/
niceSOFT
/
make
Обзор
Документация
Войти
/
niceSOFT
/
make
Код
Задачи
Вики
Пакеты
0
Релизы
0
Аналитика
Безопасность
4.2
read.c
3 355 строк
99 KB
Paul Smith
* read.c (eval): [SV 47960] Skip record waiting files when ignoring.
21 май 2016, 23:22
21 май 2016, 23:22
e2ebea3
Код
Авторство
О чём код?
/* Reading and parsing of makefiles for GNU Make. Copyright (C) 1988-2016 Free Software Foundation, Inc. This file is part of GNU Make. GNU Make is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 3 of the License, or (at your option) any later version. GNU Make is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see <http://www.gnu.org/licenses/>. */ #include "makeint.h" #include <assert.h> #include "filedef.h" #include "dep.h" #include "job.h" #include "commands.h" #include "variable.h" #include "rule.h" #include "debug.h" #include "hash.h" #ifdef WINDOWS32 #include <windows.h> #include "sub_proc.h" #else /* !WINDOWS32 */ #ifndef _AMIGA #ifndef VMS #include <pwd.h> #else struct passwd *getpwnam (char *name); #endif #endif #endif /* !WINDOWS32 */ /* A 'struct ebuffer' controls the origin of the makefile we are currently eval'ing. */ struct ebuffer { char *buffer; /* Start of the current line in the buffer. */ char *bufnext; /* Start of the next line in the buffer. */ char *bufstart; /* Start of the entire buffer. */ unsigned int size; /* Malloc'd size of buffer. */ FILE *fp; /* File, or NULL if this is an internal buffer. */ floc floc; /* Info on the file in fp (if any). */ }; /* Track the modifiers we can have on variable assignments */ struct vmodifiers { unsigned int assign_v:1; unsigned int define_v:1; unsigned int undefine_v:1; unsigned int export_v:1; unsigned int override_v:1; unsigned int private_v:1; }; /* Types of "words" that can be read in a makefile. */ enum make_word_type { w_bogus, w_eol, w_static, w_variable, w_colon, w_dcolon, w_semicolon, w_varassign }; /* A 'struct conditionals' contains the information describing all the active conditionals in a makefile. The global variable 'conditionals' contains the conditionals information for the current makefile. It is initialized from the static structure 'toplevel_conditionals' and is later changed to new structures for included makefiles. */ struct conditionals { unsigned int if_cmds; /* Depth of conditional nesting. */ unsigned int allocated; /* Elts allocated in following arrays. */ char *ignoring; /* Are we ignoring or interpreting? 0=interpreting, 1=not yet interpreted, 2=already interpreted */ char *seen_else; /* Have we already seen an 'else'? */ }; static struct conditionals toplevel_conditionals; static struct conditionals *conditionals = &toplevel_conditionals; /* Default directories to search for include files in */ static const char *default_include_directories[] = { #if defined(WINDOWS32) && !defined(INCLUDEDIR) /* This completely up to the user when they install MSVC or other packages. This is defined as a placeholder. */ # define INCLUDEDIR "." #endif INCLUDEDIR, #ifndef _AMIGA "/usr/gnu/include", "/usr/local/include", "/usr/include", #endif 0 }; /* List of directories to search for include files in */ static const char **include_directories; /* Maximum length of an element of the above. */ static unsigned int max_incl_len; /* The filename and pointer to line number of the makefile currently being read in. */ const floc *reading_file = 0; /* The chain of files read by read_all_makefiles. */ static struct goaldep *read_files = 0; static struct goaldep *eval_makefile (const char *filename, int flags); static void eval (struct ebuffer *buffer, int flags); static long readline (struct ebuffer *ebuf); static void do_undefine (char *name, enum variable_origin origin, struct ebuffer *ebuf); static struct variable *do_define (char *name, enum variable_origin origin, struct ebuffer *ebuf); static int conditional_line (char *line, int len, const floc *flocp); static void record_files (struct nameseq *filenames, const char *pattern, const char *pattern_percent, char *depstr, unsigned int cmds_started, char *commands, unsigned int commands_idx, int two_colon, char prefix, const floc *flocp); static void record_target_var (struct nameseq *filenames, char *defn, enum variable_origin origin, struct vmodifiers *vmod, const floc *flocp); static enum make_word_type get_next_mword (char *buffer, char *delim, char **startp, unsigned int *length); static void remove_comments (char *line); static char *find_char_unquote (char *string, int map); static char *unescape_char (char *string, int c); /* Compare a word, both length and contents. P must point to the word to be tested, and WLEN must be the length. */ #define word1eq(s) (wlen == CSTRLEN (s) && strneq (s, p, CSTRLEN (s))) /* Read in all the makefiles and return a chain of targets to rebuild. */ struct goaldep * read_all_makefiles (const char **makefiles) { unsigned int num_makefiles = 0; /* Create *_LIST variables, to hold the makefiles, targets, and variables we will be reading. */ define_variable_cname ("MAKEFILE_LIST", "", o_file, 0); DB (DB_BASIC, (_("Reading makefiles...\n"))); /* If there's a non-null variable MAKEFILES, its value is a list of files to read first thing. But don't let it prevent reading the default makefiles and don't let the default goal come from there. */ { char *value; char *name, *p; unsigned int length; { /* Turn off --warn-undefined-variables while we expand MAKEFILES. */ int save = warn_undefined_variables_flag; warn_undefined_variables_flag = 0; value = allocated_variable_expand ("$(MAKEFILES)"); warn_undefined_variables_flag = save; } /* Set NAME to the start of next token and LENGTH to its length. MAKEFILES is updated for finding remaining tokens. */ p = value; while ((name = find_next_token ((const char **)&p, &length)) != 0) { if (*p != '\0') *p++ = '\0'; eval_makefile (name, RM_NO_DEFAULT_GOAL|RM_INCLUDED|RM_DONTCARE); } free (value); } /* Read makefiles specified with -f switches. */ if (makefiles != 0) while (*makefiles != 0) { struct goaldep *d = eval_makefile (*makefiles, 0); if (errno) perror_with_name ("", *makefiles); /* Reuse the storage allocated for the read_file. */ *makefiles = dep_name (d); ++num_makefiles; ++makefiles; } /* If there were no -f switches, try the default names. */ if (num_makefiles == 0) { static const char *default_makefiles[] = #ifdef VMS /* all lower case since readdir() (the vms version) 'lowercasifies' */ /* TODO: Above is not always true, this needs more work */ { "makefile.vms", "gnumakefile", "makefile", 0 }; #else #ifdef _AMIGA { "GNUmakefile", "Makefile", "SMakefile", 0 }; #else /* !Amiga && !VMS */ #ifdef WINDOWS32 { "GNUmakefile", "makefile", "Makefile", "makefile.mak", 0 }; #else /* !Amiga && !VMS && !WINDOWS32 */ { "GNUmakefile", "makefile", "Makefile", 0 }; #endif /* !Amiga && !VMS && !WINDOWS32 */ #endif /* AMIGA */ #endif /* VMS */ const char **p = default_makefiles; while (*p != 0 && !file_exists_p (*p)) ++p; if (*p != 0) { eval_makefile (*p, 0); if (errno) perror_with_name ("", *p); } else { /* No default makefile was found. Add the default makefiles to the 'read_files' chain so they will be updated if possible. */ struct goaldep *tail = read_files; /* Add them to the tail, after any MAKEFILES variable makefiles. */ while (tail != 0 && tail->next != 0) tail = tail->next; for (p = default_makefiles; *p != 0; ++p) { struct goaldep *d = alloc_goaldep (); d->file = enter_file (strcache_add (*p)); /* Tell update_goal_chain to bail out as soon as this file is made, and main not to die if we can't make this file. */ d->flags = RM_DONTCARE; if (tail == 0) read_files = d; else tail->next = d; tail = d; } if (tail != 0) tail->next = 0; } } return read_files; } /* Install a new conditional and return the previous one. */ static struct conditionals * install_conditionals (struct conditionals *new) { struct conditionals *save = conditionals; memset (new, '\0', sizeof (*new)); conditionals = new; return save; } /* Free the current conditionals and reinstate a saved one. */ static void restore_conditionals (struct conditionals *saved) { /* Free any space allocated by conditional_line. */ free (conditionals->ignoring); free (conditionals->seen_else); /* Restore state. */ conditionals = saved; } static struct goaldep * eval_makefile (const char *filename, int flags) { struct goaldep *deps; struct ebuffer ebuf; const floc *curfile; char *expanded = 0; int makefile_errno; ebuf.floc.filenm = filename; /* Use the original file name. */ ebuf.floc.lineno = 1; ebuf.floc.offset = 0; if (ISDB (DB_VERBOSE)) { printf (_("Reading makefile '%s'"), filename); if (flags & RM_NO_DEFAULT_GOAL) printf (_(" (no default goal)")); if (flags & RM_INCLUDED) printf (_(" (search path)")); if (flags & RM_DONTCARE) printf (_(" (don't care)")); if (flags & RM_NO_TILDE) printf (_(" (no ~ expansion)")); puts ("..."); } /* First, get a stream to read. */ /* Expand ~ in FILENAME unless it came from 'include', in which case it was already done. */ if (!(flags & RM_NO_TILDE) && filename[0] == '~') { expanded = tilde_expand (filename); if (expanded != 0) filename = expanded; } ENULLLOOP (ebuf.fp, fopen (filename, "r")); /* Save the error code so we print the right message later. */ makefile_errno = errno; /* Check for unrecoverable errors: out of mem or FILE slots. */ switch (makefile_errno) { #ifdef EMFILE case EMFILE: #endif #ifdef ENFILE case ENFILE: #endif case ENOMEM: { const char *err = strerror (makefile_errno); OS (fatal, reading_file, "%s", err); } } /* If the makefile wasn't found and it's either a makefile from the 'MAKEFILES' variable or an included makefile, search the included makefile search path for this makefile. */ if (ebuf.fp == 0 && (flags & RM_INCLUDED) && *filename != '/') { unsigned int i; for (i = 0; include_directories[i] != 0; ++i) { const char *included = concat (3, include_directories[i], "/", filename); ebuf.fp = fopen (included, "r"); if (ebuf.fp) { filename = included; break; } } } /* Now we have the final name for this makefile. Enter it into the cache. */ filename = strcache_add (filename); /* Add FILENAME to the chain of read makefiles. */ deps = alloc_goaldep (); deps->next = read_files; read_files = deps; deps->file = lookup_file (filename); if (deps->file == 0) deps->file = enter_file (filename); filename = deps->file->name; deps->flags = flags; free (expanded); /* If the makefile can't be found at all, give up entirely. */ if (ebuf.fp == 0) { /* If we did some searching, errno has the error from the last attempt, rather from FILENAME itself. Store it in case the caller wants to use it in a message. */ errno = makefile_errno; return deps; } /* Set close-on-exec to avoid leaking the makefile to children, such as $(shell ...). */ #ifdef HAVE_FILENO CLOSE_ON_EXEC (fileno (ebuf.fp)); #endif /* Add this makefile to the list. */ do_variable_definition (&ebuf.floc, "MAKEFILE_LIST", filename, o_file, f_append, 0); /* Evaluate the makefile */ ebuf.size = 200; ebuf.buffer = ebuf.bufnext = ebuf.bufstart = xmalloc (ebuf.size); curfile = reading_file; reading_file = &ebuf.floc; eval (&ebuf, !(flags & RM_NO_DEFAULT_GOAL)); reading_file = curfile; fclose (ebuf.fp); free (ebuf.bufstart); alloca (0); errno = 0; return deps; } void eval_buffer (char *buffer, const floc *flocp) { struct ebuffer ebuf; struct conditionals *saved; struct conditionals new; const floc *curfile; /* Evaluate the buffer */ ebuf.size = strlen (buffer); ebuf.buffer = ebuf.bufnext = ebuf.bufstart = buffer; ebuf.fp = NULL; if (flocp) ebuf.floc = *flocp; else if (reading_file) ebuf.floc = *reading_file; else { ebuf.floc.filenm = NULL; ebuf.floc.lineno = 1; ebuf.floc.offset = 0; } curfile = reading_file; reading_file = &ebuf.floc; saved = install_conditionals (&new); eval (&ebuf, 1); restore_conditionals (saved); reading_file = curfile; alloca (0); } /* Check LINE to see if it's a variable assignment or undefine. It might use one of the modifiers "export", "override", "private", or it might be one of the conditional tokens like "ifdef", "include", etc. If it's not a variable assignment or undefine, VMOD.V_ASSIGN is 0. Returns LINE. Returns a pointer to the first non-modifier character, and sets VMOD based on the modifiers found if any, plus V_ASSIGN is 1. */ static char * parse_var_assignment (const char *line, struct vmodifiers *vmod) { const char *p; memset (vmod, '\0', sizeof (*vmod)); /* Find the start of the next token. If there isn't one we're done. */ NEXT_TOKEN (line); if (*line == '\0') return (char *)line; p = line; while (1) { int wlen; const char *p2; struct variable v; p2 = parse_variable_definition (p, &v); /* If this is a variable assignment, we're done. */ if (p2) break; /* It's not a variable; see if it's a modifier. */ p2 = end_of_token (p); wlen = p2 - p; if (word1eq ("export")) vmod->export_v = 1; else if (word1eq ("override")) vmod->override_v = 1; else if (word1eq ("private")) vmod->private_v = 1; else if (word1eq ("define")) { /* We can't have modifiers after 'define' */ vmod->define_v = 1; p = next_token (p2); break; } else if (word1eq ("undefine")) { /* We can't have modifiers after 'undefine' */ vmod->undefine_v = 1; p = next_token (p2); break; } else /* Not a variable or modifier: this is not a variable assignment. */ return (char *)line; /* It was a modifier. Try the next word. */ p = next_token (p2); if (*p == '\0') return (char *)line; } /* Found a variable assignment or undefine. */ vmod->assign_v = 1; return (char *)p; } /* Read file FILENAME as a makefile and add its contents to the data base. SET_DEFAULT is true if we are allowed to set the default goal. */ static void eval (struct ebuffer *ebuf, int set_default) { char *collapsed = 0; unsigned int collapsed_length = 0; unsigned int commands_len = 200; char *commands; unsigned int commands_idx = 0; unsigned int cmds_started, tgts_started; int ignoring = 0, in_ignored_define = 0; int no_targets = 0; /* Set when reading a rule without targets. */ struct nameseq *filenames = 0; char *depstr = 0; long nlines = 0; int two_colon = 0; char prefix = cmd_prefix; const char *pattern = 0; const char *pattern_percent; floc *fstart; floc fi; #define record_waiting_files() \ do \ { \ if (filenames != 0) \ { \ fi.lineno = tgts_started; \ fi.offset = 0; \ record_files (filenames, pattern, pattern_percent, depstr, \ cmds_started, commands, commands_idx, two_colon, \ prefix, &fi); \ filenames = 0; \ } \ commands_idx = 0; \ no_targets = 0; \ pattern = 0; \ } while (0) pattern_percent = 0; cmds_started = tgts_started = 1; fstart = &ebuf->floc; fi.filenm = ebuf->floc.filenm; /* Loop over lines in the file. The strategy is to accumulate target names in FILENAMES, dependencies in DEPS and commands in COMMANDS. These are used to define a rule when the start of the next rule (or eof) is encountered. When you see a "continue" in the loop below, that means we are moving on to the next line. If you see record_waiting_files(), then the statement we are parsing also finishes the previous rule. */ commands = xmalloc (200); while (1) { unsigned int linelen; char *line; unsigned int wlen; char *p; char *p2; struct vmodifiers vmod; /* At the top of this loop, we are starting a brand new line. */ /* Grab the next line to be evaluated */ ebuf->floc.lineno += nlines; nlines = readline (ebuf); /* If there is nothing left to eval, we're done. */ if (nlines < 0) break; line = ebuf->buffer; /* If this is the first line, check for a UTF-8 BOM and skip it. */ if (ebuf->floc.lineno == 1 && line[0] == (char)0xEF && line[1] == (char)0xBB && line[2] == (char)0xBF) { line += 3; if (ISDB(DB_BASIC)) { if (ebuf->floc.filenm) printf (_("Skipping UTF-8 BOM in makefile '%s'\n"), ebuf->floc.filenm); else printf (_("Skipping UTF-8 BOM in makefile buffer\n")); } } /* If this line is empty, skip it. */ if (line[0] == '\0') continue; linelen = strlen (line); /* Check for a shell command line first. If it is not one, we can stop treating cmd_prefix specially. */ if (line[0] == cmd_prefix) { if (no_targets) /* Ignore the commands in a rule with no targets. */ continue; /* If there is no preceding rule line, don't treat this line as a command, even though it begins with a recipe prefix. SunOS 4 make appears to behave this way. */ if (filenames != 0) { if (ignoring) /* Yep, this is a shell command, and we don't care. */ continue; if (commands_idx == 0) cmds_started = ebuf->floc.lineno; /* Append this command line to the line being accumulated. Skip the initial command prefix character. */ if (linelen + commands_idx > commands_len) { commands_len = (linelen + commands_idx) * 2; commands = xrealloc (commands, commands_len); } memcpy (&commands[commands_idx], line + 1, linelen - 1); commands_idx += linelen - 1; commands[commands_idx++] = '\n'; continue; } } /* This line is not a shell command line. Don't worry about whitespace. Get more space if we need it; we don't need to preserve the current contents of the buffer. */ if (collapsed_length < linelen+1) { collapsed_length = linelen+1; free (collapsed); /* Don't need xrealloc: we don't need to preserve the content. */ collapsed = xmalloc (collapsed_length); } strcpy (collapsed, line); /* Collapse continuation lines. */ collapse_continuations (collapsed); remove_comments (collapsed); /* Get rid if starting space (including formfeed, vtab, etc.) */ p = collapsed; NEXT_TOKEN (p); /* See if this is a variable assignment. We need to do this early, to allow variables with names like 'ifdef', 'export', 'private', etc. */ p = parse_var_assignment (p, &vmod); if (vmod.assign_v) { struct variable *v; enum variable_origin origin = vmod.override_v ? o_override : o_file; /* If we're ignoring then we're done now. */ if (ignoring) { if (vmod.define_v) in_ignored_define = 1; continue; } /* Variable assignment ends the previous rule. */ record_waiting_files (); if (vmod.undefine_v) { do_undefine (p, origin, ebuf); continue; } else if (vmod.define_v) v = do_define (p, origin, ebuf); else v = try_variable_definition (fstart, p, origin, 0); assert (v != NULL); if (vmod.export_v) v->export = v_export; if (vmod.private_v) v->private_var = 1; /* This line has been dealt with. */ continue; } /* If this line is completely empty, ignore it. */ if (*p == '\0') continue; p2 = end_of_token (p); wlen = p2 - p; NEXT_TOKEN (p2); /* If we're in an ignored define, skip this line (but maybe get out). */ if (in_ignored_define) { /* See if this is an endef line (plus optional comment). */ if (word1eq ("endef") && STOP_SET (*p2, MAP_COMMENT|MAP_NUL)) in_ignored_define = 0; continue; } /* Check for conditional state changes. */ { int i = conditional_line (p, wlen, fstart); if (i != -2) { if (i == -1) O (fatal, fstart, _("invalid syntax in conditional")); ignoring = i; continue; } } /* Nothing to see here... move along. */ if (ignoring) continue; /* Manage the "export" keyword used outside of variable assignment as well as "unexport". */ if (word1eq ("export") || word1eq ("unexport")) { int exporting = *p == 'u' ? 0 : 1; /* Export/unexport ends the previous rule. */ record_waiting_files (); /* (un)export by itself causes everything to be (un)exported. */ if (*p2 == '\0') export_all_variables = exporting; else { unsigned int l; const char *cp; char *ap; /* Expand the line so we can use indirect and constructed variable names in an (un)export command. */ cp = ap = allocated_variable_expand (p2); for (p = find_next_token (&cp, &l); p != 0; p = find_next_token (&cp, &l)) { struct variable *v = lookup_variable (p, l); if (v == 0) v = define_variable_global (p, l, "", o_file, 0, fstart); v->export = exporting ? v_export : v_noexport; } free (ap); } continue; } /* Handle the special syntax for vpath. */ if (word1eq ("vpath")) { const char *cp; char *vpat; unsigned int l; /* vpath ends the previous rule. */ record_waiting_files (); cp = variable_expand (p2); p = find_next_token (&cp, &l); if (p != 0) { vpat = xstrndup (p, l); p = find_next_token (&cp, &l); /* No searchpath means remove all previous selective VPATH's with the same pattern. */ } else /* No pattern means remove all previous selective VPATH's. */ vpat = 0; construct_vpath_list (vpat, p); free (vpat); continue; } /* Handle include and variants. */ if (word1eq ("include") || word1eq ("-include") || word1eq ("sinclude")) { /* We have found an 'include' line specifying a nested makefile to be read at this point. */ struct conditionals *save; struct conditionals new_conditionals; struct nameseq *files; /* "-include" (vs "include") says no error if the file does not exist. "sinclude" is an alias for this from SGI. */ int noerror = (p[0] != 'i'); /* Include ends the previous rule. */ record_waiting_files (); p = allocated_variable_expand (p2); /* If no filenames, it's a no-op. */ if (*p == '\0') { free (p); continue; } /* Parse the list of file names. Don't expand archive references! */ p2 = p; files = PARSE_FILE_SEQ (&p2, struct nameseq, MAP_NUL, NULL, PARSEFS_NOAR); free (p); /* Save the state of conditionals and start the included makefile with a clean slate. */ save = install_conditionals (&new_conditionals); /* Record the rules that are waiting so they will determine the default goal before those in the included makefile. */ record_waiting_files (); /* Read each included makefile. */ while (files != 0) { struct nameseq *next = files->next; int flags = (RM_INCLUDED | RM_NO_TILDE | (noerror ? RM_DONTCARE : 0) | (set_default ? 0 : RM_NO_DEFAULT_GOAL)); struct goaldep *d = eval_makefile (files->name, flags); if (errno) { d->error = (unsigned short)errno; d->floc = *fstart; } free_ns (files); files = next; } /* Restore conditional state. */ restore_conditionals (save); continue; } /* Handle the load operations. */ if (word1eq ("load") || word1eq ("-load")) { /* A 'load' line specifies a dynamic object to load. */ struct nameseq *files; int noerror = (p[0] == '-'); /* Load ends the previous rule. */ record_waiting_files (); p = allocated_variable_expand (p2); /* If no filenames, it's a no-op. */ if (*p == '\0') { free (p); continue; } /* Parse the list of file names. Don't expand archive references or strip "./" */ p2 = p; files = PARSE_FILE_SEQ (&p2, struct nameseq, MAP_NUL, NULL, PARSEFS_NOAR); free (p); /* Load each file. */ while (files != 0) { struct nameseq *next = files->next; const char *name = files->name; struct goaldep *deps; int r; /* Load the file. 0 means failure. */ r = load_file (&ebuf->floc, &name, noerror); if (! r && ! noerror) OS (fatal, &ebuf->floc, _("%s: failed to load"), name); free_ns (files); files = next; /* Return of -1 means a special load: don't rebuild it. */ if (r == -1) continue; /* It succeeded, so add it to the list "to be rebuilt". */ deps = alloc_goaldep (); deps->next = read_files; read_files = deps; deps->file = lookup_file (name); if (deps->file == 0) deps->file = enter_file (name); deps->file->loaded = 1; } continue; } /* This line starts with a tab but was not caught above because there was no preceding target, and the line might have been usable as a variable definition. But now we know it is definitely lossage. */ if (line[0] == cmd_prefix) O (fatal, fstart, _("recipe commences before first target")); /* This line describes some target files. This is complicated by the existence of target-specific variables, because we can't expand the entire line until we know if we have one or not. So we expand the line word by word until we find the first ':', then check to see if it's a target-specific variable. In this algorithm, 'lb_next' will point to the beginning of the unexpanded parts of the input buffer, while 'p2' points to the parts of the expanded buffer we haven't searched yet. */ { enum make_word_type wtype; char *cmdleft, *semip, *lb_next; unsigned int plen = 0; char *colonp; const char *end, *beg; /* Helpers for whitespace stripping. */ /* Record the previous rule. */ record_waiting_files (); tgts_started = fstart->lineno; /* Search the line for an unquoted ; that is not after an unquoted #. */ cmdleft = find_char_unquote (line, MAP_SEMI|MAP_COMMENT|MAP_VARIABLE); if (cmdleft != 0 && *cmdleft == '#') { /* We found a comment before a semicolon. */ *cmdleft = '\0'; cmdleft = 0; } else if (cmdleft != 0) /* Found one. Cut the line short there before expanding it. */ *(cmdleft++) = '\0'; semip = cmdleft; collapse_continuations (line); /* We can't expand the entire line, since if it's a per-target variable we don't want to expand it. So, walk from the beginning, expanding as we go, and looking for "interesting" chars. The first word is always expandable. */ wtype = get_next_mword (line, NULL, &lb_next, &wlen); switch (wtype) { case w_eol: if (cmdleft != 0) O (fatal, fstart, _("missing rule before recipe")); /* This line contained something but turned out to be nothing but whitespace (a comment?). */ continue; case w_colon: case w_dcolon: /* We accept and ignore rules without targets for compatibility with SunOS 4 make. */ no_targets = 1; continue; default: break; } p2 = variable_expand_string (NULL, lb_next, wlen); while (1) { lb_next += wlen; if (cmdleft == 0) { /* Look for a semicolon in the expanded line. */ cmdleft = find_char_unquote (p2, MAP_SEMI); if (cmdleft != 0) { unsigned long p2_off = p2 - variable_buffer; unsigned long cmd_off = cmdleft - variable_buffer; char *pend = p2 + strlen (p2); /* Append any remnants of lb, then cut the line short at the semicolon. */ *cmdleft = '\0'; /* One school of thought says that you shouldn't expand here, but merely copy, since now you're beyond a ";" and into a command script. However, the old parser expanded the whole line, so we continue that for backwards-compatibility. Also, it wouldn't be entirely consistent, since we do an unconditional expand below once we know we don't have a target-specific variable. */ (void)variable_expand_string (pend, lb_next, (long)-1); lb_next += strlen (lb_next); p2 = variable_buffer + p2_off; cmdleft = variable_buffer + cmd_off + 1; } } colonp = find_char_unquote (p2, MAP_COLON); #ifdef HAVE_DOS_PATHS /* The drive spec brain-damage strikes again... */ /* Note that the only separators of targets in this context are whitespace and a left paren. If others are possible, they should be added to the string in the call to index. */ while (colonp && (colonp[1] == '/' || colonp[1] == '\\') && colonp > p2 && isalpha ((unsigned char)colonp[-1]) && (colonp == p2 + 1 || strchr (" \t(", colonp[-2]) != 0)) colonp = find_char_unquote (colonp + 1, MAP_COLON); #endif if (colonp != 0) break; wtype = get_next_mword (lb_next, NULL, &lb_next, &wlen); if (wtype == w_eol) break; p2 += strlen (p2); *(p2++) = ' '; p2 = variable_expand_string (p2, lb_next, wlen); /* We don't need to worry about cmdleft here, because if it was found in the variable_buffer the entire buffer has already been expanded... we'll never get here. */ } p2 = next_token (variable_buffer); /* If the word we're looking at is EOL, see if there's _anything_ on the line. If not, a variable expanded to nothing, so ignore it. If so, we can't parse this line so punt. */ if (wtype == w_eol) { if (*p2 == '\0') continue; /* There's no need to be ivory-tower about this: check for one of the most common bugs found in makefiles... */ if (cmd_prefix == '\t' && strneq (line, " ", 8)) O (fatal, fstart, _("missing separator (did you mean TAB instead of 8 spaces?)")); else O (fatal, fstart, _("missing separator")); } /* Make the colon the end-of-string so we know where to stop looking for targets. Start there again once we're done. */ *colonp = '\0'; filenames = PARSE_SIMPLE_SEQ (&p2, struct nameseq); *colonp = ':'; p2 = colonp; if (!filenames) { /* We accept and ignore rules without targets for compatibility with SunOS 4 make. */ no_targets = 1; continue; } /* This should never be possible; we handled it above. */ assert (*p2 != '\0'); ++p2; /* Is this a one-colon or two-colon entry? */ two_colon = *p2 == ':'; if (two_colon) p2++; /* Test to see if it's a target-specific variable. Copy the rest of the buffer over, possibly temporarily (we'll expand it later if it's not a target-specific variable). PLEN saves the length of the unparsed section of p2, for later. */ if (*lb_next != '\0') { unsigned int l = p2 - variable_buffer; plen = strlen (p2); variable_buffer_output (p2+plen, lb_next, strlen (lb_next)+1); p2 = variable_buffer + l; } p2 = parse_var_assignment (p2, &vmod); if (vmod.assign_v) { /* If there was a semicolon found, add it back, plus anything after it. */ if (semip) { unsigned int l = p2 - variable_buffer; *(--semip) = ';'; collapse_continuations (semip); variable_buffer_output (p2 + strlen (p2), semip, strlen (semip)+1); p2 = variable_buffer + l; } record_target_var (filenames, p2, vmod.override_v ? o_override : o_file, &vmod, fstart); filenames = 0; continue; } /* This is a normal target, _not_ a target-specific variable. Unquote any = in the dependency list. */ find_char_unquote (lb_next, MAP_EQUALS); /* Remember the command prefix for this target. */ prefix = cmd_prefix; /* We have some targets, so don't ignore the following commands. */ no_targets = 0; /* Expand the dependencies, etc. */ if (*lb_next != '\0') { unsigned int l = p2 - variable_buffer; (void) variable_expand_string (p2 + plen, lb_next, (long)-1); p2 = variable_buffer + l; /* Look for a semicolon in the expanded line. */ if (cmdleft == 0) { cmdleft = find_char_unquote (p2, MAP_SEMI); if (cmdleft != 0) *(cmdleft++) = '\0'; } } /* Is this a static pattern rule: 'target: %targ: %dep; ...'? */ p = strchr (p2, ':'); while (p != 0 && p[-1] == '\\') { char *q = &p[-1]; int backslash = 0; while (*q-- == '\\') backslash = !backslash; if (backslash) p = strchr (p + 1, ':'); else break; } #ifdef _AMIGA /* Here, the situation is quite complicated. Let's have a look at a couple of targets: install: dev:make dev:make: make dev:make:: xyz The rule is that it's only a target, if there are TWO :'s OR a space around the :. */ if (p && !(ISSPACE (p[1]) || !p[1] || ISSPACE (p[-1]))) p = 0; #endif #ifdef HAVE_DOS_PATHS { int check_again; do { check_again = 0; /* For DOS-style paths, skip a "C:\..." or a "C:/..." */ if (p != 0 && (p[1] == '\\' || p[1] == '/') && isalpha ((unsigned char)p[-1]) && (p == p2 + 1 || strchr (" \t:(", p[-2]) != 0)) { p = strchr (p + 1, ':'); check_again = 1; } } while (check_again); } #endif if (p != 0) { struct nameseq *target; target = PARSE_FILE_SEQ (&p2, struct nameseq, MAP_COLON, NULL, PARSEFS_NOGLOB); ++p2; if (target == 0) O (fatal, fstart, _("missing target pattern")); else if (target->next != 0) O (fatal, fstart, _("multiple target patterns")); pattern_percent = find_percent_cached (&target->name); pattern = target->name; if (pattern_percent == 0) O (fatal, fstart, _("target pattern contains no '%%'")); free_ns (target); } else pattern = 0; /* Strip leading and trailing whitespaces. */ beg = p2; end = beg + strlen (beg) - 1; strip_whitespace (&beg, &end); /* Put all the prerequisites here; they'll be parsed later. */ if (beg <= end && *beg != '\0') depstr = xstrndup (beg, end - beg + 1); else depstr = 0; commands_idx = 0; if (cmdleft != 0) { /* Semicolon means rest of line is a command. */ unsigned int l = strlen (cmdleft); cmds_started = fstart->lineno; /* Add this command line to the buffer. */ if (l + 2 > commands_len) { commands_len = (l + 2) * 2; commands = xrealloc (commands, commands_len); } memcpy (commands, cmdleft, l); commands_idx += l; commands[commands_idx++] = '\n'; } /* Determine if this target should be made default. We used to do this in record_files() but because of the delayed target recording and because preprocessor directives are legal in target's commands it is too late. Consider this fragment for example: foo: ifeq ($(.DEFAULT_GOAL),foo) ... endif Because the target is not recorded until after ifeq directive is evaluated the .DEFAULT_GOAL does not contain foo yet as one would expect. Because of this we have to move the logic here. */ if (set_default && default_goal_var->value[0] == '\0') { struct dep *d; struct nameseq *t = filenames; for (; t != 0; t = t->next) { int reject = 0; const char *name = t->name; /* We have nothing to do if this is an implicit rule. */ if (strchr (name, '%') != 0) break; /* See if this target's name does not start with a '.', unless it contains a slash. */ if (*name == '.' && strchr (name, '/') == 0 #ifdef HAVE_DOS_PATHS && strchr (name, '\\') == 0 #endif ) continue; /* If this file is a suffix, don't let it be the default goal file. */ for (d = suffix_file->deps; d != 0; d = d->next) { register struct dep *d2; if (*dep_name (d) != '.' && streq (name, dep_name (d))) { reject = 1; break; } for (d2 = suffix_file->deps; d2 != 0; d2 = d2->next) { unsigned int l = strlen (dep_name (d2)); if (!strneq (name, dep_name (d2), l)) continue; if (streq (name + l, dep_name (d))) { reject = 1; break; } } if (reject) break; } if (!reject) { define_variable_global (".DEFAULT_GOAL", 13, t->name, o_file, 0, NILF); break; } } } continue; } /* We get here except in the case that we just read a rule line. Record now the last rule we read, so following spurious commands are properly diagnosed. */ record_waiting_files (); } #undef word1eq if (conditionals->if_cmds) O (fatal, fstart, _("missing 'endif'")); /* At eof, record the last rule. */ record_waiting_files (); free (collapsed); free (commands); } /* Remove comments from LINE. This is done by copying the text at LINE onto itself. */ static void remove_comments (char *line) { char *comment; comment = find_char_unquote (line, MAP_COMMENT); if (comment != 0) /* Cut off the line at the #. */ *comment = '\0'; } /* Execute a 'undefine' directive. The undefine line has already been read, and NAME is the name of the variable to be undefined. */ static void do_undefine (char *name, enum variable_origin origin, struct ebuffer *ebuf) { char *p, *var; /* Expand the variable name and find the beginning (NAME) and end. */ var = allocated_variable_expand (name); name = next_token (var); if (*name == '\0') O (fatal, &ebuf->floc, _("empty variable name")); p = name + strlen (name) - 1; while (p > name && ISBLANK (*p)) --p; p[1] = '\0'; undefine_variable_global (name, p - name + 1, origin); free (var); } /* Execute a 'define' directive. The first line has already been read, and NAME is the name of the variable to be defined. The following lines remain to be read. */ static struct variable * do_define (char *name, enum variable_origin origin, struct ebuffer *ebuf) { struct variable *v; struct variable var; floc defstart; int nlevels = 1; unsigned int length = 100; char *definition = xmalloc (length); unsigned int idx = 0; char *p, *n; defstart = ebuf->floc; p = parse_variable_definition (name, &var); if (p == NULL) /* No assignment token, so assume recursive. */ var.flavor = f_recursive; else { if (var.value[0] != '\0') O (error, &defstart, _("extraneous text after 'define' directive")); /* Chop the string before the assignment token to get the name. */ var.name[var.length] = '\0'; } /* Expand the variable name and find the beginning (NAME) and end. */ n = allocated_variable_expand (name); name = next_token (n); if (name[0] == '\0') O (fatal, &defstart, _("empty variable name")); p = name + strlen (name) - 1; while (p > name && ISBLANK (*p)) --p; p[1] = '\0'; /* Now read the value of the variable. */ while (1) { unsigned int len; char *line; long nlines = readline (ebuf); /* If there is nothing left to be eval'd, there's no 'endef'!! */ if (nlines < 0) O (fatal, &defstart, _("missing 'endef', unterminated 'define'")); ebuf->floc.lineno += nlines; line = ebuf->buffer; collapse_continuations (line); /* If the line doesn't begin with a tab, test to see if it introduces another define, or ends one. Stop if we find an 'endef' */ if (line[0] != cmd_prefix) { p = next_token (line); len = strlen (p); /* If this is another 'define', increment the level count. */ if ((len == 6 || (len > 6 && ISBLANK (p[6]))) && strneq (p, "define", 6)) ++nlevels; /* If this is an 'endef', decrement the count. If it's now 0, we've found the last one. */ else if ((len == 5 || (len > 5 && ISBLANK (p[5]))) && strneq (p, "endef", 5)) { p += 5; remove_comments (p); if (*(next_token (p)) != '\0') O (error, &ebuf->floc, _("extraneous text after 'endef' directive")); if (--nlevels == 0) break; } } /* Add this line to the variable definition. */ len = strlen (line); if (idx + len + 1 > length) { length = (idx + len) * 2; definition = xrealloc (definition, length + 1); } memcpy (&definition[idx], line, len); idx += len; /* Separate lines with a newline. */ definition[idx++] = '\n'; } /* We've got what we need; define the variable. */ if (idx == 0) definition[0] = '\0'; else definition[idx - 1] = '\0'; v = do_variable_definition (&defstart, name, definition, origin, var.flavor, 0); free (definition); free (n); return (v); } /* Interpret conditional commands "ifdef", "ifndef", "ifeq", "ifneq", "else" and "endif". LINE is the input line, with the command as its first word. FILENAME and LINENO are the filename and line number in the current makefile. They are used for error messages. Value is -2 if the line is not a conditional at all, -1 if the line is an invalid conditional, 0 if following text should be interpreted, 1 if following text should be ignored. */ static int conditional_line (char *line, int len, const floc *flocp) { const char *cmdname; enum { c_ifdef, c_ifndef, c_ifeq, c_ifneq, c_else, c_endif } cmdtype; unsigned int i; unsigned int o; /* Compare a word, both length and contents. */ #define word1eq(s) (len == CSTRLEN (s) && strneq (s, line, CSTRLEN (s))) #define chkword(s, t) if (word1eq (s)) { cmdtype = (t); cmdname = (s); } /* Make sure this line is a conditional. */ chkword ("ifdef", c_ifdef) else chkword ("ifndef", c_ifndef) else chkword ("ifeq", c_ifeq) else chkword ("ifneq", c_ifneq) else chkword ("else", c_else) else chkword ("endif", c_endif) else return -2; /* Found one: skip past it and any whitespace after it. */ line += len; NEXT_TOKEN (line); #define EXTRATEXT() OS (error, flocp, _("extraneous text after '%s' directive"), cmdname) #define EXTRACMD() OS (fatal, flocp, _("extraneous '%s'"), cmdname) /* An 'endif' cannot contain extra text, and reduces the if-depth by 1 */ if (cmdtype == c_endif) { if (*line != '\0') EXTRATEXT (); if (!conditionals->if_cmds) EXTRACMD (); --conditionals->if_cmds; goto DONE; } /* An 'else' statement can either be simple, or it can have another conditional after it. */ if (cmdtype == c_else) { const char *p; if (!conditionals->if_cmds) EXTRACMD (); o = conditionals->if_cmds - 1; if (conditionals->seen_else[o]) O (fatal, flocp, _("only one 'else' per conditional")); /* Change the state of ignorance. */ switch (conditionals->ignoring[o]) { case 0: /* We've just been interpreting. Never do it again. */ conditionals->ignoring[o] = 2; break; case 1: /* We've never interpreted yet. Maybe this time! */ conditionals->ignoring[o] = 0; break; } /* It's a simple 'else'. */ if (*line == '\0') { conditionals->seen_else[o] = 1; goto DONE; } /* The 'else' has extra text. That text must be another conditional and cannot be an 'else' or 'endif'. */ /* Find the length of the next word. */ for (p = line+1; ! STOP_SET (*p, MAP_SPACE|MAP_NUL); ++p) ; len = p - line; /* If it's 'else' or 'endif' or an illegal conditional, fail. */ if (word1eq ("else") || word1eq ("endif") || conditional_line (line, len, flocp) < 0) EXTRATEXT (); else { /* conditional_line() created a new level of conditional. Raise it back to this level. */ if (conditionals->ignoring[o] < 2) conditionals->ignoring[o] = conditionals->ignoring[o+1]; --conditionals->if_cmds; } goto DONE; } if (conditionals->allocated == 0) { conditionals->allocated = 5; conditionals->ignoring = xmalloc (conditionals->allocated); conditionals->seen_else = xmalloc (conditionals->allocated); } o = conditionals->if_cmds++; if (conditionals->if_cmds > conditionals->allocated) { conditionals->allocated += 5; conditionals->ignoring = xrealloc (conditionals->ignoring, conditionals->allocated); conditionals->seen_else = xrealloc (conditionals->seen_else, conditionals->allocated); } /* Record that we have seen an 'if...' but no 'else' so far. */ conditionals->seen_else[o] = 0; /* Search through the stack to see if we're already ignoring. */ for (i = 0; i < o; ++i) if (conditionals->ignoring[i]) { /* We are already ignoring, so just push a level to match the next "else" or "endif", and keep ignoring. We don't want to expand variables in the condition. */ conditionals->ignoring[o] = 1; return 1; } if (cmdtype == c_ifdef || cmdtype == c_ifndef) { char *var; struct variable *v; char *p; /* Expand the thing we're looking up, so we can use indirect and constructed variable names. */ var = allocated_variable_expand (line); /* Make sure there's only one variable name to test. */ p = end_of_token (var); i = p - var; NEXT_TOKEN (p); if (*p != '\0') return -1; var[i] = '\0'; v = lookup_variable (var, i); conditionals->ignoring[o] = ((v != 0 && *v->value != '\0') == (cmdtype == c_ifndef)); free (var); } else { /* "ifeq" or "ifneq". */ char *s1, *s2; unsigned int l; char termin = *line == '(' ? ',' : *line; if (termin != ',' && termin != '"' && termin != '\'') return -1; s1 = ++line; /* Find the end of the first string. */ if (termin == ',') { int count = 0; for (; *line != '\0'; ++line) if (*line == '(') ++count; else if (*line == ')') --count; else if (*line == ',' && count <= 0) break; } else while (*line != '\0' && *line != termin) ++line; if (*line == '\0') return -1; if (termin == ',') { /* Strip blanks after the first string. */ char *p = line++; while (ISBLANK (p[-1])) --p; *p = '\0'; } else *line++ = '\0'; s2 = variable_expand (s1); /* We must allocate a new copy of the expanded string because variable_expand re-uses the same buffer. */ l = strlen (s2); s1 = alloca (l + 1); memcpy (s1, s2, l + 1); if (termin != ',') /* Find the start of the second string. */ NEXT_TOKEN (line); termin = termin == ',' ? ')' : *line; if (termin != ')' && termin != '"' && termin != '\'') return -1; /* Find the end of the second string. */ if (termin == ')') { int count = 0; s2 = next_token (line); for (line = s2; *line != '\0'; ++line) { if (*line == '(') ++count; else if (*line == ')') { if (count <= 0) break; else --count; } } } else { ++line; s2 = line; while (*line != '\0' && *line != termin) ++line; } if (*line == '\0') return -1; *(line++) = '\0'; NEXT_TOKEN (line); if (*line != '\0') EXTRATEXT (); s2 = variable_expand (s2); conditionals->ignoring[o] = (streq (s1, s2) == (cmdtype == c_ifneq)); } DONE: /* Search through the stack to see if we're ignoring. */ for (i = 0; i < conditionals->if_cmds; ++i) if (conditionals->ignoring[i]) return 1; return 0; } /* Record target-specific variable values for files FILENAMES. TWO_COLON is nonzero if a double colon was used. The links of FILENAMES are freed, and so are any names in it that are not incorporated into other data structures. If the target is a pattern, add the variable to the pattern-specific variable value list. */ static void record_target_var (struct nameseq *filenames, char *defn, enum variable_origin origin, struct vmodifiers *vmod, const floc *flocp) { struct nameseq *nextf; struct variable_set_list *global; global = current_variable_set_list; /* If the variable is an append version, store that but treat it as a normal recursive variable. */ for (; filenames != 0; filenames = nextf) { struct variable *v; const char *name = filenames->name; const char *percent; struct pattern_var *p; nextf = filenames->next; free_ns (filenames); /* If it's a pattern target, then add it to the pattern-specific variable list. */ percent = find_percent_cached (&name); if (percent) { /* Get a reference for this pattern-specific variable struct. */ p = create_pattern_var (name, percent); p->variable.fileinfo = *flocp; /* I don't think this can fail since we already determined it was a variable definition. */ v = assign_variable_definition (&p->variable, defn); assert (v != 0); v->origin = origin; if (v->flavor == f_simple) v->value = allocated_variable_expand (v->value); else v->value = xstrdup (v->value); } else { struct file *f; /* Get a file reference for this file, and initialize it. We don't want to just call enter_file() because that allocates a new entry if the file is a double-colon, which we don't want in this situation. */ f = lookup_file (name); if (!f) f = enter_file (strcache_add (name)); else if (f->double_colon) f = f->double_colon; initialize_file_variables (f, 1); current_variable_set_list = f->variables; v = try_variable_definition (flocp, defn, origin, 1); if (!v) O (fatal, flocp, _("Malformed target-specific variable definition")); current_variable_set_list = global; } /* Set up the variable to be *-specific. */ v->per_target = 1; v->private_var = vmod->private_v; v->export = vmod->export_v ? v_export : v_default; /* If it's not an override, check to see if there was a command-line setting. If so, reset the value. */ if (v->origin != o_override) { struct variable *gv; int len = strlen (v->name); gv = lookup_variable (v->name, len); if (gv && v != gv && (gv->origin == o_env_override || gv->origin == o_command)) { free (v->value); v->value = xstrdup (gv->value); v->origin = gv->origin; v->recursive = gv->recursive; v->append = 0; } } } } /* Record a description line for files FILENAMES, with dependencies DEPS, commands to execute described by COMMANDS and COMMANDS_IDX, coming from FILENAME:COMMANDS_STARTED. TWO_COLON is nonzero if a double colon was used. If not nil, PATTERN is the '%' pattern to make this a static pattern rule, and PATTERN_PERCENT is a pointer to the '%' within it. The links of FILENAMES are freed, and so are any names in it that are not incorporated into other data structures. */ static void record_files (struct nameseq *filenames, const char *pattern, const char *pattern_percent, char *depstr, unsigned int cmds_started, char *commands, unsigned int commands_idx, int two_colon, char prefix, const floc *flocp) { struct commands *cmds; struct dep *deps; const char *implicit_percent; const char *name; /* If we've already snapped deps, that means we're in an eval being resolved after the makefiles have been read in. We can't add more rules at this time, since they won't get snapped and we'll get core dumps. See Savannah bug # 12124. */ if (snapped_deps) O (fatal, flocp, _("prerequisites cannot be defined in recipes")); /* Determine if this is a pattern rule or not. */ name = filenames->name; implicit_percent = find_percent_cached (&name); /* If there's a recipe, set up a struct for it. */ if (commands_idx > 0) { cmds = xmalloc (sizeof (struct commands)); cmds->fileinfo.filenm = flocp->filenm; cmds->fileinfo.lineno = cmds_started; cmds->fileinfo.offset = 0; cmds->commands = xstrndup (commands, commands_idx); cmds->command_lines = 0; cmds->recipe_prefix = prefix; } else cmds = 0; /* If there's a prereq string then parse it--unless it's eligible for 2nd expansion: if so, snap_deps() will do it. */ if (depstr == 0) deps = 0; else { depstr = unescape_char (depstr, ':'); if (second_expansion && strchr (depstr, '$')) { deps = alloc_dep (); deps->name = depstr; deps->need_2nd_expansion = 1; deps->staticpattern = pattern != 0; } else { deps = split_prereqs (depstr); free (depstr); /* We'll enter static pattern prereqs later when we have the stem. We don't want to enter pattern rules at all so that we don't think that they ought to exist (make manual "Implicit Rule Search Algorithm", item 5c). */ if (! pattern && ! implicit_percent) deps = enter_prereqs (deps, NULL); } } /* For implicit rules, _all_ the targets must have a pattern. That means we can test the first one to see if we're working with an implicit rule; if so we handle it specially. */ if (implicit_percent) { struct nameseq *nextf; const char **targets, **target_pats; unsigned int c; if (pattern != 0) O (fatal, flocp, _("mixed implicit and static pattern rules")); /* Count the targets to create an array of target names. We already have the first one. */ nextf = filenames->next; free_ns (filenames); filenames = nextf; for (c = 1; nextf; ++c, nextf = nextf->next) ; targets = xmalloc (c * sizeof (const char *)); target_pats = xmalloc (c * sizeof (const char *)); targets[0] = name; target_pats[0] = implicit_percent; c = 1; while (filenames) { name = filenames->name; implicit_percent = find_percent_cached (&name); if (implicit_percent == 0) O (fatal, flocp, _("mixed implicit and normal rules")); targets[c] = name; target_pats[c] = implicit_percent; ++c; nextf = filenames->next; free_ns (filenames); filenames = nextf; } create_pattern_rule (targets, target_pats, c, two_colon, deps, cmds, 1); return; } /* Walk through each target and create it in the database. We already set up the first target, above. */ while (1) { struct nameseq *nextf = filenames->next; struct file *f; struct dep *this = 0; free_ns (filenames); /* Check for special targets. Do it here instead of, say, snap_deps() so that we can immediately use the value. */ if (streq (name, ".POSIX")) { posix_pedantic = 1; define_variable_cname (".SHELLFLAGS", "-ec", o_default, 0); /* These default values are based on IEEE Std 1003.1-2008. */ define_variable_cname ("ARFLAGS", "-rv", o_default, 0); define_variable_cname ("CC", "c99", o_default, 0); define_variable_cname ("CFLAGS", "-O", o_default, 0); define_variable_cname ("FC", "fort77", o_default, 0); define_variable_cname ("FFLAGS", "-O 1", o_default, 0); define_variable_cname ("SCCSGETFLAGS", "-s", o_default, 0); } else if (streq (name, ".SECONDEXPANSION")) second_expansion = 1; #if !defined (__MSDOS__) && !defined (__EMX__) else if (streq (name, ".ONESHELL")) one_shell = 1; #endif /* If this is a static pattern rule: 'targets: target%pattern: prereq%pattern; recipe', make sure the pattern matches this target name. */ if (pattern && !pattern_matches (pattern, pattern_percent, name)) OS (error, flocp, _("target '%s' doesn't match the target pattern"), name); else if (deps) /* If there are multiple targets, copy the chain DEPS for all but the last one. It is not safe for the same deps to go in more than one place in the database. */ this = nextf != 0 ? copy_dep_chain (deps) : deps; /* Find or create an entry in the file database for this target. */ if (!two_colon) { /* Single-colon. Combine this rule with the file's existing record, if any. */ f = enter_file (strcache_add (name)); if (f->double_colon) OS (fatal, flocp, _("target file '%s' has both : and :: entries"), f->name); /* If CMDS == F->CMDS, this target was listed in this rule more than once. Just give a warning since this is harmless. */ if (cmds != 0 && cmds == f->cmds) OS (error, flocp, _("target '%s' given more than once in the same rule"), f->name); /* Check for two single-colon entries both with commands. Check is_target so that we don't lose on files such as .c.o whose commands were preinitialized. */ else if (cmds != 0 && f->cmds != 0 && f->is_target) { size_t l = strlen (f->name); error (&cmds->fileinfo, l, _("warning: overriding recipe for target '%s'"), f->name); error (&f->cmds->fileinfo, l, _("warning: ignoring old recipe for target '%s'"), f->name); } /* Defining .DEFAULT with no deps or cmds clears it. */ if (f == default_file && this == 0 && cmds == 0) f->cmds = 0; if (cmds != 0) f->cmds = cmds; /* Defining .SUFFIXES with no dependencies clears out the list of suffixes. */ if (f == suffix_file && this == 0) { free_dep_chain (f->deps); f->deps = 0; } } else { /* Double-colon. Make a new record even if there already is one. */ f = lookup_file (name); /* Check for both : and :: rules. Check is_target so we don't lose on default suffix rules or makefiles. */ if (f != 0 && f->is_target && !f->double_colon) OS (fatal, flocp, _("target file '%s' has both : and :: entries"), f->name); f = enter_file (strcache_add (name)); /* If there was an existing entry and it was a double-colon entry, enter_file will have returned a new one, making it the prev pointer of the old one, and setting its double_colon pointer to the first one. */ if (f->double_colon == 0) /* This is the first entry for this name, so we must set its double_colon pointer to itself. */ f->double_colon = f; f->cmds = cmds; } f->is_target = 1; /* If this is a static pattern rule, set the stem to the part of its name that matched the '%' in the pattern, so you can use $* in the commands. If we didn't do it before, enter the prereqs now. */ if (pattern) { static const char *percent = "%"; char *buffer = variable_expand (""); char *o = patsubst_expand_pat (buffer, name, pattern, percent, pattern_percent+1, percent+1); f->stem = strcache_add_len (buffer, o - buffer); if (this) { if (! this->need_2nd_expansion) this = enter_prereqs (this, f->stem); else this->stem = f->stem; } } /* Add the dependencies to this file entry. */ if (this != 0) { /* Add the file's old deps and the new ones in THIS together. */ if (f->deps == 0) f->deps = this; else if (cmds != 0) { struct dep *d = this; /* If this rule has commands, put these deps first. */ while (d->next != 0) d = d->next; d->next = f->deps; f->deps = this; } else { struct dep *d = f->deps; /* A rule without commands: put its prereqs at the end. */ while (d->next != 0) d = d->next; d->next = this; } } name = f->name; /* All done! Set up for the next one. */ if (nextf == 0) break; filenames = nextf; /* Reduce escaped percents. If there are any unescaped it's an error */ name = filenames->name; if (find_percent_cached (&name)) O (error, flocp, _("*** mixed implicit and normal rules: deprecated syntax")); } } /* Search STRING for an unquoted STOPCHAR or blank (if BLANK is nonzero). Backslashes quote STOPCHAR, blanks if BLANK is nonzero, and backslash. Quoting backslashes are removed from STRING by compacting it into itself. Returns a pointer to the first unquoted STOPCHAR if there is one, or nil if there are none. STOPCHARs inside variable references are ignored if IGNOREVARS is true. STOPCHAR _cannot_ be '$' if IGNOREVARS is true. */ static char * find_char_unquote (char *string, int map) { unsigned int string_len = 0; char *p = string; /* Always stop on NUL. */ map |= MAP_NUL; while (1) { while (! STOP_SET (*p, map)) ++p; if (*p == '\0') break; /* If we stopped due to a variable reference, skip over its contents. */ if (STOP_SET (*p, MAP_VARIABLE)) { char openparen = p[1]; /* Check if '$' is the last character in the string. */ if (openparen == '\0') break; p += 2; /* Skip the contents of a non-quoted, multi-char variable ref. */ if (openparen == '(' || openparen == '{') { unsigned int pcount = 1; char closeparen = (openparen == '(' ? ')' : '}'); while (*p) { if (*p == openparen) ++pcount; else if (*p == closeparen) if (--pcount == 0) { ++p; break; } ++p; } } /* Skipped the variable reference: look for STOPCHARS again. */ continue; } if (p > string && p[-1] == '\\') { /* Search for more backslashes. */ int i = -2; while (&p[i] >= string && p[i] == '\\') --i; ++i; /* Only compute the length if really needed. */ if (string_len == 0) string_len = strlen (string); /* The number of backslashes is now -I. Copy P over itself to swallow half of them. */ memmove (&p[i], &p[i/2], (string_len - (p - string)) - (i/2) + 1); p += i/2; if (i % 2 == 0) /* All the backslashes quoted each other; the STOPCHAR was unquoted. */ return p; /* The STOPCHAR was quoted by a backslash. Look for another. */ } else /* No backslash in sight. */ return p; } /* Never hit a STOPCHAR or blank (with BLANK nonzero). */ return 0; } /* Unescape a character in a string. The string is compressed onto itself. */ static char * unescape_char (char *string, int c) { char *p = string; char *s = string; while (*s != '\0') { if (*s == '\\') { char *e = s; int l; /* We found a backslash. See if it's escaping our character. */ while (*e == '\\') ++e; l = e - s; if (*e != c || l%2 == 0) { /* It's not; just take it all without unescaping. */ memmove (p, s, l); p += l; // If we hit the end of the string, we're done if (*e == '\0') break; } else if (l > 1) { /* It is, and there's >1 backslash. Take half of them. */ l /= 2; memmove (p, s, l); p += l; } s = e; } *(p++) = *(s++); } *p = '\0'; return string; } /* Search PATTERN for an unquoted % and handle quoting. */ char * find_percent (char *pattern) { return find_char_unquote (pattern, MAP_PERCENT); } /* Search STRING for an unquoted % and handle quoting. Returns a pointer to the % or NULL if no % was found. This version is used with strings in the string cache: if there's a need to modify the string a new version will be added to the string cache and *STRING will be set to that. */ const char * find_percent_cached (const char **string) { const char *p = *string; char *new = 0; int slen = 0; /* If the first char is a % return now. This lets us avoid extra tests inside the loop. */ if (*p == '%') return p; while (1) { while (! STOP_SET (*p, MAP_PERCENT|MAP_NUL)) ++p; if (*p == '\0') break; /* See if this % is escaped with a backslash; if not we're done. */ if (p[-1] != '\\') break; { /* Search for more backslashes. */ char *pv; int i = -2; while (&p[i] >= *string && p[i] == '\\') --i; ++i; /* At this point we know we'll need to allocate a new string. Make a copy if we haven't yet done so. */ if (! new) { slen = strlen (*string); new = alloca (slen + 1); memcpy (new, *string, slen + 1); p = new + (p - *string); *string = new; } /* At this point *string, p, and new all point into the same string. Get a non-const version of p so we can modify new. */ pv = new + (p - *string); /* The number of backslashes is now -I. Copy P over itself to swallow half of them. */ memmove (&pv[i], &pv[i/2], (slen - (pv - new)) - (i/2) + 1); p += i/2; /* If the backslashes quoted each other; the % was unquoted. */ if (i % 2 == 0) break; } } /* If we had to change STRING, add it to the strcache. */ if (new) { *string = strcache_add (*string); p = *string + (p - new); } /* If we didn't find a %, return NULL. Otherwise return a ptr to it. */ return (*p == '\0') ? NULL : p; } /* Find the next line of text in an eval buffer, combining continuation lines into one line. Return the number of actual lines read (> 1 if continuation lines). Returns -1 if there's nothing left in the buffer. After this function, ebuf->buffer points to the first character of the line we just found. */ /* Read a line of text from a STRING. Since we aren't really reading from a file, don't bother with linenumbers. */ static long readstring (struct ebuffer *ebuf) { char *eol; /* If there is nothing left in this buffer, return 0. */ if (ebuf->bufnext >= ebuf->bufstart + ebuf->size) return -1; /* Set up a new starting point for the buffer, and find the end of the next logical line (taking into account backslash/newline pairs). */ eol = ebuf->buffer = ebuf->bufnext; while (1) { int backslash = 0; const char *bol = eol; const char *p; /* Find the next newline. At EOS, stop. */ p = eol = strchr (eol , '\n'); if (!eol) { ebuf->bufnext = ebuf->bufstart + ebuf->size + 1; return 0; } /* Found a newline; if it's escaped continue; else we're done. */ while (p > bol && *(--p) == '\\') backslash = !backslash; if (!backslash) break; ++eol; } /* Overwrite the newline char. */ *eol = '\0'; ebuf->bufnext = eol+1; return 0; } static long readline (struct ebuffer *ebuf) { char *p; char *end; char *start; long nlines = 0; /* The behaviors between string and stream buffers are different enough to warrant different functions. Do the Right Thing. */ if (!ebuf->fp) return readstring (ebuf); /* When reading from a file, we always start over at the beginning of the buffer for each new line. */ p = start = ebuf->bufstart; end = p + ebuf->size; *p = '\0'; while (fgets (p, end - p, ebuf->fp) != 0) { char *p2; unsigned long len; int backslash; len = strlen (p); if (len == 0) { /* This only happens when the first thing on the line is a '\0'. It is a pretty hopeless case, but (wonder of wonders) Athena lossage strikes again! (xmkmf puts NULs in its makefiles.) There is nothing really to be done; we synthesize a newline so the following line doesn't appear to be part of this line. */ O (error, &ebuf->floc, _("warning: NUL character seen; rest of line ignored")); p[0] = '\n'; len = 1; } /* Jump past the text we just read. */ p += len; /* If the last char isn't a newline, the whole line didn't fit into the buffer. Get some more buffer and try again. */ if (p[-1] != '\n') goto more_buffer; /* We got a newline, so add one to the count of lines. */ ++nlines; #if !defined(WINDOWS32) && !defined(__MSDOS__) && !defined(__EMX__) /* Check to see if the line was really ended with CRLF; if so ignore the CR. */ if ((p - start) > 1 && p[-2] == '\r') { --p; memmove (p-1, p, strlen (p) + 1); } #endif backslash = 0; for (p2 = p - 2; p2 >= start; --p2) { if (*p2 != '\\') break; backslash = !backslash; } if (!backslash) { p[-1] = '\0'; break; } /* It was a backslash/newline combo. If we have more space, read another line. */ if (end - p >= 80) continue; /* We need more space at the end of our buffer, so realloc it. Make sure to preserve the current offset of p. */ more_buffer: { unsigned long off = p - start; ebuf->size *= 2; start = ebuf->buffer = ebuf->bufstart = xrealloc (start, ebuf->size); p = start + off; end = start + ebuf->size; *p = '\0'; } } if (ferror (ebuf->fp)) pfatal_with_name (ebuf->floc.filenm); /* If we found some lines, return how many. If we didn't, but we did find _something_, that indicates we read the last line of a file with no final newline; return 1. If we read nothing, we're at EOF; return -1. */ return nlines ? nlines : p == ebuf->bufstart ? -1 : 1; } /* Parse the next "makefile word" from the input buffer, and return info about it. A "makefile word" is one of: w_bogus Should never happen w_eol End of input w_static A static word; cannot be expanded w_variable A word containing one or more variables/functions w_colon A colon w_dcolon A double-colon w_semicolon A semicolon w_varassign A variable assignment operator (=, :=, ::=, +=, ?=, or !=) Note that this function is only used when reading certain parts of the makefile. Don't use it where special rules hold sway (RHS of a variable, in a command list, etc.) */ static enum make_word_type get_next_mword (char *buffer, char *delim, char **startp, unsigned int *length) { enum make_word_type wtype = w_bogus; char *p = buffer, *beg; char c; /* Skip any leading whitespace. */ while (ISBLANK (*p)) ++p; beg = p; c = *(p++); switch (c) { case '\0': wtype = w_eol; break; case ';': wtype = w_semicolon; break; case '=': wtype = w_varassign; break; case ':': wtype = w_colon; switch (*p) { case ':': ++p; if (p[1] != '=') wtype = w_dcolon; else { wtype = w_varassign; ++p; } break; case '=': ++p; wtype = w_varassign; break; } break; case '+': case '?': case '!': if (*p == '=') { ++p; wtype = w_varassign; break; } default: if (delim && strchr (delim, c)) wtype = w_static; break; } /* Did we find something? If so, return now. */ if (wtype != w_bogus) goto done; /* This is some non-operator word. A word consists of the longest string of characters that doesn't contain whitespace, one of [:=#], or [?+!]=, or one of the chars in the DELIM string. */ /* We start out assuming a static word; if we see a variable we'll adjust our assumptions then. */ wtype = w_static; /* We already found the first value of "c", above. */ while (1) { char closeparen; int count; switch (c) { case '\0': case ' ': case '\t': case '=': goto done_word; case ':': #ifdef HAVE_DOS_PATHS /* A word CAN include a colon in its drive spec. The drive spec is allowed either at the beginning of a word, or as part of the archive member name, like in "libfoo.a(d:/foo/bar.o)". */ if (!(p - beg >= 2 && (*p == '/' || *p == '\\') && isalpha ((unsigned char)p[-2]) && (p - beg == 2 || p[-3] == '('))) #endif goto done_word; case '$': c = *(p++); if (c == '$') break; if (c == '\0') goto done_word; /* This is a variable reference, so note that it's expandable. Then read it to the matching close paren. */ wtype = w_variable; if (c == '(') closeparen = ')'; else if (c == '{') closeparen = '}'; else /* This is a single-letter variable reference. */ break; for (count=0; *p != '\0'; ++p) { if (*p == c) ++count; else if (*p == closeparen && --count < 0) { ++p; break; } } break; case '?': case '+': if (*p == '=') goto done_word; break; case '\\': switch (*p) { case ':': case ';': case '=': case '\\': ++p; break; } break; default: if (delim && strchr (delim, c)) goto done_word; break; } c = *(p++); } done_word: --p; done: if (startp) *startp = beg; if (length) *length = p - beg; return wtype; } /* Construct the list of include directories from the arguments and the default list. */ void construct_include_path (const char **arg_dirs) { #ifdef VAXC /* just don't ask ... */ stat_t stbuf; #else struct stat stbuf; #endif const char **dirs; const char **cpp; unsigned int idx; /* Compute the number of pointers we need in the table. */ idx = sizeof (default_include_directories) / sizeof (const char *); if (arg_dirs) for (cpp = arg_dirs; *cpp != 0; ++cpp) ++idx; #ifdef __MSDOS__ /* Add one for $DJDIR. */ ++idx; #endif dirs = xmalloc (idx * sizeof (const char *)); idx = 0; max_incl_len = 0; /* First consider any dirs specified with -I switches. Ignore any that don't exist. Remember the maximum string length. */ if (arg_dirs) while (*arg_dirs != 0) { const char *dir = *(arg_dirs++); char *expanded = 0; int e; if (dir[0] == '~') { expanded = tilde_expand (dir); if (expanded != 0) dir = expanded; } EINTRLOOP (e, stat (dir, &stbuf)); if (e == 0 && S_ISDIR (stbuf.st_mode)) { unsigned int len = strlen (dir); /* If dir name is written with trailing slashes, discard them. */ while (len > 1 && dir[len - 1] == '/') --len; if (len > max_incl_len) max_incl_len = len; dirs[idx++] = strcache_add_len (dir, len); } free (expanded); } /* Now add the standard default dirs at the end. */ #ifdef __MSDOS__ { /* The environment variable $DJDIR holds the root of the DJGPP directory tree; add ${DJDIR}/include. */ struct variable *djdir = lookup_variable ("DJDIR", 5); if (djdir) { unsigned int len = strlen (djdir->value) + 8; char *defdir = alloca (len + 1); strcat (strcpy (defdir, djdir->value), "/include"); dirs[idx++] = strcache_add (defdir); if (len > max_incl_len) max_incl_len = len; } } #endif for (cpp = default_include_directories; *cpp != 0; ++cpp) { int e; EINTRLOOP (e, stat (*cpp, &stbuf)); if (e == 0 && S_ISDIR (stbuf.st_mode)) { unsigned int len = strlen (*cpp); /* If dir name is written with trailing slashes, discard them. */ while (len > 1 && (*cpp)[len - 1] == '/') --len; if (len > max_incl_len) max_incl_len = len; dirs[idx++] = strcache_add_len (*cpp, len); } } dirs[idx] = 0; /* Now add each dir to the .INCLUDE_DIRS variable. */ for (cpp = dirs; *cpp != 0; ++cpp) do_variable_definition (NILF, ".INCLUDE_DIRS", *cpp, o_default, f_append, 0); include_directories = dirs; } /* Expand ~ or ~USER at the beginning of NAME. Return a newly malloc'd string or 0. */ char * tilde_expand (const char *name) { #ifndef VMS if (name[1] == '/' || name[1] == '\0') { char *home_dir; int is_variable; { /* Turn off --warn-undefined-variables while we expand HOME. */ int save = warn_undefined_variables_flag; warn_undefined_variables_flag = 0; home_dir = allocated_variable_expand ("$(HOME)"); warn_undefined_variables_flag = save; } is_variable = home_dir[0] != '\0'; if (!is_variable) { free (home_dir); home_dir = getenv ("HOME"); } # if !defined(_AMIGA) && !defined(WINDOWS32) if (home_dir == 0 || home_dir[0] == '\0') { char *logname = getlogin (); home_dir = 0; if (logname != 0) { struct passwd *p = getpwnam (logname); if (p != 0) home_dir = p->pw_dir; } } # endif /* !AMIGA && !WINDOWS32 */ if (home_dir != 0) { char *new = xstrdup (concat (2, home_dir, name + 1)); if (is_variable) free (home_dir); return new; } } # if !defined(_AMIGA) && !defined(WINDOWS32) else { struct passwd *pwent; char *userend = strchr (name + 1, '/'); if (userend != 0) *userend = '\0'; pwent = getpwnam (name + 1); if (pwent != 0) { if (userend == 0) return xstrdup (pwent->pw_dir); else return xstrdup (concat (3, pwent->pw_dir, "/", userend + 1)); } else if (userend != 0) *userend = '/'; } # endif /* !AMIGA && !WINDOWS32 */ #endif /* !VMS */ return 0; } /* Parse a string into a sequence of filenames represented as a chain of struct nameseq's and return that chain. Optionally expand the strings via glob(). The string is passed as STRINGP, the address of a string pointer. The string pointer is updated to point at the first character not parsed, which either is a null char or equals STOPCHAR. SIZE is how big to construct chain elements. This is useful if we want them actually to be other structures that have room for additional info. PREFIX, if non-null, is added to the beginning of each filename. FLAGS allows one or more of the following bitflags to be set: PARSEFS_NOSTRIP - Do no strip './'s off the beginning PARSEFS_NOAR - Do not check filenames for archive references PARSEFS_NOGLOB - Do not expand globbing characters PARSEFS_EXISTS - Only return globbed files that actually exist (cannot also set NOGLOB) PARSEFS_NOCACHE - Do not add filenames to the strcache (caller frees) */ void * parse_file_seq (char **stringp, unsigned int size, int stopmap, const char *prefix, int flags) { /* tmp points to tmpbuf after the prefix, if any. tp is the end of the buffer. */ static char *tmpbuf = NULL; int cachep = NONE_SET (flags, PARSEFS_NOCACHE); struct nameseq *new = 0; struct nameseq **newp = &new; #define NEWELT(_n) do { \ const char *__n = (_n); \ *newp = xcalloc (size); \ (*newp)->name = (cachep ? strcache_add (__n) : xstrdup (__n)); \ newp = &(*newp)->next; \ } while(0) char *p; glob_t gl; char *tp; /* Always stop on NUL. */ stopmap |= MAP_NUL; if (size < sizeof (struct nameseq)) size = sizeof (struct nameseq); if (NONE_SET (flags, PARSEFS_NOGLOB)) dir_setup_glob (&gl); /* Get enough temporary space to construct the largest possible target. */ { static int tmpbuf_len = 0; int l = strlen (*stringp) + 1; if (l > tmpbuf_len) { tmpbuf = xrealloc (tmpbuf, l); tmpbuf_len = l; } } tp = tmpbuf; /* Parse STRING. P will always point to the end of the parsed content. */ p = *stringp; while (1) { const char *name; const char **nlist = 0; char *tildep = 0; int globme = 1; #ifndef NO_ARCHIVES char *arname = 0; char *memname = 0; #endif char *s; int nlen; int i; /* Skip whitespace; at the end of the string or STOPCHAR we're done. */ NEXT_TOKEN (p); if (STOP_SET (*p, stopmap)) break; /* There are names left, so find the end of the next name. Throughout this iteration S points to the start. */ s = p; p = find_char_unquote (p, stopmap|MAP_VMSCOMMA|MAP_BLANK); #ifdef VMS /* convert comma separated list to space separated */ if (p && *p == ',') *p =' '; #endif #ifdef _AMIGA if (p && STOP_SET (*p, stopmap & MAP_COLON) && !(ISSPACE (p[1]) || !p[1] || ISSPACE (p[-1]))) p = find_char_unquote (p+1, stopmap|MAP_VMSCOMMA|MAP_BLANK); #endif #ifdef HAVE_DOS_PATHS /* For DOS paths, skip a "C:\..." or a "C:/..." until we find the first colon which isn't followed by a slash or a backslash. Note that tokens separated by spaces should be treated as separate tokens since make doesn't allow path names with spaces */ if (stopmap | MAP_COLON) while (p != 0 && !ISSPACE (*p) && (p[1] == '\\' || p[1] == '/') && isalpha ((unsigned char)p[-1])) p = find_char_unquote (p + 1, stopmap|MAP_VMSCOMMA|MAP_BLANK); #endif if (p == 0) p = s + strlen (s); /* Strip leading "this directory" references. */ if (NONE_SET (flags, PARSEFS_NOSTRIP)) #ifdef VMS /* Skip leading '[]'s. should only be one set or bug somwhere else */ if (p - s > 2 && s[0] == '[' && s[1] == ']') s += 2; /* Skip leading '<>'s. should only be one set or bug somwhere else */ if (p - s > 2 && s[0] == '<' && s[1] == '>') s += 2; #endif /* Skip leading './'s. */ while (p - s > 2 && s[0] == '.' && s[1] == '/') { /* Skip "./" and all following slashes. */ s += 2; while (*s == '/') ++s; } /* Extract the filename just found, and skip it. Set NAME to the string, and NLEN to its length. */ if (s == p) { /* The name was stripped to empty ("./"). */ #if defined(_AMIGA) /* PDS-- This cannot be right!! */ tp[0] = '\0'; nlen = 0; #else tp[0] = '.'; tp[1] = '/'; tp[2] = '\0'; nlen = 2; #endif } else { #ifdef VMS /* VMS filenames can have a ':' in them but they have to be '\'ed but we need * to remove this '\' before we can use the filename. * xstrdup called because S may be read-only string constant. */ char *n = tp; while (s < p) { if (s[0] == '\\' && s[1] == ':') ++s; *(n++) = *(s++); } n[0] = '\0'; nlen = strlen (tp); #else nlen = p - s; memcpy (tp, s, nlen); tp[nlen] = '\0'; #endif } /* At this point, TP points to the element and NLEN is its length. */ #ifndef NO_ARCHIVES /* If this is the start of an archive group that isn't complete, set up to add the archive prefix for future files. A file list like: "libf.a(x.o y.o z.o)" needs to be expanded as: "libf.a(x.o) libf.a(y.o) libf.a(z.o)" TP == TMP means we're not already in an archive group. Ignore something starting with '(', as that cannot actually be an archive-member reference (and treating it as such results in an empty file name, which causes much lossage). Also if it ends in ")" then it's a complete reference so we don't need to treat it specially. Finally, note that archive groups must end with ')' as the last character, so ensure there's some word ending like that before considering this an archive group. */ if (NONE_SET (flags, PARSEFS_NOAR) && tp == tmpbuf && tp[0] != '(' && tp[nlen-1] != ')') { char *n = strchr (tp, '('); if (n) { /* This looks like the first element in an open archive group. A valid group MUST have ')' as the last character. */ const char *e = p; do { const char *o = e; NEXT_TOKEN (e); /* Find the end of this word. We don't want to unquote and we don't care about quoting since we're looking for the last char in the word. */ while (! STOP_SET (*e, stopmap|MAP_BLANK|MAP_VMSCOMMA)) ++e; /* If we didn't move, we're done now. */ if (e == o) break; if (e[-1] == ')') { /* Found the end, so this is the first element in an open archive group. It looks like "lib(mem". Reset TP past the open paren. */ nlen -= (n + 1) - tp; tp = n + 1; /* We can stop looking now. */ break; } } while (*e != '\0'); /* If we have just "lib(", part of something like "lib( a b)", go to the next item. */ if (! nlen) continue; } } /* If we are inside an archive group, make sure it has an end. */ if (tp > tmpbuf) { if (tp[nlen-1] == ')') { /* This is the natural end; reset TP. */ tp = tmpbuf; /* This is just ")", something like "lib(a b )": skip it. */ if (nlen == 1) continue; } else { /* Not the end, so add a "fake" end. */ tp[nlen++] = ')'; tp[nlen] = '\0'; } } #endif /* If we're not globbing we're done: add it to the end of the chain. Go to the next item in the string. */ if (ANY_SET (flags, PARSEFS_NOGLOB)) { NEWELT (concat (2, prefix, tmpbuf)); continue; } /* If we get here we know we're doing glob expansion. TP is a string in tmpbuf. NLEN is no longer used. We may need to do more work: after this NAME will be set. */ name = tmpbuf; /* Expand tilde if applicable. */ if (tmpbuf[0] == '~') { tildep = tilde_expand (tmpbuf); if (tildep != 0) name = tildep; } #ifndef NO_ARCHIVES /* If NAME is an archive member reference replace it with the archive file name, and save the member name in MEMNAME. We will glob on the archive name and then reattach MEMNAME later. */ if (NONE_SET (flags, PARSEFS_NOAR) && ar_name (name)) { ar_parse_name (name, &arname, &memname); name = arname; } #endif /* !NO_ARCHIVES */ /* glob() is expensive: don't call it unless we need to. */ if (NONE_SET (flags, PARSEFS_EXISTS) && strpbrk (name, "?*[") == NULL) { globme = 0; i = 1; nlist = &name; } else switch (glob (name, GLOB_NOSORT|GLOB_ALTDIRFUNC, NULL, &gl)) { case GLOB_NOSPACE: OUT_OF_MEM(); case 0: /* Success. */ i = gl.gl_pathc; nlist = (const char **)gl.gl_pathv; break; case GLOB_NOMATCH: /* If we want only existing items, skip this one. */ if (ANY_SET (flags, PARSEFS_EXISTS)) { i = 0; break; } /* FALLTHROUGH */ default: /* By default keep this name. */ i = 1; nlist = &name; break; } /* For each matched element, add it to the list. */ while (i-- > 0) #ifndef NO_ARCHIVES if (memname != 0) { /* Try to glob on MEMNAME within the archive. */ struct nameseq *found = ar_glob (nlist[i], memname, size); if (! found) /* No matches. Use MEMNAME as-is. */ NEWELT (concat (5, prefix, nlist[i], "(", memname, ")")); else { /* We got a chain of items. Attach them. */ if (*newp) (*newp)->next = found; else *newp = found; /* Find and set the new end. Massage names if necessary. */ while (1) { if (! cachep) found->name = xstrdup (concat (2, prefix, name)); else if (prefix) found->name = strcache_add (concat (2, prefix, name)); if (found->next == 0) break; found = found->next; } newp = &found->next; } } else #endif /* !NO_ARCHIVES */ NEWELT (concat (2, prefix, nlist[i])); if (globme) globfree (&gl); #ifndef NO_ARCHIVES free (arname); #endif free (tildep); } *stringp = p; return new; }