d2fb | 2452 | throw parse_error (_("expected ')'")); |
731a5359 | 2453 | swallow (); |
2f1a1aea FCE |
2454 | |
2455 | s->thenblock = parse_statement (); | |
2456 | ||
2457 | t = peek (); | |
6e213f58 | 2458 | if (t && t->type == tok_keyword && t->content == "else") |
2f1a1aea | 2459 | { |
731a5359 | 2460 | swallow (); |
2f1a1aea FCE |
2461 | s->elseblock = parse_statement (); |
2462 | } | |
ed10c639 FCE |
2463 | else |
2464 | s->elseblock = 0; // in case not otherwise initialized | |
2f1a1aea FCE |
2465 | |
2466 | return s; | |
2467 | } | |
2468 | ||
2469 | ||
69c68955 FCE |
2470 | expr_statement* |
2471 | parser::parse_expr_statement () | |
2472 | { | |
2473 | expr_statement *es = new expr_statement; | |
2474 | const token* t = peek (); | |
5e58d11c MW |
2475 | if (t == NULL) |
2476 | throw parse_error (_("expression statement expected")); | |
731a5359 MW |
2477 | // Copy, we only peeked, parse_expression might swallow. |
2478 | es->tok = new token (*t); | |
69c68955 FCE |
2479 | es->value = parse_expression (); |
2480 | return es; | |
2481 | } | |
2482 | ||
2483 | ||
56099f08 FCE |
2484 | return_statement* |
2485 | parser::parse_return_statement () | |
2486 | { | |
2487 | const token* t = next (); | |
6e213f58 | 2488 | if (! (t->type == tok_keyword && t->content == "return")) |
2677d2fb | 2489 | throw parse_error (_("expected 'return'")); |
6e213f58 | 2490 | if (context != con_function) |
2677d2fb | 2491 | throw parse_error (_("found 'return' not in function context")); |
56099f08 FCE |
2492 | return_statement* s = new return_statement; |
2493 | s->tok = t; | |
2494 | s->value = parse_expression (); | |
2495 | return s; | |
2496 | } | |
2497 | ||
2498 | ||
2499 | delete_statement* | |
2500 | parser::parse_delete_statement () | |
2501 | { | |
2502 | const token* t = next (); | |
6e213f58 | 2503 | if (! (t->type == tok_keyword && t->content == "delete")) |
2677d2fb | 2504 | throw parse_error (_("expected 'delete'")); |
56099f08 FCE |
2505 | delete_statement* s = new delete_statement; |
2506 | s->tok = t; | |
2507 | s->value = parse_expression (); | |
2508 | return s; | |
2509 | } | |
2510 | ||
2511 | ||
f3c26ea5 FCE |
2512 | next_statement* |
2513 | parser::parse_next_statement () | |
2514 | { | |
2515 | const token* t = next (); | |
6e213f58 | 2516 | if (! (t->type == tok_keyword && t->content == "next")) |
2677d2fb | 2517 | throw parse_error (_("expected 'next'")); |
6e213f58 | 2518 | if (context != con_probe) |
2677d2fb | 2519 | throw parse_error (_("found 'next' not in probe context")); |
f3c26ea5 FCE |
2520 | next_statement* s = new next_statement; |
2521 | s->tok = t; | |
2522 | return s; | |
2523 | } | |
2524 | ||
2525 | ||
2526 | break_statement* | |
2527 | parser::parse_break_statement () | |
2528 | { | |
2529 | const token* t = next (); | |
6e213f58 | 2530 | if (! (t->type == tok_keyword && t->content == "break")) |
2677d2fb | 2531 | throw parse_error (_("expected 'break'")); |
f3c26ea5 FCE |
2532 | break_statement* s = new break_statement; |
2533 | s->tok = t; | |
2534 | return s; | |
2535 | } | |
2536 | ||
2537 | ||
2538 | continue_statement* | |
2539 | parser::parse_continue_statement () | |
2540 | { | |
2541 | const token* t = next (); | |
6e213f58 | 2542 | if (! (t->type == tok_keyword && t->content == "continue")) |
2677d2fb | 2543 | throw parse_error (_("expected 'continue'")); |
f3c26ea5 FCE |
2544 | continue_statement* s = new continue_statement; |
2545 | s->tok = t; | |
2546 | return s; | |
2547 | } | |
2548 | ||
2549 | ||
69c68955 FCE |
2550 | for_loop* |
2551 | parser::parse_for_loop () | |
2552 | { | |
f3c26ea5 | 2553 | const token* t = next (); |
6e213f58 | 2554 | if (! (t->type == tok_keyword && t->content == "for")) |
2677d2fb | 2555 | throw parse_error (_("expected 'for'")); |
f3c26ea5 FCE |
2556 | for_loop* s = new for_loop; |
2557 | s->tok = t; | |
2558 | ||
2559 | t = next (); | |
2560 | if (! (t->type == tok_operator && t->content == "(")) | |
2677d2fb | 2561 | throw parse_error (_("expected '('")); |
731a5359 | 2562 | swallow (); |
f3c26ea5 FCE |
2563 | |
2564 | // initializer + ";" | |
2565 | t = peek (); | |
2566 | if (t && t->type == tok_operator && t->content == ";") | |
2567 | { | |
cbfbbf69 | 2568 | s->init = 0; |
731a5359 | 2569 | swallow (); |
f3c26ea5 FCE |
2570 | } |
2571 | else | |
2572 | { | |
2573 | s->init = parse_expr_statement (); | |
2574 | t = next (); | |
2575 | if (! (t->type == tok_operator && t->content == ";")) | |
2677d2fb | 2576 | throw parse_error (_("expected ';'")); |
731a5359 | 2577 | swallow (); |
f3c26ea5 FCE |
2578 | } |
2579 | ||
2580 | // condition + ";" | |
2581 | t = peek (); | |
2582 | if (t && t->type == tok_operator && t->content == ";") | |
2583 | { | |
2584 | literal_number* l = new literal_number(1); | |
2585 | s->cond = l; | |
2586 | s->cond->tok = next (); | |
2587 | } | |
2588 | else | |
2589 | { | |
2590 | s->cond = parse_expression (); | |
2591 | t = next (); | |
2592 | if (! (t->type == tok_operator && t->content == ";")) | |
2677d2fb | 2593 | throw parse_error (_("expected ';'")); |
731a5359 | 2594 | swallow (); |
f3c26ea5 | 2595 | } |
dff50e09 | 2596 | |
f3c26ea5 FCE |
2597 | // increment + ")" |
2598 | t = peek (); | |
2599 | if (t && t->type == tok_operator && t->content == ")") | |
2600 | { | |
cbfbbf69 | 2601 | s->incr = 0; |
731a5359 | 2602 | swallow (); |
f3c26ea5 FCE |
2603 | } |
2604 | else | |
2605 | { | |
2606 | s->incr = parse_expr_statement (); | |
2607 | t = next (); | |
2608 | if (! (t->type == tok_operator && t->content == ")")) | |
2677d2fb | 2609 | throw parse_error (_("expected ')'")); |
731a5359 | 2610 | swallow (); |
f3c26ea5 FCE |
2611 | } |
2612 | ||
2613 | // block | |
2614 | s->block = parse_statement (); | |
2615 | ||
2616 | return s; | |
2617 | } | |
2618 | ||
2619 | ||
2620 | for_loop* | |
2621 | parser::parse_while_loop () | |
2622 | { | |
2623 | const token* t = next (); | |
6e213f58 | 2624 | if (! (t->type == tok_keyword && t->content == "while")) |
2677d2fb | 2625 | throw parse_error (_("expected 'while'")); |
f3c26ea5 FCE |
2626 | for_loop* s = new for_loop; |
2627 | s->tok = t; | |
2628 | ||
2629 | t = next (); | |
2630 | if (! (t->type == tok_operator && t->content == "(")) | |
2677d2fb | 2631 | throw parse_error (_("expected '('")); |
731a5359 | 2632 | swallow (); |
f3c26ea5 FCE |
2633 | |
2634 | // dummy init and incr fields | |
cbfbbf69 FCE |
2635 | s->init = 0; |
2636 | s->incr = 0; | |
f3c26ea5 FCE |
2637 | |
2638 | // condition | |
2639 | s->cond = parse_expression (); | |
2640 | ||
f3c26ea5 FCE |
2641 | t = next (); |
2642 | if (! (t->type == tok_operator && t->content == ")")) | |
2677d2fb | 2643 | throw parse_error (_("expected ')'")); |
731a5359 | 2644 | swallow (); |
dff50e09 | 2645 | |
f3c26ea5 FCE |
2646 | // block |
2647 | s->block = parse_statement (); | |
2648 | ||
2649 | return s; | |
69c68955 FCE |
2650 | } |
2651 | ||
2652 | ||
2653 | foreach_loop* | |
2654 | parser::parse_foreach_loop () | |
2655 | { | |
2656 | const token* t = next (); | |
6e213f58 | 2657 | if (! (t->type == tok_keyword && t->content == "foreach")) |
2677d2fb | 2658 | throw parse_error (_("expected 'foreach'")); |
69c68955 FCE |
2659 | foreach_loop* s = new foreach_loop; |
2660 | s->tok = t; | |
93484556 | 2661 | s->sort_direction = 0; |
fd5689dc | 2662 | s->sort_aggr = sc_none; |
c261711d | 2663 | s->value = NULL; |
27f21e8c | 2664 | s->limit = NULL; |
69c68955 FCE |
2665 | |
2666 | t = next (); | |
2667 | if (! (t->type == tok_operator && t->content == "(")) | |
2677d2fb | 2668 | throw parse_error (_("expected '('")); |
731a5359 | 2669 | swallow (); |
69c68955 | 2670 | |
c261711d JS |
2671 | symbol* lookahead_sym = NULL; |
2672 | int lookahead_sort = 0; | |
2673 | ||
2674 | t = peek (); | |
2675 | if (t && t->type == tok_identifier) | |
2676 | { | |
2677 | next (); | |
2678 | lookahead_sym = new symbol; | |
2679 | lookahead_sym->tok = t; | |
2680 | lookahead_sym->name = t->content; | |
2681 | ||
2682 | t = peek (); | |
2683 | if (t && t->type == tok_operator && | |
2684 | (t->content == "+" || t->content == "-")) | |
2685 | { | |
c261711d | 2686 | lookahead_sort = (t->content == "+") ? 1 : -1; |
731a5359 | 2687 | swallow (); |
c261711d JS |
2688 | } |
2689 | ||
2690 | t = peek (); | |
2691 | if (t && t->type == tok_operator && t->content == "=") | |
2692 | { | |
731a5359 | 2693 | swallow (); |
c261711d JS |
2694 | s->value = lookahead_sym; |
2695 | if (lookahead_sort) | |
2696 | { | |
2697 | s->sort_direction = lookahead_sort; | |
2698 | s->sort_column = 0; | |
2699 | } | |
2700 | lookahead_sym = NULL; | |
2701 | } | |
2702 | } | |
2703 | ||
69c68955 FCE |
2704 | // see also parse_array_in |
2705 | ||
2706 | bool parenthesized = false; | |
2707 | t = peek (); | |
c261711d | 2708 | if (!lookahead_sym && t && t->type == tok_operator && t->content == "[") |
69c68955 | 2709 | { |
731a5359 | 2710 | swallow (); |
69c68955 FCE |
2711 | parenthesized = true; |
2712 | } | |
2713 | ||
c261711d JS |
2714 | if (lookahead_sym) |
2715 | { | |
2716 | s->indexes.push_back (lookahead_sym); | |
2717 | if (lookahead_sort) | |
2718 | { | |
2719 | s->sort_direction = lookahead_sort; | |
2720 | s->sort_column = 1; | |
2721 | } | |
2722 | lookahead_sym = NULL; | |
2723 | } | |
2724 | else while (1) | |
69c68955 FCE |
2725 | { |
2726 | t = next (); | |
2727 | if (! (t->type == tok_identifier)) | |
2677d2fb | 2728 | throw parse_error (_("expected identifier")); |
69c68955 FCE |
2729 | symbol* sym = new symbol; |
2730 | sym->tok = t; | |
2731 | sym->name = t->content; | |
2732 | s->indexes.push_back (sym); | |
2733 | ||
93484556 FCE |
2734 | t = peek (); |
2735 | if (t && t->type == tok_operator && | |
2736 | (t->content == "+" || t->content == "-")) | |
2737 | { | |
2738 | if (s->sort_direction) | |
2677d2fb | 2739 | throw parse_error (_("multiple sort directives")); |
93484556 FCE |
2740 | s->sort_direction = (t->content == "+") ? 1 : -1; |
2741 | s->sort_column = s->indexes.size(); | |
731a5359 | 2742 | swallow (); |
93484556 FCE |
2743 | } |
2744 | ||
69c68955 FCE |
2745 | if (parenthesized) |
2746 | { | |
93484556 | 2747 | t = peek (); |
69c68955 FCE |
2748 | if (t && t->type == tok_operator && t->content == ",") |
2749 | { | |
731a5359 | 2750 | swallow (); |
69c68955 FCE |
2751 | continue; |
2752 | } | |
2753 | else if (t && t->type == tok_operator && t->content == "]") | |
2754 | { | |
731a5359 | 2755 | swallow (); |
69c68955 FCE |
2756 | break; |
2757 | } | |
dff50e09 | 2758 | else |
2677d2fb | 2759 | throw parse_error (_("expected ',' or ']'")); |
69c68955 FCE |
2760 | } |
2761 | else | |
2762 | break; // expecting only one expression | |
2763 | } | |
2764 | ||
2765 | t = next (); | |
6e213f58 | 2766 | if (! (t->type == tok_keyword && t->content == "in")) |
2677d2fb | 2767 | throw parse_error (_("expected 'in'")); |
731a5359 | 2768 | swallow (); |
dff50e09 | 2769 | |
d02548c0 | 2770 | s->base = parse_indexable(); |
69c68955 | 2771 | |
fd5689dc FCE |
2772 | // check for atword, see also expect_ident_or_atword, |
2773 | t = peek (); | |
2774 | if (t && t->type == tok_operator && t->content[0] == '@') | |
2775 | { | |
2776 | if (t->content == "@avg") s->sort_aggr = sc_average; | |
2777 | else if (t->content == "@min") s->sort_aggr = sc_min; | |
2778 | else if (t->content == "@max") s->sort_aggr = sc_max; | |
2779 | else if (t->content == "@count") s->sort_aggr = sc_count; | |
2780 | else if (t->content == "@sum") s->sort_aggr = sc_sum; | |
2781 | else throw parse_error(_("expected statistical operation")); | |
2782 | swallow(); | |
2783 | ||
2784 | t = peek (); | |
2785 | if (! (t && t->type == tok_operator && (t->content == "+" || t->content == "-"))) | |
2786 | throw parse_error(_("expected sort directive")); | |
2787 | } | |
2788 | ||
93484556 FCE |
2789 | t = peek (); |
2790 | if (t && t->type == tok_operator && | |
2791 | (t->content == "+" || t->content == "-")) | |
2792 | { | |
2793 | if (s->sort_direction) | |
2677d2fb | 2794 | throw parse_error (_("multiple sort directives")); |
93484556 FCE |
2795 | s->sort_direction = (t->content == "+") ? 1 : -1; |
2796 | s->sort_column = 0; | |
731a5359 | 2797 | swallow (); |
93484556 FCE |
2798 | } |
2799 | ||
27f21e8c DS |
2800 | t = peek (); |
2801 | if (tok_is(t, tok_keyword, "limit")) | |
2802 | { | |
731a5359 | 2803 | swallow (); // get past the "limit" |
27f21e8c DS |
2804 | s->limit = parse_expression (); |
2805 | } | |
2806 | ||
69c68955 FCE |
2807 | t = next (); |
2808 | if (! (t->type == tok_operator && t->content == ")")) | |
2809 | throw parse_error ("expected ')'"); | |
731a5359 | 2810 | swallow (); |
69c68955 FCE |
2811 | |
2812 | s->block = parse_statement (); | |
2813 | return s; | |
2814 | } | |
2815 | ||
2816 | ||
2f1a1aea FCE |
2817 | expression* |
2818 | parser::parse_expression () | |
2819 | { | |
2820 | return parse_assignment (); | |
2821 | } | |
2822 | ||
2f1a1aea FCE |
2823 | |
2824 | expression* | |
2825 | parser::parse_assignment () | |
2826 | { | |
2827 | expression* op1 = parse_ternary (); | |
2828 | ||
2829 | const token* t = peek (); | |
82919855 | 2830 | // right-associative operators |
dff50e09 | 2831 | if (t && t->type == tok_operator |
2f1a1aea | 2832 | && (t->content == "=" || |
82919855 | 2833 | t->content == "<<<" || |
2f1a1aea | 2834 | t->content == "+=" || |
bb2e3076 FCE |
2835 | t->content == "-=" || |
2836 | t->content == "*=" || | |
2837 | t->content == "/=" || | |
2838 | t->content == "%=" || | |
2839 | t->content == "<<=" || | |
2840 | t->content == ">>=" || | |
2841 | t->content == "&=" || | |
2842 | t->content == "^=" || | |
2843 | t->content == "|=" || | |
d5d7c2cc | 2844 | t->content == ".=" || |
dff50e09 | 2845 | false)) |
2f1a1aea | 2846 | { |
bb2e3076 | 2847 | // NB: lvalueness is checked during elaboration / translation |
2f1a1aea | 2848 | assignment* e = new assignment; |
56099f08 | 2849 | e->left = op1; |
2f1a1aea | 2850 | e->op = t->content; |
56099f08 | 2851 | e->tok = t; |
2f1a1aea | 2852 | next (); |
82919855 | 2853 | e->right = parse_expression (); |
56099f08 | 2854 | op1 = e; |
2f1a1aea | 2855 | } |
56099f08 FCE |
2856 | |
2857 | return op1; | |
2f1a1aea FCE |
2858 | } |
2859 | ||
2860 | ||
2861 | expression* | |
2862 | parser::parse_ternary () | |
2863 | { | |
2864 | expression* op1 = parse_logical_or (); | |
2865 | ||
2866 | const token* t = peek (); | |
2867 | if (t && t->type == tok_operator && t->content == "?") | |
2868 | { | |
2f1a1aea | 2869 | ternary_expression* e = new ternary_expression; |
56099f08 | 2870 | e->tok = t; |
2f1a1aea | 2871 | e->cond = op1; |
56099f08 FCE |
2872 | next (); |
2873 | e->truevalue = parse_expression (); // XXX | |
2f1a1aea FCE |
2874 | |
2875 | t = next (); | |
2876 | if (! (t->type == tok_operator && t->content == ":")) | |
2677d2fb | 2877 | throw parse_error (_("expected ':'")); |
731a5359 | 2878 | swallow (); |
2f1a1aea | 2879 | |
56099f08 | 2880 | e->falsevalue = parse_expression (); // XXX |
2f1a1aea FCE |
2881 | return e; |
2882 | } | |
2883 | else | |
2884 | return op1; | |
2885 | } | |
2886 | ||
2887 | ||
2888 | expression* | |
2889 | parser::parse_logical_or () | |
2890 | { | |
2891 | expression* op1 = parse_logical_and (); | |
dff50e09 | 2892 | |
2f1a1aea | 2893 | const token* t = peek (); |
56099f08 | 2894 | while (t && t->type == tok_operator && t->content == "||") |
2f1a1aea | 2895 | { |
2f1a1aea | 2896 | logical_or_expr* e = new logical_or_expr; |
56099f08 FCE |
2897 | e->tok = t; |
2898 | e->op = t->content; | |
2f1a1aea | 2899 | e->left = op1; |
56099f08 FCE |
2900 | next (); |
2901 | e->right = parse_logical_and (); | |
2902 | op1 = e; | |
2903 | t = peek (); | |
2f1a1aea | 2904 | } |
56099f08 FCE |
2905 | |
2906 | return op1; | |
2f1a1aea FCE |
2907 | } |
2908 | ||
2909 | ||
2910 | expression* | |
2911 | parser::parse_logical_and () | |
2912 | { | |
bb2e3076 | 2913 | expression* op1 = parse_boolean_or (); |
2f1a1aea FCE |
2914 | |
2915 | const token* t = peek (); | |
56099f08 | 2916 | while (t && t->type == tok_operator && t->content == "&&") |
2f1a1aea | 2917 | { |
2f1a1aea FCE |
2918 | logical_and_expr *e = new logical_and_expr; |
2919 | e->left = op1; | |
56099f08 FCE |
2920 | e->op = t->content; |
2921 | e->tok = t; | |
2922 | next (); | |
bb2e3076 FCE |
2923 | e->right = parse_boolean_or (); |
2924 | op1 = e; | |
2925 | t = peek (); | |
2926 | } | |
2927 | ||
2928 | return op1; | |
2929 | } | |
2930 | ||
2931 | ||
2932 | expression* | |
2933 | parser::parse_boolean_or () | |
2934 | { | |
2935 | expression* op1 = parse_boolean_xor (); | |
2936 | ||
2937 | const token* t = peek (); | |
2938 | while (t && t->type == tok_operator && t->content == "|") | |
2939 | { | |
2940 | binary_expression* e = new binary_expression; | |
2941 | e->left = op1; | |
2942 | e->op = t->content; | |
2943 | e->tok = t; | |
2944 | next (); | |
2945 | e->right = parse_boolean_xor (); | |
2946 | op1 = e; | |
2947 | t = peek (); | |
2948 | } | |
2949 | ||
2950 | return op1; | |
2951 | } | |
2952 | ||
2953 | ||
2954 | expression* | |
2955 | parser::parse_boolean_xor () | |
2956 | { | |
2957 | expression* op1 = parse_boolean_and (); | |
2958 | ||
2959 | const token* t = peek (); | |
2960 | while (t && t->type == tok_operator && t->content == "^") | |
2961 | { | |
2962 | binary_expression* e = new binary_expression; | |
2963 | e->left = op1; | |
2964 | e->op = t->content; | |
2965 | e->tok = t; | |
2966 | next (); | |
2967 | e->right = parse_boolean_and (); | |
2968 | op1 = e; | |
2969 | t = peek (); | |
2970 | } | |
2971 | ||
2972 | return op1; | |
2973 | } | |
2974 | ||
2975 | ||
2976 | expression* | |
2977 | parser::parse_boolean_and () | |
2978 | { | |
2979 | expression* op1 = parse_array_in (); | |
2980 | ||
2981 | const token* t = peek (); | |
2982 | while (t && t->type == tok_operator && t->content == "&") | |
2983 | { | |
2984 | binary_expression* e = new binary_expression; | |
2985 | e->left = op1; | |
2986 | e->op = t->content; | |
2987 | e->tok = t; | |
2988 | next (); | |
56099f08 FCE |
2989 | e->right = parse_array_in (); |
2990 | op1 = e; | |
2991 | t = peek (); | |
2f1a1aea | 2992 | } |
56099f08 FCE |
2993 | |
2994 | return op1; | |
2f1a1aea FCE |
2995 | } |
2996 | ||
2997 | ||
2998 | expression* | |
2999 | parser::parse_array_in () | |
3000 | { | |
ce10591c | 3001 | // This is a very tricky case. All these are legit expressions: |
69c68955 | 3002 | // "a in b" "a+0 in b" "[a,b] in c" "[c,(d+0)] in b" |
ce10591c FCE |
3003 | vector<expression*> indexes; |
3004 | bool parenthesized = false; | |
2f1a1aea FCE |
3005 | |
3006 | const token* t = peek (); | |
69c68955 | 3007 | if (t && t->type == tok_operator && t->content == "[") |
ce10591c | 3008 | { |
731a5359 | 3009 | swallow (); |
ce10591c FCE |
3010 | parenthesized = true; |
3011 | } | |
3012 | ||
3013 | while (1) | |
3014 | { | |
93daaca8 | 3015 | expression* op1 = parse_comparison_or_regex_query (); |
ce10591c FCE |
3016 | indexes.push_back (op1); |
3017 | ||
3018 | if (parenthesized) | |
3019 | { | |
3020 | const token* t = peek (); | |
3021 | if (t && t->type == tok_operator && t->content == ",") | |
3022 | { | |
731a5359 | 3023 | swallow (); |
ce10591c FCE |
3024 | continue; |
3025 | } | |
69c68955 | 3026 | else if (t && t->type == tok_operator && t->content == "]") |
ce10591c | 3027 | { |
731a5359 | 3028 | swallow (); |
ce10591c FCE |
3029 | break; |
3030 | } | |
dff50e09 | 3031 | else |
2677d2fb | 3032 | throw parse_error (_("expected ',' or ']'")); |
ce10591c FCE |
3033 | } |
3034 | else | |
3035 | break; // expecting only one expression | |
3036 | } | |
3037 | ||
3038 | t = peek (); | |
6e213f58 | 3039 | if (t && t->type == tok_keyword && t->content == "in") |
2f1a1aea | 3040 | { |
2f1a1aea | 3041 | array_in *e = new array_in; |
56099f08 | 3042 | e->tok = t; |
731a5359 | 3043 | next (); |
ce10591c FCE |
3044 | |
3045 | arrayindex* a = new arrayindex; | |
3046 | a->indexes = indexes; | |
d02548c0 | 3047 | a->base = parse_indexable(); |
d15d767c | 3048 | a->tok = a->base->tok; |
ce10591c | 3049 | e->operand = a; |
2f1a1aea FCE |
3050 | return e; |
3051 | } | |
ce10591c FCE |
3052 | else if (indexes.size() == 1) // no "in" - need one expression only |
3053 | return indexes[0]; | |
2f1a1aea | 3054 | else |
2677d2fb | 3055 | throw parse_error (_("unexpected comma-separated expression list")); |
2f1a1aea FCE |
3056 | } |
3057 | ||
3058 | ||
3059 | expression* | |
93daaca8 | 3060 | parser::parse_comparison_or_regex_query () |
2f1a1aea | 3061 | { |
bb2e3076 | 3062 | expression* op1 = parse_shift (); |
2f1a1aea | 3063 | |
93daaca8 SM |
3064 | // TODOXXX for now, =~ is nonassociative |
3065 | // TODOXXX maybe instead a =~ b == c =~ d --> (a =~ b) == (c =~ d) ?? | |
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 (); | |
d24f1ff4 | 3076 | r->right = r->re = 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; |
cc9001af | 3621 | tsym->target_name = ""; |
bfa7e523 | 3622 | tsym->cu_name = ""; |
30263a73 | 3623 | parse_target_symbol_components(tsym); |
d48afc20 | 3624 | tsym->addressof = addressof; |
30263a73 FCE |
3625 | return tsym; |
3626 | } | |
3627 | ||
06219d6f | 3628 | if (t->type == tok_operator && t->content == "@var") |
cc9001af MW |
3629 | { |
3630 | target_symbol *tsym = new target_symbol; | |
3631 | tsym->tok = t; | |
3632 | tsym->name = t->content; | |
3633 | expect_op("("); | |
3634 | expect_unknown(tok_string, tsym->target_name); | |
bfa7e523 MW |
3635 | size_t found_at = tsym->target_name.find("@"); |
3636 | if (found_at != string::npos) | |
3637 | tsym->cu_name = tsym->target_name.substr(found_at + 1); | |
3638 | else | |
3639 | tsym->cu_name = ""; | |
cc9001af MW |
3640 | expect_op(")"); |
3641 | parse_target_symbol_components(tsym); | |
3642 | tsym->addressof = addressof; | |
3643 | return tsym; | |
3644 | } | |
3645 | ||
3646 | throw parse_error (_("expected @cast, @var or $var")); | |
30263a73 FCE |
3647 | } |
3648 | ||
3649 | ||
3650 | // Parse a @defined(). Given head token has already been consumed. | |
3651 | expression* parser::parse_defined_op (const token* t) | |
3652 | { | |
3653 | defined_op* dop = new defined_op; | |
3654 | dop->tok = t; | |
3655 | expect_op("("); | |
30263a73 | 3656 | // no need for parse_hist_op... etc., as @defined takes only target_symbols as its operand. |
d48afc20 | 3657 | const token* tt = next (); |
30263a73 FCE |
3658 | dop->operand = parse_target_symbol (tt); |
3659 | expect_op(")"); | |
3660 | return dop; | |
3661 | } | |
3662 | ||
3663 | ||
8cc799a5 JS |
3664 | // Parse a @entry(). Given head token has already been consumed. |
3665 | expression* parser::parse_entry_op (const token* t) | |
3666 | { | |
3667 | entry_op* eop = new entry_op; | |
3668 | eop->tok = t; | |
3669 | expect_op("("); | |
3670 | eop->operand = parse_expression (); | |
3671 | expect_op(")"); | |
3672 | return eop; | |
3673 | } | |
3674 | ||
3675 | ||
3689db05 SC |
3676 | // Parse a @perf(). Given head token has already been consumed. |
3677 | expression* parser::parse_perf_op (const token* t) | |
3678 | { | |
3679 | perf_op* pop = new perf_op; | |
3e6a17ee SC |
3680 | |
3681 | if (strverscmp(session.compatible.c_str(), "2.1") < 0) | |
3682 | throw parse_error (_("expected @cast, @var or $var")); | |
3683 | ||
3689db05 SC |
3684 | pop->tok = t; |
3685 | expect_op("("); | |
ace7c23f FCE |
3686 | pop->operand = parse_literal_string (); |
3687 | if (pop->operand->value == "") | |
3688 | throw parse_error (_("expected non-empty string")); | |
3689db05 SC |
3689 | expect_op(")"); |
3690 | return pop; | |
3691 | } | |
3692 | ||
3693 | ||
30263a73 | 3694 | |
81931eab JS |
3695 | void |
3696 | parser::parse_target_symbol_components (target_symbol* e) | |
3697 | { | |
5f36109e JS |
3698 | bool pprint = false; |
3699 | ||
3700 | // check for pretty-print in the form $foo$ | |
277c21bc | 3701 | string &base = e->name; |
5f36109e JS |
3702 | size_t pprint_pos = base.find_last_not_of('$'); |
3703 | if (0 < pprint_pos && pprint_pos < base.length() - 1) | |
3704 | { | |
3705 | string pprint_val = base.substr(pprint_pos + 1); | |
3706 | base.erase(pprint_pos + 1); | |
3707 | e->components.push_back (target_symbol::component(e->tok, pprint_val, true)); | |
3708 | pprint = true; | |
3709 | } | |
3710 | ||
3711 | while (!pprint) | |
81931eab | 3712 | { |
81931eab JS |
3713 | if (peek_op ("->")) |
3714 | { | |
c67847a0 JS |
3715 | const token* t = next(); |
3716 | string member; | |
3717 | expect_ident_or_keyword (member); | |
5f36109e JS |
3718 | |
3719 | // check for pretty-print in the form $foo->$ or $foo->bar$ | |
3720 | pprint_pos = member.find_last_not_of('$'); | |
3721 | string pprint_val; | |
3722 | if (pprint_pos == string::npos || pprint_pos < member.length() - 1) | |
3723 | { | |
3724 | pprint_val = member.substr(pprint_pos + 1); | |
3725 | member.erase(pprint_pos + 1); | |
3726 | pprint = true; | |
3727 | } | |
3728 | ||
3729 | if (!member.empty()) | |
3730 | e->components.push_back (target_symbol::component(t, member)); | |
3731 | if (pprint) | |
3732 | e->components.push_back (target_symbol::component(t, pprint_val, true)); | |
81931eab JS |
3733 | } |
3734 | else if (peek_op ("[")) | |
3735 | { | |
c67847a0 | 3736 | const token* t = next(); |
6fda2dff JS |
3737 | expression* index = parse_expression(); |
3738 | literal_number* ln = dynamic_cast<literal_number*>(index); | |
3739 | if (ln) | |
3740 | e->components.push_back (target_symbol::component(t, ln->value)); | |
3741 | else | |
3742 | e->components.push_back (target_symbol::component(t, index)); | |
81931eab | 3743 | expect_op ("]"); |
81931eab JS |
3744 | } |
3745 | else | |
3746 | break; | |
3747 | } | |
5f36109e JS |
3748 | |
3749 | if (!pprint) | |
3750 | { | |
3751 | // check for pretty-print in the form $foo $ | |
3752 | // i.e. as a separate token, esp. for $foo[i]$ and @cast(...)$ | |
3753 | const token* t = peek(); | |
3819d181 | 3754 | if (t != NULL && t->type == tok_identifier && |
5f36109e JS |
3755 | t->content.find_first_not_of('$') == string::npos) |
3756 | { | |
3757 | t = next(); | |
3758 | e->components.push_back (target_symbol::component(t, t->content, true)); | |
3759 | pprint = true; | |
3760 | } | |
3761 | } | |
3762 | ||
3763 | if (pprint && (peek_op ("->") || peek_op("["))) | |
ce0f6648 | 3764 | throw parse_error(_("-> and [ are not accepted for a pretty-printing variable")); |
81931eab JS |
3765 | } |
3766 | ||
73267b89 | 3767 | /* vim: set sw=2 ts=8 cino=>4,n-2,{2,^-2,t0,(0,u0,w1,M1 : */ |