gdb/completer.c

   1 /* Line completion stuff for GDB, the GNU debugger.
   2    Copyright 2000, 2001 Free Software Foundation, Inc.
   3
   4    This file is part of GDB.
   5
   6    This program is free software; you can redistribute it and/or modify
   7    it under the terms of the GNU General Public License as published by
   8    the Free Software Foundation; either version 2 of the License, or
   9    (at your option) any later version.
  10
  11    This program is distributed in the hope that it will be useful,
  12    but WITHOUT ANY WARRANTY; without even the implied warranty of
  13    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  14    GNU General Public License for more details.
  15
  16    You should have received a copy of the GNU General Public License
  17    along with this program; if not, write to the Free Software
  18    Foundation, Inc., 59 Temple Place - Suite 330,
  19    Boston, MA 02111-1307, USA.  */
  20
  21 #include "defs.h"
  22 #include "symtab.h"
  23 #include "gdbtypes.h"
  24 #include "expression.h"
  25 #include "filenames.h"          /* for DOSish file names */
  26
  27 /* FIXME: This is needed because of lookup_cmd_1().
  28    We should be calling a hook instead so we eliminate the CLI dependency. */
  29 #include "gdbcmd.h"
  30
  31 /* Needed for rl_completer_word_break_characters() and for
  32    filename_completion_function.  */
  33 #include <readline/readline.h>
  34
  35 /* readline defines this.  */
  36 #undef savestring
  37
  38 #include "completer.h"
  39
  40 /* Prototypes for local functions */
  41
  42 /* readline uses the word breaks for two things:
  43    (1) In figuring out where to point the TEXT parameter to the
  44    rl_completion_entry_function.  Since we don't use TEXT for much,
  45    it doesn't matter a lot what the word breaks are for this purpose, but
  46    it does affect how much stuff M-? lists.
  47    (2) If one of the matches contains a word break character, readline
  48    will quote it.  That's why we switch between
  49    gdb_completer_word_break_characters and
  50    gdb_completer_command_word_break_characters.  I'm not sure when
  51    we need this behavior (perhaps for funky characters in C++ symbols?).  */
  52
  53 /* Variables which are necessary for fancy command line editing.  */
  54 static char *gdb_completer_word_break_characters =
  55 " \t\n!@#$%^&*()+=|~`}{[]\"';:?/>.<,-";
  56
  57 /* When completing on command names, we remove '-' from the list of
  58    word break characters, since we use it in command names.  If the
  59    readline library sees one in any of the current completion strings,
  60    it thinks that the string needs to be quoted and automatically supplies
  61    a leading quote. */
  62 static char *gdb_completer_command_word_break_characters =
  63 " \t\n!@#$%^&*()+=|~`}{[]\"';:?/>.<,";
  64
  65 /* When completing on file names, we remove from the list of word
  66    break characters any characters that are commonly used in file
  67    names, such as '-', '+', '~', etc.  Otherwise, readline displays
  68    incorrect completion candidates.  */
  69 #ifdef HAVE_DOS_BASED_FILE_SYSTEM
  70 /* MS-DOS and MS-Windows use colon as part of the drive spec, and most
  71    programs support @foo style response files.  */
  72 static char *gdb_completer_file_name_break_characters = " \t\n*|\"';?><@";
  73 #else
  74 static char *gdb_completer_file_name_break_characters = " \t\n*|\"';:?><";
  75 #endif
  76
  77 /* These are used when completing on locations, which can mix file
  78    names and symbol names separated by a colon.  */
  79 static char *gdb_completer_loc_break_characters = " \t\n*|\"';:?><,";
  80
  81 /* Characters that can be used to quote completion strings.  Note that we
  82    can't include '"' because the gdb C parser treats such quoted sequences
  83    as strings. */
  84 static char *gdb_completer_quote_characters = "'";
  85 \f
  86 /* Accessor for some completer data that may interest other files. */
  87
  88 char *
  89 get_gdb_completer_word_break_characters (void)
  90 {
  91   return gdb_completer_word_break_characters;
  92 }
  93
  94 char *
  95 get_gdb_completer_quote_characters (void)
  96 {
  97   return gdb_completer_quote_characters;
  98 }
  99
 100 /* Complete on filenames.  */
 101 char **
 102 filename_completer (char *text, char *word)
 103 {
 104   int subsequent_name;
 105   char **return_val;
 106   int return_val_used;
 107   int return_val_alloced;
 108
 109   return_val_used = 0;
 110   /* Small for testing.  */
 111   return_val_alloced = 1;
 112   return_val = (char **) xmalloc (return_val_alloced * sizeof (char *));
 113
 114   subsequent_name = 0;
 115   while (1)
 116     {
 117       char *p;
 118       p = filename_completion_function (text, subsequent_name);
 119       if (return_val_used >= return_val_alloced)
 120         {
 121           return_val_alloced *= 2;
 122           return_val =
 123             (char **) xrealloc (return_val,
 124                                 return_val_alloced * sizeof (char *));
 125         }
 126       if (p == NULL)
 127         {
 128           return_val[return_val_used++] = p;
 129           break;
 130         }
 131       /* We need to set subsequent_name to a non-zero value before the
 132          continue line below, because otherwise, if the first file seen
 133          by GDB is a backup file whose name ends in a `~', we will loop
 134          indefinitely.  */
 135       subsequent_name = 1;
 136       /* Like emacs, don't complete on old versions.  Especially useful
 137          in the "source" command.  */
 138       if (p[strlen (p) - 1] == '~')
 139         continue;
 140
 141       {
 142         char *q;
 143         if (word == text)
 144           /* Return exactly p.  */
 145           return_val[return_val_used++] = p;
 146         else if (word > text)
 147           {
 148             /* Return some portion of p.  */
 149             q = xmalloc (strlen (p) + 5);
 150             strcpy (q, p + (word - text));
 151             return_val[return_val_used++] = q;
 152             xfree (p);
 153           }
 154         else
 155           {
 156             /* Return some of TEXT plus p.  */
 157             q = xmalloc (strlen (p) + (text - word) + 5);
 158             strncpy (q, word, text - word);
 159             q[text - word] = '\0';
 160             strcat (q, p);
 161             return_val[return_val_used++] = q;
 162             xfree (p);
 163           }
 164       }
 165     }
 166 #if 0
 167   /* There is no way to do this just long enough to affect quote inserting
 168      without also affecting the next completion.  This should be fixed in
 169      readline.  FIXME.  */
 170   /* Insure that readline does the right thing
 171      with respect to inserting quotes.  */
 172   rl_completer_word_break_characters = "";
 173 #endif
 174   return return_val;
 175 }
 176
 177 /* Complete on locations, which might be of two possible forms:
 178
 179        file:line
 180    or
 181        symbol+offset
 182
 183    This is intended to be used in commands that set breakpoints etc.  */
 184 char **
 185 location_completer (char *text, char *word)
 186 {
 187   int n_syms = 0, n_files = 0;
 188   char ** fn_list = NULL;
 189   char ** list = NULL;
 190   char *p;
 191   int quote_found = 0;
 192   int quoted = *text == '\'' || *text == '"';
 193   int quote_char = '\0';
 194   char *colon = NULL;
 195   char *file_to_match = NULL;
 196   char *symbol_start = text;
 197   char *orig_text = text;
 198   size_t text_len;
 199
 200   /* Do we have an unquoted colon, as in "break foo.c::bar"?  */
 201   for (p = text; *p != '\0'; ++p)
 202     {
 203       if (*p == '\\' && p[1] == '\'')
 204         p++;
 205       else if (*p == '\'' || *p == '"')
 206         {
 207           quote_found = *p;
 208           quote_char = *p++;
 209           while (*p != '\0' && *p != quote_found)
 210             {
 211               if (*p == '\\' && p[1] == quote_found)
 212                 p++;
 213               p++;
 214             }
 215
 216           if (*p == quote_found)
 217             quote_found = 0;
 218           else
 219             break;              /* hit the end of text */
 220         }
 221 #if HAVE_DOS_BASED_FILE_SYSTEM
 222       /* If we have a DOS-style absolute file name at the beginning of
 223          TEXT, and the colon after the drive letter is the only colon
 224          we found, pretend the colon is not there.  */
 225       else if (p < text + 3 && *p == ':' && p == text + 1 + quoted)
 226         ;
 227 #endif
 228       else if (*p == ':' && !colon)
 229         {
 230           colon = p;
 231           symbol_start = p + 1;
 232         }
 233       else if (strchr (gdb_completer_word_break_characters, *p))
 234         symbol_start = p + 1;
 235     }
 236
 237   if (quoted)
 238     text++;
 239   text_len = strlen (text);
 240
 241   /* Where is the file name?  */
 242   if (colon)
 243     {
 244       char *s;
 245
 246       file_to_match = (char *) xmalloc (colon - text + 1);
 247       strncpy (file_to_match, text, colon - text + 1);
 248       /* Remove trailing colons and quotes from the file name.  */
 249       for (s = file_to_match + (colon - text);
 250            s > file_to_match;
 251            s--)
 252         if (*s == ':' || *s == quote_char)
 253           *s = '\0';
 254     }
 255   /* If the text includes a colon, they want completion only on a
 256      symbol name after the colon.  Otherwise, we need to complete on
 257      symbols as well as on files.  */
 258   if (colon)
 259     {
 260       list = make_file_symbol_completion_list (symbol_start, word,
 261                                                file_to_match);
 262       xfree (file_to_match);
 263     }
 264   else
 265     {
 266       list = make_symbol_completion_list (symbol_start, word);
 267       /* If text includes characters which cannot appear in a file
 268          name, they cannot be asking for completion on files.  */
 269       if (strcspn (text, gdb_completer_file_name_break_characters) == text_len)
 270         fn_list = make_source_files_completion_list (text, text);
 271     }
 272
 273   /* How many completions do we have in both lists?  */
 274   if (fn_list)
 275     for ( ; fn_list[n_files]; n_files++)
 276       ;
 277   if (list)
 278     for ( ; list[n_syms]; n_syms++)
 279       ;
 280
 281   /* Make list[] large enough to hold both lists, then catenate
 282      fn_list[] onto the end of list[].  */
 283   if (n_syms && n_files)
 284     {
 285       list = xrealloc (list, (n_syms + n_files + 1) * sizeof (char *));
 286       memcpy (list + n_syms, fn_list, (n_files + 1) * sizeof (char *));
 287       xfree (fn_list);
 288     }
 289   else if (n_files)
 290     {
 291       /* If we only have file names as possible completion, we should
 292          bring them in sync with what rl_complete expects.  The
 293          problem is that if the user types "break /foo/b TAB", and the
 294          possible completions are "/foo/bar" and "/foo/baz"
 295          rl_complete expects us to return "bar" and "baz", without the
 296          leading directories, as possible completions, because `word'
 297          starts at the "b".  But we ignore the value of `word' when we
 298          call make_source_files_completion_list above (because that
 299          would not DTRT when the completion results in both symbols
 300          and file names), so make_source_files_completion_list returns
 301          the full "/foo/bar" and "/foo/baz" strings.  This produces
 302          wrong results when, e.g., there's only one possible
 303          completion, because rl_complete will prepend "/foo/" to each
 304          candidate completion.  The loop below removes that leading
 305          part.  */
 306       for (n_files = 0; fn_list[n_files]; n_files++)
 307         {
 308           memmove (fn_list[n_files], fn_list[n_files] + (word - text),
 309                    strlen (fn_list[n_files]) + 1 - (word - text));
 310         }
 311       /* Return just the file-name list as the result.  */
 312       list = fn_list;
 313     }
 314   else if (!n_syms)
 315     {
 316       /* No completions at all.  As the final resort, try completing
 317          on the entire text as a symbol.  */
 318       list = make_symbol_completion_list (orig_text, word);
 319     }
 320
 321   return list;
 322 }
 323
 324 /* Here are some useful test cases for completion.  FIXME: These should
 325    be put in the test suite.  They should be tested with both M-? and TAB.
 326
 327    "show output-" "radix"
 328    "show output" "-radix"
 329    "p" ambiguous (commands starting with p--path, print, printf, etc.)
 330    "p "  ambiguous (all symbols)
 331    "info t foo" no completions
 332    "info t " no completions
 333    "info t" ambiguous ("info target", "info terminal", etc.)
 334    "info ajksdlfk" no completions
 335    "info ajksdlfk " no completions
 336    "info" " "
 337    "info " ambiguous (all info commands)
 338    "p \"a" no completions (string constant)
 339    "p 'a" ambiguous (all symbols starting with a)
 340    "p b-a" ambiguous (all symbols starting with a)
 341    "p b-" ambiguous (all symbols)
 342    "file Make" "file" (word break hard to screw up here)
 343    "file ../gdb.stabs/we" "ird" (needs to not break word at slash)
 344  */
 345
 346 /* Generate completions one by one for the completer.  Each time we are
 347    called return another potential completion to the caller.
 348    line_completion just completes on commands or passes the buck to the
 349    command's completer function, the stuff specific to symbol completion
 350    is in make_symbol_completion_list.
 351
 352    TEXT is the caller's idea of the "word" we are looking at.
 353
 354    MATCHES is the number of matches that have currently been collected from
 355    calling this completion function.  When zero, then we need to initialize,
 356    otherwise the initialization has already taken place and we can just
 357    return the next potential completion string.
 358
 359    LINE_BUFFER is available to be looked at; it contains the entire text
 360    of the line.  POINT is the offset in that line of the cursor.  You
 361    should pretend that the line ends at POINT.
 362
 363    Returns NULL if there are no more completions, else a pointer to a string
 364    which is a possible completion, it is the caller's responsibility to
 365    free the string.  */
 366
 367 char *
 368 line_completion_function (char *text, int matches, char *line_buffer, int point)
 369 {
 370   static char **list = (char **) NULL;  /* Cache of completions */
 371   static int index;             /* Next cached completion */
 372   char *output = NULL;
 373   char *tmp_command, *p;
 374   /* Pointer within tmp_command which corresponds to text.  */
 375   char *word;
 376   struct cmd_list_element *c, *result_list;
 377
 378   if (matches == 0)
 379     {
 380       /* The caller is beginning to accumulate a new set of completions, so
 381          we need to find all of them now, and cache them for returning one at
 382          a time on future calls. */
 383
 384       if (list)
 385         {
 386           /* Free the storage used by LIST, but not by the strings inside.
 387              This is because rl_complete_internal () frees the strings. */
 388           xfree (list);
 389         }
 390       list = 0;
 391       index = 0;
 392
 393       /* Choose the default set of word break characters to break completions.
 394          If we later find out that we are doing completions on command strings
 395          (as opposed to strings supplied by the individual command completer
 396          functions, which can be any string) then we will switch to the
 397          special word break set for command strings, which leaves out the
 398          '-' character used in some commands.  */
 399
 400       rl_completer_word_break_characters =
 401         gdb_completer_word_break_characters;
 402
 403       /* Decide whether to complete on a list of gdb commands or on symbols. */
 404       tmp_command = (char *) alloca (point + 1);
 405       p = tmp_command;
 406
 407       strncpy (tmp_command, line_buffer, point);
 408       tmp_command[point] = '\0';
 409       /* Since text always contains some number of characters leading up
 410          to point, we can find the equivalent position in tmp_command
 411          by subtracting that many characters from the end of tmp_command.  */
 412       word = tmp_command + point - strlen (text);
 413
 414       if (point == 0)
 415         {
 416           /* An empty line we want to consider ambiguous; that is, it
 417              could be any command.  */
 418           c = (struct cmd_list_element *) -1;
 419           result_list = 0;
 420         }
 421       else
 422         {
 423           c = lookup_cmd_1 (&p, cmdlist, &result_list, 1);
 424         }
 425
 426       /* Move p up to the next interesting thing.  */
 427       while (*p == ' ' || *p == '\t')
 428         {
 429           p++;
 430         }
 431
 432       if (!c)
 433         {
 434           /* It is an unrecognized command.  So there are no
 435              possible completions.  */
 436           list = NULL;
 437         }
 438       else if (c == (struct cmd_list_element *) -1)
 439         {
 440           char *q;
 441
 442           /* lookup_cmd_1 advances p up to the first ambiguous thing, but
 443              doesn't advance over that thing itself.  Do so now.  */
 444           q = p;
 445           while (*q && (isalnum (*q) || *q == '-' || *q == '_'))
 446             ++q;
 447           if (q != tmp_command + point)
 448             {
 449               /* There is something beyond the ambiguous
 450                  command, so there are no possible completions.  For
 451                  example, "info t " or "info t foo" does not complete
 452                  to anything, because "info t" can be "info target" or
 453                  "info terminal".  */
 454               list = NULL;
 455             }
 456           else
 457             {
 458               /* We're trying to complete on the command which was ambiguous.
 459                  This we can deal with.  */
 460               if (result_list)
 461                 {
 462                   list = complete_on_cmdlist (*result_list->prefixlist, p,
 463                                               word);
 464                 }
 465               else
 466                 {
 467                   list = complete_on_cmdlist (cmdlist, p, word);
 468                 }
 469               /* Insure that readline does the right thing with respect to
 470                  inserting quotes.  */
 471               rl_completer_word_break_characters =
 472                 gdb_completer_command_word_break_characters;
 473             }
 474         }
 475       else
 476         {
 477           /* We've recognized a full command.  */
 478
 479           if (p == tmp_command + point)
 480             {
 481               /* There is no non-whitespace in the line beyond the command.  */
 482
 483               if (p[-1] == ' ' || p[-1] == '\t')
 484                 {
 485                   /* The command is followed by whitespace; we need to complete
 486                      on whatever comes after command.  */
 487                   if (c->prefixlist)
 488                     {
 489                       /* It is a prefix command; what comes after it is
 490                          a subcommand (e.g. "info ").  */
 491                       list = complete_on_cmdlist (*c->prefixlist, p, word);
 492
 493                       /* Insure that readline does the right thing
 494                          with respect to inserting quotes.  */
 495                       rl_completer_word_break_characters =
 496                         gdb_completer_command_word_break_characters;
 497                     }
 498                   else if (c->enums)
 499                     {
 500                       list = complete_on_enum (c->enums, p, word);
 501                       rl_completer_word_break_characters =
 502                         gdb_completer_command_word_break_characters;
 503                     }
 504                   else
 505                     {
 506                       /* It is a normal command; what comes after it is
 507                          completed by the command's completer function.  */
 508                       if (c->completer == filename_completer)
 509                         {
 510                           /* Many commands which want to complete on
 511                              file names accept several file names, as
 512                              in "run foo bar >>baz".  So we don't want
 513                              to complete the entire text after the
 514                              command, just the last word.  To this
 515                              end, we need to find the beginning of the
 516                              file name by starting at `word' and going
 517                              backwards.  */
 518                           for (p = word;
 519                                p > tmp_command
 520                                  && strchr (gdb_completer_file_name_break_characters, p[-1]) == NULL;
 521                                p--)
 522                             ;
 523                           rl_completer_word_break_characters =
 524                             gdb_completer_file_name_break_characters;
 525                         }
 526                       else if (c->completer == location_completer)
 527                         {
 528                           /* Commands which complete on locations want to
 529                              see the entire argument.  */
 530                           for (p = word;
 531                                p > tmp_command
 532                                  && p[-1] != ' ' && p[-1] != '\t';
 533                                p--)
 534                             ;
 535                         }
 536                       list = (*c->completer) (p, word);
 537                     }
 538                 }
 539               else
 540                 {
 541                   /* The command is not followed by whitespace; we need to
 542                      complete on the command itself.  e.g. "p" which is a
 543                      command itself but also can complete to "print", "ptype"
 544                      etc.  */
 545                   char *q;
 546
 547                   /* Find the command we are completing on.  */
 548                   q = p;
 549                   while (q > tmp_command)
 550                     {
 551                       if (isalnum (q[-1]) || q[-1] == '-' || q[-1] == '_')
 552                         --q;
 553                       else
 554                         break;
 555                     }
 556
 557                   list = complete_on_cmdlist (result_list, q, word);
 558
 559                   /* Insure that readline does the right thing
 560                      with respect to inserting quotes.  */
 561                   rl_completer_word_break_characters =
 562                     gdb_completer_command_word_break_characters;
 563                 }
 564             }
 565           else
 566             {
 567               /* There is non-whitespace beyond the command.  */
 568
 569               if (c->prefixlist && !c->allow_unknown)
 570                 {
 571                   /* It is an unrecognized subcommand of a prefix command,
 572                      e.g. "info adsfkdj".  */
 573                   list = NULL;
 574                 }
 575               else if (c->enums)
 576                 {
 577                   list = complete_on_enum (c->enums, p, word);
 578                 }
 579               else
 580                 {
 581                   /* It is a normal command.  */
 582                   if (c->completer == filename_completer)
 583                     {
 584                       /* See the commentary above about the specifics
 585                          of file-name completion.  */
 586                       for (p = word;
 587                            p > tmp_command
 588                              && strchr (gdb_completer_file_name_break_characters, p[-1]) == NULL;
 589                            p--)
 590                         ;
 591                       rl_completer_word_break_characters =
 592                         gdb_completer_file_name_break_characters;
 593                     }
 594                   else if (c->completer == location_completer)
 595                     {
 596                       for (p = word;
 597                            p > tmp_command
 598                              && p[-1] != ' ' && p[-1] != '\t';
 599                            p--)
 600                         ;
 601                     }
 602                   list = (*c->completer) (p, word);
 603                 }
 604             }
 605         }
 606     }
 607
 608   /* If we found a list of potential completions during initialization then
 609      dole them out one at a time.  The vector of completions is NULL
 610      terminated, so after returning the last one, return NULL (and continue
 611      to do so) each time we are called after that, until a new list is
 612      available. */
 613
 614   if (list)
 615     {
 616       output = list[index];
 617       if (output)
 618         {
 619           index++;
 620         }
 621     }
 622
 623 #if 0
 624   /* Can't do this because readline hasn't yet checked the word breaks
 625      for figuring out whether to insert a quote.  */
 626   if (output == NULL)
 627     /* Make sure the word break characters are set back to normal for the
 628        next time that readline tries to complete something.  */
 629     rl_completer_word_break_characters =
 630       gdb_completer_word_break_characters;
 631 #endif
 632
 633   return (output);
 634 }
 635 /* Skip over a possibly quoted word (as defined by the quote characters
 636    and word break characters the completer uses).  Returns pointer to the
 637    location after the "word". */
 638
 639 char *
 640 skip_quoted (char *str)
 641 {
 642   char quote_char = '\0';
 643   char *scan;
 644
 645   for (scan = str; *scan != '\0'; scan++)
 646     {
 647       if (quote_char != '\0')
 648         {
 649           /* Ignore everything until the matching close quote char */
 650           if (*scan == quote_char)
 651             {
 652               /* Found matching close quote. */
 653               scan++;
 654               break;
 655             }
 656         }
 657       else if (strchr (gdb_completer_quote_characters, *scan))
 658         {
 659           /* Found start of a quoted string. */
 660           quote_char = *scan;
 661         }
 662       else if (strchr (gdb_completer_word_break_characters, *scan))
 663         {
 664           break;
 665         }
 666     }
 667   return (scan);
 668 }
 669