/* $NetBSD: parse.c,v 1.73 2001/09/24 13:22:35 wiz Exp $ */ /* * Copyright (c) 1988, 1989, 1990, 1993 * The Regents of the University of California. All rights reserved. * Copyright (c) 1989 by Berkeley Softworks * All rights reserved. * * This code is derived from software contributed to Berkeley by * Adam de Boor. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. All advertising materials mentioning features or use of this software * must display the following acknowledgement: * This product includes software developed by the University of * California, Berkeley and its contributors. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #ifdef MAKE_BOOTSTRAP static char rcsid[] = "$NetBSD: parse.c,v 1.73 2001/09/24 13:22:35 wiz Exp $"; #else #include #ifndef lint #if 0 static char sccsid[] = "@(#)parse.c 8.3 (Berkeley) 3/19/94"; #else __RCSID("$NetBSD: parse.c,v 1.73 2001/09/24 13:22:35 wiz Exp $"); #endif #endif /* not lint */ #endif /*- * parse.c -- * Functions to parse a makefile. * * One function, Parse_Init, must be called before any functions * in this module are used. After that, the function Parse_File is the * main entry point and controls most of the other functions in this * module. * * Most important structures are kept in Lsts. Directories for * the #include "..." function are kept in the 'parseIncPath' Lst, while * those for the #include <...> are kept in the 'sysIncPath' Lst. The * targets currently being defined are kept in the 'targets' Lst. * * The variables 'fname' and 'lineno' are used to track the name * of the current file and the line number in that file so that error * messages can be more meaningful. * * Interface: * Parse_Init Initialization function which must be * called before anything else in this module * is used. * * Parse_End Cleanup the module * * Parse_File Function used to parse a makefile. It must * be given the name of the file, which should * already have been opened, and a function * to call to read a character from the file. * * Parse_IsVar Returns TRUE if the given line is a * variable assignment. Used by MainParseArgs * to determine if an argument is a target * or a variable assignment. Used internally * for pretty much the same thing... * * Parse_Error Function called when an error occurs in * parsing. Used by the variable and * conditional modules. * Parse_MainName Returns a Lst of the main target to create. */ #ifdef __STDC__ #include #else #include #endif #include #include #include #include "make.h" #include "hash.h" #include "dir.h" #include "job.h" #include "buf.h" #include "pathnames.h" /* * These values are returned by ParseEOF to tell Parse_File whether to * CONTINUE parsing, i.e. it had only reached the end of an include file, * or if it's DONE. */ #define CONTINUE 1 #define DONE 0 static Lst targets; /* targets we're working on */ #ifdef CLEANUP static Lst targCmds; /* command lines for targets */ #endif static Boolean inLine; /* true if currently in a dependency * line or its commands */ typedef struct { char *str; char *ptr; } PTR; static char *fname; /* name of current file (for errors) */ static int lineno; /* line number in current file */ static FILE *curFILE = NULL; /* current makefile */ static PTR *curPTR = NULL; /* current makefile */ static int fatals = 0; static GNode *mainNode; /* The main target to create. This is the * first target on the first dependency * line in the first makefile */ /* * Definitions for handling #include specifications */ typedef struct IFile { char *fname; /* name of previous file */ int lineno; /* saved line number */ FILE * F; /* the open stream */ PTR * p; /* the char pointer */ } IFile; static Lst includes; /* stack of IFiles generated by * #includes */ Lst parseIncPath; /* list of directories for "..." includes */ Lst sysIncPath; /* list of directories for <...> includes */ /*- * specType contains the SPECial TYPE of the current target. It is * Not if the target is unspecial. If it *is* special, however, the children * are linked as children of the parent but not vice versa. This variable is * set in ParseDoDependency */ typedef enum { Begin, /* .BEGIN */ Default, /* .DEFAULT */ End, /* .END */ Ignore, /* .IGNORE */ Includes, /* .INCLUDES */ Interrupt, /* .INTERRUPT */ Libs, /* .LIBS */ MFlags, /* .MFLAGS or .MAKEFLAGS */ Main, /* .MAIN and we don't have anything user-specified to * make */ NoExport, /* .NOEXPORT */ NoPath, /* .NOPATH */ Not, /* Not special */ NotParallel, /* .NOTPARALELL */ Null, /* .NULL */ Order, /* .ORDER */ Parallel, /* .PARALLEL */ ExPath, /* .PATH */ Phony, /* .PHONY */ #ifdef POSIX Posix, /* .POSIX */ #endif Precious, /* .PRECIOUS */ ExShell, /* .SHELL */ Silent, /* .SILENT */ SingleShell, /* .SINGLESHELL */ Suffixes, /* .SUFFIXES */ Wait, /* .WAIT */ Attribute /* Generic attribute */ } ParseSpecial; static ParseSpecial specType; static int waiting; /* * Predecessor node for handling .ORDER. Initialized to NILGNODE when .ORDER * seen, then set to each successive source on the line. */ static GNode *predecessor; /* * The parseKeywords table is searched using binary search when deciding * if a target or source is special. The 'spec' field is the ParseSpecial * type of the keyword ("Not" if the keyword isn't special as a target) while * the 'op' field is the operator to apply to the list of targets if the * keyword is used as a source ("0" if the keyword isn't special as a source) */ static struct { char *name; /* Name of keyword */ ParseSpecial spec; /* Type when used as a target */ int op; /* Operator when used as a source */ } parseKeywords[] = { { ".BEGIN", Begin, 0 }, { ".DEFAULT", Default, 0 }, { ".END", End, 0 }, { ".EXEC", Attribute, OP_EXEC }, { ".IGNORE", Ignore, OP_IGNORE }, { ".INCLUDES", Includes, 0 }, { ".INTERRUPT", Interrupt, 0 }, { ".INVISIBLE", Attribute, OP_INVISIBLE }, { ".JOIN", Attribute, OP_JOIN }, { ".LIBS", Libs, 0 }, { ".MADE", Attribute, OP_MADE }, { ".MAIN", Main, 0 }, { ".MAKE", Attribute, OP_MAKE }, { ".MAKEFLAGS", MFlags, 0 }, { ".MFLAGS", MFlags, 0 }, { ".NOPATH", NoPath, OP_NOPATH }, { ".NOTMAIN", Attribute, OP_NOTMAIN }, { ".NOTPARALLEL", NotParallel, 0 }, { ".NO_PARALLEL", NotParallel, 0 }, { ".NULL", Null, 0 }, { ".OPTIONAL", Attribute, OP_OPTIONAL }, { ".ORDER", Order, 0 }, { ".PARALLEL", Parallel, 0 }, { ".PATH", ExPath, 0 }, { ".PHONY", Phony, OP_PHONY }, #ifdef POSIX { ".POSIX", Posix, 0 }, #endif { ".PRECIOUS", Precious, OP_PRECIOUS }, { ".RECURSIVE", Attribute, OP_MAKE }, { ".SHELL", ExShell, 0 }, { ".SILENT", Silent, OP_SILENT }, { ".SINGLESHELL", SingleShell, 0 }, { ".SUFFIXES", Suffixes, 0 }, { ".USE", Attribute, OP_USE }, { ".USEBEFORE", Attribute, OP_USEBEFORE }, { ".WAIT", Wait, 0 }, }; static void ParseErrorInternal __P((char *, size_t, int, char *, ...)) __attribute__((__format__(__printf__, 4, 5))); static void ParseVErrorInternal __P((char *, size_t, int, char *, va_list)) __attribute__((__format__(__printf__, 4, 0))); static int ParseFindKeyword __P((char *)); static int ParseLinkSrc __P((ClientData, ClientData)); static int ParseDoOp __P((ClientData, ClientData)); static int ParseAddDep __P((ClientData, ClientData)); static void ParseDoSrc __P((int, char *, Lst)); static int ParseFindMain __P((ClientData, ClientData)); static int ParseAddDir __P((ClientData, ClientData)); static int ParseClearPath __P((ClientData, ClientData)); static void ParseDoDependency __P((char *)); static int ParseAddCmd __P((ClientData, ClientData)); static __inline int ParseReadc __P((void)); static void ParseUnreadc __P((int)); static void ParseHasCommands __P((ClientData)); static void ParseDoInclude __P((char *)); static void ParseSetParseFile __P((char *)); #ifdef SYSVINCLUDE static void ParseTraditionalInclude __P((char *)); #endif static int ParseEOF __P((int)); static char *ParseReadLine __P((void)); static char *ParseSkipLine __P((int)); static void ParseFinishLine __P((void)); static void ParseMark __P((GNode *)); extern int maxJobs; /*- *---------------------------------------------------------------------- * ParseFindKeyword -- * Look in the table of keywords for one matching the given string. * * Results: * The index of the keyword, or -1 if it isn't there. * * Side Effects: * None *---------------------------------------------------------------------- */ static int ParseFindKeyword (str) char *str; /* String to find */ { register int start, end, cur; register int diff; start = 0; end = (sizeof(parseKeywords)/sizeof(parseKeywords[0])) - 1; do { cur = start + ((end - start) / 2); diff = strcmp (str, parseKeywords[cur].name); if (diff == 0) { return (cur); } else if (diff < 0) { end = cur - 1; } else { start = cur + 1; } } while (start <= end); return (-1); } /*- * ParseVErrorInternal -- * Error message abort function for parsing. Prints out the context * of the error (line number and file) as well as the message with * two optional arguments. * * Results: * None * * Side Effects: * "fatals" is incremented if the level is PARSE_FATAL. */ /* VARARGS */ static void #ifdef __STDC__ ParseVErrorInternal(char *cfname, size_t clineno, int type, char *fmt, va_list ap) #else ParseVErrorInternal(va_alist) va_dcl #endif { static Boolean fatal_warning_error_printed = FALSE; (void)fprintf(stderr, "%s: \"", progname); if (*cfname != '/') { char *cp, *dir; /* * Nothing is more anoying than not knowing which Makefile * is the culprit. */ dir = Var_Value(".PARSEDIR", VAR_GLOBAL, &cp); if (dir == NULL || *dir == '\0' || (*dir == '.' && dir[1] == '\0')) dir = Var_Value(".CURDIR", VAR_GLOBAL, &cp); if (dir == NULL) dir = "."; (void)fprintf(stderr, "%s/%s", dir, cfname); } else (void)fprintf(stderr, "%s", cfname); (void)fprintf(stderr, "\" line %d: ", (int)clineno); if (type == PARSE_WARNING) (void)fprintf(stderr, "warning: "); (void)vfprintf(stderr, fmt, ap); (void)fprintf(stderr, "\n"); (void)fflush(stderr); if (type == PARSE_FATAL || parseWarnFatal) fatals += 1; if (parseWarnFatal && !fatal_warning_error_printed) { Error("parsing warnings being treated as errors"); fatal_warning_error_printed = TRUE; } } /*- * ParseErrorInternal -- * Error function * * Results: * None * * Side Effects: * None */ /* VARARGS */ static void #ifdef __STDC__ ParseErrorInternal(char *cfname, size_t clineno, int type, char *fmt, ...) #else ParseErrorInternal(va_alist) va_dcl #endif { va_list ap; #ifdef __STDC__ va_start(ap, fmt); #else int type; /* Error type (PARSE_WARNING, PARSE_FATAL) */ char *fmt; char *cfname; size_t clineno; va_start(ap); cfname = va_arg(ap, char *); clineno = va_arg(ap, size_t); type = va_arg(ap, int); fmt = va_arg(ap, char *); #endif ParseVErrorInternal(cfname, clineno, type, fmt, ap); va_end(ap); } /*- * Parse_Error -- * External interface to ParseErrorInternal; uses the default filename * Line number. * * Results: * None * * Side Effects: * None */ /* VARARGS */ void #ifdef __STDC__ Parse_Error(int type, char *fmt, ...) #else Parse_Error(va_alist) va_dcl #endif { va_list ap; #ifdef __STDC__ va_start(ap, fmt); #else int type; /* Error type (PARSE_WARNING, PARSE_FATAL) */ char *fmt; va_start(ap); type = va_arg(ap, int); fmt = va_arg(ap, char *); #endif ParseVErrorInternal(fname, lineno, type, fmt, ap); va_end(ap); } /*- *--------------------------------------------------------------------- * ParseLinkSrc -- * Link the parent node to its new child. Used in a Lst_ForEach by * ParseDoDependency. If the specType isn't 'Not', the parent * isn't linked as a parent of the child. * * Results: * Always = 0 * * Side Effects: * New elements are added to the parents list of cgn and the * children list of cgn. the unmade field of pgn is updated * to reflect the additional child. *--------------------------------------------------------------------- */ static int ParseLinkSrc (pgnp, cgnp) ClientData pgnp; /* The parent node */ ClientData cgnp; /* The child node */ { GNode *pgn = (GNode *) pgnp; GNode *cgn = (GNode *) cgnp; if ((pgn->type & OP_DOUBLEDEP) && !Lst_IsEmpty (pgn->cohorts)) pgn = (GNode *) Lst_Datum (Lst_Last (pgn->cohorts)); (void)Lst_AtEnd (pgn->children, (ClientData)cgn); if (specType == Not) (void)Lst_AtEnd (cgn->parents, (ClientData)pgn); pgn->unmade += 1; return (0); } /*- *--------------------------------------------------------------------- * ParseDoOp -- * Apply the parsed operator to the given target node. Used in a * Lst_ForEach call by ParseDoDependency once all targets have * been found and their operator parsed. If the previous and new * operators are incompatible, a major error is taken. * * Results: * Always 0 * * Side Effects: * The type field of the node is altered to reflect any new bits in * the op. *--------------------------------------------------------------------- */ static int ParseDoOp (gnp, opp) ClientData gnp; /* The node to which the operator is to be * applied */ ClientData opp; /* The operator to apply */ { GNode *gn = (GNode *) gnp; int op = *(int *) opp; /* * If the dependency mask of the operator and the node don't match and * the node has actually had an operator applied to it before, and * the operator actually has some dependency information in it, complain. */ if (((op & OP_OPMASK) != (gn->type & OP_OPMASK)) && !OP_NOP(gn->type) && !OP_NOP(op)) { Parse_Error (PARSE_FATAL, "Inconsistent operator for %s", gn->name); return (1); } if ((op == OP_DOUBLEDEP) && ((gn->type & OP_OPMASK) == OP_DOUBLEDEP)) { /* * If the node was the object of a :: operator, we need to create a * new instance of it for the children and commands on this dependency * line. The new instance is placed on the 'cohorts' list of the * initial one (note the initial one is not on its own cohorts list) * and the new instance is linked to all parents of the initial * instance. */ register GNode *cohort; /* * Propagate copied bits to the initial node. They'll be propagated * back to the rest of the cohorts later. */ gn->type |= op & ~OP_OPMASK; cohort = Targ_NewGN(gn->name); /* * Make the cohort invisible as well to avoid duplicating it into * other variables. True, parents of this target won't tend to do * anything with their local variables, but better safe than * sorry. (I think this is pointless now, since the relevant list * traversals will no longer see this node anyway. -mycroft) */ cohort->type = op | OP_INVISIBLE; (void)Lst_AtEnd(gn->cohorts, (ClientData)cohort); } else { /* * We don't want to nuke any previous flags (whatever they were) so we * just OR the new operator into the old */ gn->type |= op; } return (0); } /*- *--------------------------------------------------------------------- * ParseAddDep -- * Check if the pair of GNodes given needs to be synchronized. * This has to be when two nodes are on different sides of a * .WAIT directive. * * Results: * Returns 1 if the two targets need to be ordered, 0 otherwise. * If it returns 1, the search can stop * * Side Effects: * A dependency can be added between the two nodes. * *--------------------------------------------------------------------- */ static int ParseAddDep(pp, sp) ClientData pp; ClientData sp; { GNode *p = (GNode *) pp; GNode *s = (GNode *) sp; if (p->order < s->order) { /* * XXX: This can cause loops, and loops can cause unmade targets, * but checking is tedious, and the debugging output can show the * problem */ (void)Lst_AtEnd(p->successors, (ClientData)s); (void)Lst_AtEnd(s->preds, (ClientData)p); return 0; } else return 1; } /*- *--------------------------------------------------------------------- * ParseDoSrc -- * Given the name of a source, figure out if it is an attribute * and apply it to the targets if it is. Else decide if there is * some attribute which should be applied *to* the source because * of some special target and apply it if so. Otherwise, make the * source be a child of the targets in the list 'targets' * * Results: * None * * Side Effects: * Operator bits may be added to the list of targets or to the source. * The targets may have a new source added to their lists of children. *--------------------------------------------------------------------- */ static void ParseDoSrc (tOp, src, allsrc) int tOp; /* operator (if any) from special targets */ char *src; /* name of the source to handle */ Lst allsrc; /* List of all sources to wait for */ { GNode *gn = NULL; if (*src == '.' && isupper ((unsigned char)src[1])) { int keywd = ParseFindKeyword(src); if (keywd != -1) { int op = parseKeywords[keywd].op; if (op != 0) { Lst_ForEach (targets, ParseDoOp, (ClientData)&op); return; } if (parseKeywords[keywd].spec == Wait) { waiting++; return; } } } switch (specType) { case Main: /* * If we have noted the existence of a .MAIN, it means we need * to add the sources of said target to the list of things * to create. The string 'src' is likely to be free, so we * must make a new copy of it. Note that this will only be * invoked if the user didn't specify a target on the command * line. This is to allow #ifmake's to succeed, or something... */ (void) Lst_AtEnd (create, (ClientData)estrdup(src)); /* * Add the name to the .TARGETS variable as well, so the user cna * employ that, if desired. */ Var_Append(".TARGETS", src, VAR_GLOBAL); return; case Order: /* * Create proper predecessor/successor links between the previous * source and the current one. */ gn = Targ_FindNode(src, TARG_CREATE); if (predecessor != NILGNODE) { (void)Lst_AtEnd(predecessor->successors, (ClientData)gn); (void)Lst_AtEnd(gn->preds, (ClientData)predecessor); } /* * The current source now becomes the predecessor for the next one. */ predecessor = gn; break; default: /* * If the source is not an attribute, we need to find/create * a node for it. After that we can apply any operator to it * from a special target or link it to its parents, as * appropriate. * * In the case of a source that was the object of a :: operator, * the attribute is applied to all of its instances (as kept in * the 'cohorts' list of the node) or all the cohorts are linked * to all the targets. */ gn = Targ_FindNode (src, TARG_CREATE); if (tOp) { gn->type |= tOp; } else { Lst_ForEach (targets, ParseLinkSrc, (ClientData)gn); } break; } gn->order = waiting; (void)Lst_AtEnd(allsrc, (ClientData)gn); if (waiting) { Lst_ForEach(allsrc, ParseAddDep, (ClientData)gn); } } /*- *----------------------------------------------------------------------- * ParseFindMain -- * Find a real target in the list and set it to be the main one. * Called by ParseDoDependency when a main target hasn't been found * yet. * * Results: * 0 if main not found yet, 1 if it is. * * Side Effects: * mainNode is changed and Targ_SetMain is called. * *----------------------------------------------------------------------- */ static int ParseFindMain(gnp, dummy) ClientData gnp; /* Node to examine */ ClientData dummy; { GNode *gn = (GNode *) gnp; if ((gn->type & OP_NOTARGET) == 0) { mainNode = gn; Targ_SetMain(gn); return (dummy ? 1 : 1); } else { return (dummy ? 0 : 0); } } /*- *----------------------------------------------------------------------- * ParseAddDir -- * Front-end for Dir_AddDir to make sure Lst_ForEach keeps going * * Results: * === 0 * * Side Effects: * See Dir_AddDir. * *----------------------------------------------------------------------- */ static int ParseAddDir(path, name) ClientData path; ClientData name; { (void) Dir_AddDir((Lst) path, (char *) name); return(0); } /*- *----------------------------------------------------------------------- * ParseClearPath -- * Front-end for Dir_ClearPath to make sure Lst_ForEach keeps going * * Results: * === 0 * * Side Effects: * See Dir_ClearPath * *----------------------------------------------------------------------- */ static int ParseClearPath(path, dummy) ClientData path; ClientData dummy; { Dir_ClearPath((Lst) path); return(dummy ? 0 : 0); } /*- *--------------------------------------------------------------------- * ParseDoDependency -- * Parse the dependency line in line. * * Results: * None * * Side Effects: * The nodes of the sources are linked as children to the nodes of the * targets. Some nodes may be created. * * We parse a dependency line by first extracting words from the line and * finding nodes in the list of all targets with that name. This is done * until a character is encountered which is an operator character. Currently * these are only ! and :. At this point the operator is parsed and the * pointer into the line advanced until the first source is encountered. * The parsed operator is applied to each node in the 'targets' list, * which is where the nodes found for the targets are kept, by means of * the ParseDoOp function. * The sources are read in much the same way as the targets were except * that now they are expanded using the wildcarding scheme of the C-Shell * and all instances of the resulting words in the list of all targets * are found. Each of the resulting nodes is then linked to each of the * targets as one of its children. * Certain targets are handled specially. These are the ones detailed * by the specType variable. * The storing of transformation rules is also taken care of here. * A target is recognized as a transformation rule by calling * Suff_IsTransform. If it is a transformation rule, its node is gotten * from the suffix module via Suff_AddTransform rather than the standard * Targ_FindNode in the target module. *--------------------------------------------------------------------- */ static void ParseDoDependency (line) char *line; /* the line to parse */ { char *cp; /* our current position */ GNode *gn; /* a general purpose temporary node */ int op; /* the operator on the line */ char savec; /* a place to save a character */ Lst paths; /* List of search paths to alter when parsing * a list of .PATH targets */ int tOp; /* operator from special target */ Lst sources; /* list of archive source names after * expansion */ Lst curTargs; /* list of target names to be found and added * to the targets list */ Lst curSrcs; /* list of sources in order */ tOp = 0; specType = Not; waiting = 0; paths = (Lst)NULL; curTargs = Lst_Init(FALSE); curSrcs = Lst_Init(FALSE); do { for (cp = line; *cp && !isspace ((unsigned char)*cp) && (*cp != '!') && (*cp != ':') && (*cp != '('); cp++) { if (*cp == '$') { /* * Must be a dynamic source (would have been expanded * otherwise), so call the Var module to parse the puppy * so we can safely advance beyond it...There should be * no errors in this, as they would have been discovered * in the initial Var_Subst and we wouldn't be here. */ int length; Boolean freeIt; char *result; result=Var_Parse(cp, VAR_CMD, TRUE, &length, &freeIt); if (freeIt) { free(result); } cp += length-1; } continue; } if (*cp == '(') { /* * Archives must be handled specially to make sure the OP_ARCHV * flag is set in their 'type' field, for one thing, and because * things like "archive(file1.o file2.o file3.o)" are permissible. * Arch_ParseArchive will set 'line' to be the first non-blank * after the archive-spec. It creates/finds nodes for the members * and places them on the given list, returning SUCCESS if all * went well and FAILURE if there was an error in the * specification. On error, line should remain untouched. */ if (Arch_ParseArchive (&line, targets, VAR_CMD) != SUCCESS) { Parse_Error (PARSE_FATAL, "Error in archive specification: \"%s\"", line); return; } else { continue; } } savec = *cp; if (!*cp) { /* * Ending a dependency line without an operator is a Bozo * no-no. As a heuristic, this is also often triggered by * undetected conflicts from cvs/rcs merges. */ if ((strncmp(line, "<<<<<<", 6) == 0) || (strncmp(line, "======", 6) == 0) || (strncmp(line, ">>>>>>", 6) == 0)) Parse_Error (PARSE_FATAL, "Makefile appears to contain unresolved cvs/rcs/??? merge conflicts"); else Parse_Error (PARSE_FATAL, "Need an operator"); return; } *cp = '\0'; /* * Have a word in line. See if it's a special target and set * specType to match it. */ if (*line == '.' && isupper ((unsigned char)line[1])) { /* * See if the target is a special target that must have it * or its sources handled specially. */ int keywd = ParseFindKeyword(line); if (keywd != -1) { if (specType == ExPath && parseKeywords[keywd].spec != ExPath) { Parse_Error(PARSE_FATAL, "Mismatched special targets"); return; } specType = parseKeywords[keywd].spec; tOp = parseKeywords[keywd].op; /* * Certain special targets have special semantics: * .PATH Have to set the dirSearchPath * variable too * .MAIN Its sources are only used if * nothing has been specified to * create. * .DEFAULT Need to create a node to hang * commands on, but we don't want * it in the graph, nor do we want * it to be the Main Target, so we * create it, set OP_NOTMAIN and * add it to the list, setting * DEFAULT to the new node for * later use. We claim the node is * A transformation rule to make * life easier later, when we'll * use Make_HandleUse to actually * apply the .DEFAULT commands. * .PHONY The list of targets * .NOPATH Don't search for file in the path * .BEGIN * .END * .INTERRUPT Are not to be considered the * main target. * .NOTPARALLEL Make only one target at a time. * .SINGLESHELL Create a shell for each command. * .ORDER Must set initial predecessor to NIL */ switch (specType) { case ExPath: if (paths == NULL) { paths = Lst_Init(FALSE); } (void)Lst_AtEnd(paths, (ClientData)dirSearchPath); break; case Main: if (!Lst_IsEmpty(create)) { specType = Not; } break; case Begin: case End: case Interrupt: gn = Targ_FindNode(line, TARG_CREATE); gn->type |= OP_NOTMAIN; (void)Lst_AtEnd(targets, (ClientData)gn); break; case Default: gn = Targ_NewGN(".DEFAULT"); gn->type |= (OP_NOTMAIN|OP_TRANSFORM); (void)Lst_AtEnd(targets, (ClientData)gn); DEFAULT = gn; break; case NotParallel: { maxJobs = 1; break; } case SingleShell: compatMake = TRUE; break; case Order: predecessor = NILGNODE; break; default: break; } } else if (strncmp (line, ".PATH", 5) == 0) { /* * .PATH has to be handled specially. * Call on the suffix module to give us a path to * modify. */ Lst path; specType = ExPath; path = Suff_GetPath (&line[5]); if (path == NILLST) { Parse_Error (PARSE_FATAL, "Suffix '%s' not defined (yet)", &line[5]); return; } else { if (paths == (Lst)NULL) { paths = Lst_Init(FALSE); } (void)Lst_AtEnd(paths, (ClientData)path); } } } /* * Have word in line. Get or create its node and stick it at * the end of the targets list */ if ((specType == Not) && (*line != '\0')) { if (Dir_HasWildcards(line)) { /* * Targets are to be sought only in the current directory, * so create an empty path for the thing. Note we need to * use Dir_Destroy in the destruction of the path as the * Dir module could have added a directory to the path... */ Lst emptyPath = Lst_Init(FALSE); Dir_Expand(line, emptyPath, curTargs); Lst_Destroy(emptyPath, Dir_Destroy); } else { /* * No wildcards, but we want to avoid code duplication, * so create a list with the word on it. */ (void)Lst_AtEnd(curTargs, (ClientData)line); } while(!Lst_IsEmpty(curTargs)) { char *targName = (char *)Lst_DeQueue(curTargs); if (!Suff_IsTransform (targName)) { gn = Targ_FindNode (targName, TARG_CREATE); } else { gn = Suff_AddTransform (targName); } (void)Lst_AtEnd (targets, (ClientData)gn); } } else if (specType == ExPath && *line != '.' && *line != '\0') { Parse_Error(PARSE_WARNING, "Extra target (%s) ignored", line); } *cp = savec; /* * If it is a special type and not .PATH, it's the only target we * allow on this line... */ if (specType != Not && specType != ExPath) { Boolean warn = FALSE; while ((*cp != '!') && (*cp != ':') && *cp) { if (*cp != ' ' && *cp != '\t') { warn = TRUE; } cp++; } if (warn) { Parse_Error(PARSE_WARNING, "Extra target ignored"); } } else { while (*cp && isspace ((unsigned char)*cp)) { cp++; } } line = cp; } while ((*line != '!') && (*line != ':') && *line); /* * Don't need the list of target names anymore... */ Lst_Destroy(curTargs, NOFREE); if (!Lst_IsEmpty(targets)) { switch(specType) { default: Parse_Error(PARSE_WARNING, "Special and mundane targets don't mix. Mundane ones ignored"); break; case Default: case Begin: case End: case Interrupt: /* * These four create nodes on which to hang commands, so * targets shouldn't be empty... */ case Not: /* * Nothing special here -- targets can be empty if it wants. */ break; } } /* * Have now parsed all the target names. Must parse the operator next. The * result is left in op . */ if (*cp == '!') { op = OP_FORCE; } else if (*cp == ':') { if (cp[1] == ':') { op = OP_DOUBLEDEP; cp++; } else { op = OP_DEPENDS; } } else { Parse_Error (PARSE_FATAL, "Missing dependency operator"); return; } cp++; /* Advance beyond operator */ Lst_ForEach (targets, ParseDoOp, (ClientData)&op); /* * Get to the first source */ while (*cp && isspace ((unsigned char)*cp)) { cp++; } line = cp; /* * Several special targets take different actions if present with no * sources: * a .SUFFIXES line with no sources clears out all old suffixes * a .PRECIOUS line makes all targets precious * a .IGNORE line ignores errors for all targets * a .SILENT line creates silence when making all targets * a .PATH removes all directories from the search path(s). */ if (!*line) { switch (specType) { case Suffixes: Suff_ClearSuffixes (); break; case Precious: allPrecious = TRUE; break; case Ignore: ignoreErrors = TRUE; break; case Silent: beSilent = TRUE; break; case ExPath: Lst_ForEach(paths, ParseClearPath, (ClientData)NULL); break; #ifdef POSIX case Posix: Var_Set("%POSIX", "1003.2", VAR_GLOBAL, 0); break; #endif default: break; } } else if (specType == MFlags) { /* * Call on functions in main.c to deal with these arguments and * set the initial character to a null-character so the loop to * get sources won't get anything */ Main_ParseArgLine (line); *line = '\0'; } else if (specType == ExShell) { if (Job_ParseShell (line) != SUCCESS) { Parse_Error (PARSE_FATAL, "improper shell specification"); return; } *line = '\0'; } else if ((specType == NotParallel) || (specType == SingleShell)) { *line = '\0'; } /* * NOW GO FOR THE SOURCES */ if ((specType == Suffixes) || (specType == ExPath) || (specType == Includes) || (specType == Libs) || (specType == Null)) { while (*line) { /* * If the target was one that doesn't take files as its sources * but takes something like suffixes, we take each * space-separated word on the line as a something and deal * with it accordingly. * * If the target was .SUFFIXES, we take each source as a * suffix and add it to the list of suffixes maintained by the * Suff module. * * If the target was a .PATH, we add the source as a directory * to search on the search path. * * If it was .INCLUDES, the source is taken to be the suffix of * files which will be #included and whose search path should * be present in the .INCLUDES variable. * * If it was .LIBS, the source is taken to be the suffix of * files which are considered libraries and whose search path * should be present in the .LIBS variable. * * If it was .NULL, the source is the suffix to use when a file * has no valid suffix. */ char savec; while (*cp && !isspace ((unsigned char)*cp)) { cp++; } savec = *cp; *cp = '\0'; switch (specType) { case Suffixes: Suff_AddSuffix (line, &mainNode); break; case ExPath: Lst_ForEach(paths, ParseAddDir, (ClientData)line); break; case Includes: Suff_AddInclude (line); break; case Libs: Suff_AddLib (line); break; case Null: Suff_SetNull (line); break; default: break; } *cp = savec; if (savec != '\0') { cp++; } while (*cp && isspace ((unsigned char)*cp)) { cp++; } line = cp; } if (paths) { Lst_Destroy(paths, NOFREE); } } else { while (*line) { /* * The targets take real sources, so we must beware of archive * specifications (i.e. things with left parentheses in them) * and handle them accordingly. */ while (*cp && !isspace ((unsigned char)*cp)) { if ((*cp == '(') && (cp > line) && (cp[-1] != '$')) { /* * Only stop for a left parenthesis if it isn't at the * start of a word (that'll be for variable changes * later) and isn't preceded by a dollar sign (a dynamic * source). */ break; } else { cp++; } } if (*cp == '(') { GNode *gn; sources = Lst_Init (FALSE); if (Arch_ParseArchive (&line, sources, VAR_CMD) != SUCCESS) { Parse_Error (PARSE_FATAL, "Error in source archive spec \"%s\"", line); return; } while (!Lst_IsEmpty (sources)) { gn = (GNode *) Lst_DeQueue (sources); ParseDoSrc (tOp, gn->name, curSrcs); } Lst_Destroy (sources, NOFREE); cp = line; } else { if (*cp) { *cp = '\0'; cp += 1; } ParseDoSrc (tOp, line, curSrcs); } while (*cp && isspace ((unsigned char)*cp)) { cp++; } line = cp; } } if (mainNode == NILGNODE) { /* * If we have yet to decide on a main target to make, in the * absence of any user input, we want the first target on * the first dependency line that is actually a real target * (i.e. isn't a .USE or .EXEC rule) to be made. */ Lst_ForEach (targets, ParseFindMain, (ClientData)0); } /* * Finally, destroy the list of sources */ Lst_Destroy(curSrcs, NOFREE); } /*- *--------------------------------------------------------------------- * Parse_IsVar -- * Return TRUE if the passed line is a variable assignment. A variable * assignment consists of a single word followed by optional whitespace * followed by either a += or an = operator. * This function is used both by the Parse_File function and main when * parsing the command-line arguments. * * Results: * TRUE if it is. FALSE if it ain't * * Side Effects: * none *--------------------------------------------------------------------- */ Boolean Parse_IsVar (line) register char *line; /* the line to check */ { register Boolean wasSpace = FALSE; /* set TRUE if found a space */ register Boolean haveName = FALSE; /* Set TRUE if have a variable name */ int level = 0; #define ISEQOPERATOR(c) \ (((c) == '+') || ((c) == ':') || ((c) == '?') || ((c) == '!')) /* * Skip to variable name */ for (;(*line == ' ') || (*line == '\t'); line++) continue; for (; *line != '=' || level != 0; line++) switch (*line) { case '\0': /* * end-of-line -- can't be a variable assignment. */ return FALSE; case ' ': case '\t': /* * there can be as much white space as desired so long as there is * only one word before the operator */ wasSpace = TRUE; break; case '(': case '{': level++; break; case '}': case ')': level--; break; default: if (wasSpace && haveName) { if (ISEQOPERATOR(*line)) { /* * We must have a finished word */ if (level != 0) return FALSE; /* * When an = operator [+?!:] is found, the next * character must be an = or it ain't a valid * assignment. */ if (line[1] == '=') return haveName; #ifdef SUNSHCMD /* * This is a shell command */ if (strncmp(line, ":sh", 3) == 0) return haveName; #endif } /* * This is the start of another word, so not assignment. */ return FALSE; } else { haveName = TRUE; wasSpace = FALSE; } break; } return haveName; } /*- *--------------------------------------------------------------------- * Parse_DoVar -- * Take the variable assignment in the passed line and do it in the * global context. * * Note: There is a lexical ambiguity with assignment modifier characters * in variable names. This routine interprets the character before the = * as a modifier. Therefore, an assignment like * C++=/usr/bin/CC * is interpreted as "C+ +=" instead of "C++ =". * * Results: * none * * Side Effects: * the variable structure of the given variable name is altered in the * global context. *--------------------------------------------------------------------- */ void Parse_DoVar (line, ctxt) char *line; /* a line guaranteed to be a variable * assignment. This reduces error checks */ GNode *ctxt; /* Context in which to do the assignment */ { char *cp; /* pointer into line */ enum { VAR_SUBST, VAR_APPEND, VAR_SHELL, VAR_NORMAL } type; /* Type of assignment */ char *opc; /* ptr to operator character to * null-terminate the variable name */ /* * Avoid clobbered variable warnings by forcing the compiler * to ``unregister'' variables */ #if __GNUC__ (void) &cp; (void) &line; #endif /* * Skip to variable name */ while ((*line == ' ') || (*line == '\t')) { line++; } /* * Skip to operator character, nulling out whitespace as we go */ for (cp = line + 1; *cp != '='; cp++) { if (isspace ((unsigned char)*cp)) { *cp = '\0'; } } opc = cp-1; /* operator is the previous character */ *cp++ = '\0'; /* nuke the = */ /* * Check operator type */ switch (*opc) { case '+': type = VAR_APPEND; *opc = '\0'; break; case '?': /* * If the variable already has a value, we don't do anything. */ *opc = '\0'; if (Var_Exists(line, ctxt)) { return; } else { type = VAR_NORMAL; } break; case ':': type = VAR_SUBST; *opc = '\0'; break; case '!': type = VAR_SHELL; *opc = '\0'; break; default: #ifdef SUNSHCMD while (opc > line && *opc != ':') opc--; if (strncmp(opc, ":sh", 3) == 0) { type = VAR_SHELL; *opc = '\0'; break; } #endif type = VAR_NORMAL; break; } while (isspace ((unsigned char)*cp)) { cp++; } if (type == VAR_APPEND) { Var_Append (line, cp, ctxt); } else if (type == VAR_SUBST) { /* * Allow variables in the old value to be undefined, but leave their * invocation alone -- this is done by forcing oldVars to be false. * XXX: This can cause recursive variables, but that's not hard to do, * and this allows someone to do something like * * CFLAGS = $(.INCLUDES) * CFLAGS := -I.. $(CFLAGS) * * And not get an error. */ Boolean oldOldVars = oldVars; oldVars = FALSE; /* * make sure that we set the variable the first time to nothing * so that it gets substituted! */ if (!Var_Exists(line, ctxt)) Var_Set(line, "", ctxt, 0); cp = Var_Subst(NULL, cp, ctxt, FALSE); oldVars = oldOldVars; Var_Set(line, cp, ctxt, 0); free(cp); } else if (type == VAR_SHELL) { Boolean freeCmd = FALSE; /* TRUE if the command needs to be freed, i.e. * if any variable expansion was performed */ char *res, *err; if (strchr(cp, '$') != NULL) { /* * There's a dollar sign in the command, so perform variable * expansion on the whole thing. The resulting string will need * freeing when we're done, so set freeCmd to TRUE. */ cp = Var_Subst(NULL, cp, VAR_CMD, TRUE); freeCmd = TRUE; } res = Cmd_Exec(cp, &err); Var_Set(line, res, ctxt, 0); free(res); if (err) Parse_Error(PARSE_WARNING, err, cp); if (freeCmd) free(cp); } else { /* * Normal assignment -- just do it. */ Var_Set(line, cp, ctxt, 0); } if (strcmp(line, MAKEOVERRIDES) == 0) ExportMAKEFLAGS(0); /* re-export MAKEFLAGS */ } /*- * ParseAddCmd -- * Lst_ForEach function to add a command line to all targets * * Results: * Always 0 * * Side Effects: * A new element is added to the commands list of the node. */ static int ParseAddCmd(gnp, cmd) ClientData gnp; /* the node to which the command is to be added */ ClientData cmd; /* the command to add */ { GNode *gn = (GNode *) gnp; /* if target already supplied, ignore commands */ if ((gn->type & OP_DOUBLEDEP) && !Lst_IsEmpty (gn->cohorts)) gn = (GNode *) Lst_Datum (Lst_Last (gn->cohorts)); if (!(gn->type & OP_HAS_COMMANDS)) { (void)Lst_AtEnd(gn->commands, cmd); ParseMark(gn); } else { #ifdef notyet /* XXX: We cannot do this until we fix the tree */ (void)Lst_AtEnd(gn->commands, cmd); Parse_Error (PARSE_WARNING, "overriding commands for target \"%s\"; " "previous commands defined at %s: %d ignored", gn->name, gn->fname, gn->lineno); #else Parse_Error (PARSE_WARNING, "duplicate script for target \"%s\" ignored", gn->name); ParseErrorInternal (gn->fname, gn->lineno, PARSE_WARNING, "using previous script for \"%s\" defined here", gn->name); #endif } return(0); } /*- *----------------------------------------------------------------------- * ParseHasCommands -- * Callback procedure for Parse_File when destroying the list of * targets on the last dependency line. Marks a target as already * having commands if it does, to keep from having shell commands * on multiple dependency lines. * * Results: * None * * Side Effects: * OP_HAS_COMMANDS may be set for the target. * *----------------------------------------------------------------------- */ static void ParseHasCommands(gnp) ClientData gnp; /* Node to examine */ { GNode *gn = (GNode *) gnp; if (!Lst_IsEmpty(gn->commands)) { gn->type |= OP_HAS_COMMANDS; } } /*- *----------------------------------------------------------------------- * Parse_AddIncludeDir -- * Add a directory to the path searched for included makefiles * bracketed by double-quotes. Used by functions in main.c * * Results: * None. * * Side Effects: * The directory is appended to the list. * *----------------------------------------------------------------------- */ void Parse_AddIncludeDir (dir) char *dir; /* The name of the directory to add */ { (void) Dir_AddDir (parseIncPath, dir); } /*- *--------------------------------------------------------------------- * ParseDoInclude -- * Push to another file. * * The input is the line minus the `.'. A file spec is a string * enclosed in <> or "". The former is looked for only in sysIncPath. * The latter in . and the directories specified by -I command line * options * * Results: * None * * Side Effects: * A structure is added to the includes Lst and readProc, lineno, * fname and curFILE are altered for the new file *--------------------------------------------------------------------- */ static void ParseDoInclude (line) char *line; { char *fullname; /* full pathname of file */ IFile *oldFile; /* state associated with current file */ char endc; /* the character which ends the file spec */ char *cp; /* current position in file spec */ Boolean isSystem; /* TRUE if makefile is a system makefile */ int silent = (*line != 'i') ? 1 : 0; char *file = &line[7 + silent]; /* * Skip to delimiter character so we know where to look */ while ((*file == ' ') || (*file == '\t')) { file++; } if ((*file != '"') && (*file != '<')) { Parse_Error (PARSE_FATAL, ".include filename must be delimited by '\"' or '<'"); return; } /* * Set the search path on which to find the include file based on the * characters which bracket its name. Angle-brackets imply it's * a system Makefile while double-quotes imply it's a user makefile */ if (*file == '<') { isSystem = TRUE; endc = '>'; } else { isSystem = FALSE; endc = '"'; } /* * Skip to matching delimiter */ for (cp = ++file; *cp && *cp != endc; cp++) { continue; } if (*cp != endc) { Parse_Error (PARSE_FATAL, "Unclosed %cinclude filename. '%c' expected", '.', endc); return; } *cp = '\0'; /* * Substitute for any variables in the file name before trying to * find the thing. */ file = Var_Subst (NULL, file, VAR_CMD, FALSE); /* * Now we know the file's name and its search path, we attempt to * find the durn thing. A return of NULL indicates the file don't * exist. */ if (!isSystem) { /* * Include files contained in double-quotes are first searched for * relative to the including file's location. We don't want to * cd there, of course, so we just tack on the old file's * leading path components and call Dir_FindFile to see if * we can locate the beast. */ char *prefEnd, *Fname; /* Make a temporary copy of this, to be safe. */ Fname = estrdup(fname); prefEnd = strrchr (Fname, '/'); if (prefEnd != (char *)NULL) { char *newName; *prefEnd = '\0'; if (file[0] == '/') newName = estrdup(file); else newName = str_concat (Fname, file, STR_ADDSLASH); fullname = Dir_FindFile (newName, parseIncPath); if (fullname == (char *)NULL) { fullname = Dir_FindFile(newName, dirSearchPath); } free (newName); *prefEnd = '/'; } else { fullname = (char *)NULL; } free (Fname); } else { fullname = (char *)NULL; } if (fullname == (char *)NULL) { /* * System makefile or makefile wasn't found in same directory as * included makefile. Search for it first on the -I search path, * then on the .PATH search path, if not found in a -I directory. * XXX: Suffix specific? */ fullname = Dir_FindFile (file, parseIncPath); if (fullname == (char *)NULL) { fullname = Dir_FindFile(file, dirSearchPath); } } if (fullname == (char *)NULL) { /* * Still haven't found the makefile. Look for it on the system * path as a last resort. */ fullname = Dir_FindFile(file, sysIncPath); } if (fullname == (char *) NULL) { *cp = endc; if (!silent) Parse_Error (PARSE_FATAL, "Could not find %s", file); return; } free(file); /* * Once we find the absolute path to the file, we get to save all the * state from the current file before we can start reading this * include file. The state is stored in an IFile structure which * is placed on a list with other IFile structures. The list makes * a very nice stack to track how we got here... */ oldFile = (IFile *) emalloc (sizeof (IFile)); oldFile->fname = fname; oldFile->F = curFILE; oldFile->p = curPTR; oldFile->lineno = lineno; (void) Lst_AtFront (includes, (ClientData)oldFile); /* * Once the previous state has been saved, we can get down to reading * the new file. We set up the name of the file to be the absolute * name of the include file so error messages refer to the right * place. Naturally enough, we start reading at line number 0. */ fname = fullname; lineno = 0; ParseSetParseFile(fname); curFILE = fopen (fullname, "r"); curPTR = NULL; if (curFILE == (FILE * ) NULL) { if (!silent) Parse_Error (PARSE_FATAL, "Cannot open %s", fullname); /* * Pop to previous file */ (void) ParseEOF(0); } } /*- *--------------------------------------------------------------------- * ParseSetParseFile -- * Set the .PARSEDIR and .PARSEFILE variables to the dirname and * basename of the given filename * * Results: * None * * Side Effects: * The .PARSEDIR and .PARSEFILE variables are overwritten by the * dirname and basename of the given filename. *--------------------------------------------------------------------- */ static void ParseSetParseFile(fname) char *fname; { char *slash; slash = strrchr(fname, '/'); if (slash == 0) { Var_Set(".PARSEDIR", ".", VAR_GLOBAL, 0); Var_Set(".PARSEFILE", fname, VAR_GLOBAL, 0); } else { *slash = '\0'; Var_Set(".PARSEDIR", fname, VAR_GLOBAL, 0); Var_Set(".PARSEFILE", slash+1, VAR_GLOBAL, 0); *slash = '/'; } } /*- *--------------------------------------------------------------------- * Parse_FromString -- * Start Parsing from the given string * * Results: * None * * Side Effects: * A structure is added to the includes Lst and readProc, lineno, * fname and curFILE are altered for the new file *--------------------------------------------------------------------- */ void Parse_FromString(str) char *str; { IFile *oldFile; /* state associated with this file */ if (DEBUG(FOR)) (void) fprintf(stderr, "%s\n----\n", str); oldFile = (IFile *) emalloc (sizeof (IFile)); oldFile->lineno = lineno; oldFile->fname = fname; oldFile->F = curFILE; oldFile->p = curPTR; (void) Lst_AtFront (includes, (ClientData)oldFile); curFILE = NULL; curPTR = (PTR *) emalloc (sizeof (PTR)); curPTR->str = curPTR->ptr = str; lineno = 0; fname = estrdup(fname); } #ifdef SYSVINCLUDE /*- *--------------------------------------------------------------------- * ParseTraditionalInclude -- * Push to another file. * * The input is the current line. The file name(s) are * following the "include". * * Results: * None * * Side Effects: * A structure is added to the includes Lst and readProc, lineno, * fname and curFILE are altered for the new file *--------------------------------------------------------------------- */ static void ParseTraditionalInclude (line) char *line; { char *fullname; /* full pathname of file */ IFile *oldFile; /* state associated with current file */ char *cp; /* current position in file spec */ char *prefEnd; int done = 0; int silent = (line[0] != 'i') ? 1 : 0; char *file = &line[silent + 7]; char *cfname = fname; size_t clineno = lineno; /* * Skip over whitespace */ while (isspace((unsigned char)*file)) file++; if (*file == '\0') { Parse_Error (PARSE_FATAL, "Filename missing from \"include\""); return; } for (; !done; file = cp + 1) { /* * Skip to end of line or next whitespace */ for (cp = file; *cp && !isspace((unsigned char) *cp); cp++) continue; if (*cp) *cp = '\0'; else done = 1; /* * Substitute for any variables in the file name before trying to * find the thing. */ file = Var_Subst(NULL, file, VAR_CMD, FALSE); /* * Now we know the file's name, we attempt to find the durn thing. * A return of NULL indicates the file don't exist. * * Include files are first searched for relative to the including * file's location. We don't want to cd there, of course, so we * just tack on the old file's leading path components and call * Dir_FindFile to see if we can locate the beast. * XXX - this *does* search in the current directory, right? */ prefEnd = strrchr(cfname, '/'); if (prefEnd != NULL) { char *newName; *prefEnd = '\0'; newName = str_concat(cfname, file, STR_ADDSLASH); fullname = Dir_FindFile(newName, parseIncPath); if (fullname == NULL) { fullname = Dir_FindFile(newName, dirSearchPath); } free (newName); *prefEnd = '/'; } else { fullname = NULL; } if (fullname == NULL) { /* * System makefile or makefile wasn't found in same directory as * included makefile. Search for it first on the -I search path, * then on the .PATH search path, if not found in a * -I directory. XXX: Suffix specific? */ fullname = Dir_FindFile(file, parseIncPath); if (fullname == NULL) { fullname = Dir_FindFile(file, dirSearchPath); } } if (fullname == NULL) { /* * Still haven't found the makefile. Look for it on the system * path as a last resort. */ fullname = Dir_FindFile(file, sysIncPath); } if (fullname == NULL) { if (!silent) ParseErrorInternal(cfname, clineno, PARSE_FATAL, "Could not find %s", file); free(file); continue; } free(file); /* * Once we find the absolute path to the file, we get to save all * the state from the current file before we can start reading this * include file. The state is stored in an IFile structure which * is placed on a list with other IFile structures. The list makes * a very nice stack to track how we got here... */ oldFile = (IFile *) emalloc(sizeof(IFile)); oldFile->fname = fname; oldFile->F = curFILE; oldFile->p = curPTR; oldFile->lineno = lineno; (void) Lst_AtFront(includes, (ClientData)oldFile); /* * Once the previous state has been saved, we can get down to * reading the new file. We set up the name of the file to be the * absolute name of the include file so error messages refer to the * right place. Naturally enough, we start reading at line number 0. */ fname = fullname; lineno = 0; curFILE = fopen(fullname, "r"); curPTR = NULL; if (curFILE == NULL) { if (!silent) ParseErrorInternal(cfname, clineno, PARSE_FATAL, "Cannot open %s", fullname); /* * Pop to previous file */ (void) ParseEOF(1); } } } #endif /*- *--------------------------------------------------------------------- * ParseEOF -- * Called when EOF is reached in the current file. If we were reading * an include file, the includes stack is popped and things set up * to go back to reading the previous file at the previous location. * * Results: * CONTINUE if there's more to do. DONE if not. * * Side Effects: * The old curFILE, is closed. The includes list is shortened. * lineno, curFILE, and fname are changed if CONTINUE is returned. *--------------------------------------------------------------------- */ static int ParseEOF (opened) int opened; { IFile *ifile; /* the state on the top of the includes stack */ if (Lst_IsEmpty (includes)) { Var_Delete(".PARSEDIR", VAR_GLOBAL); Var_Delete(".PARSEFILE", VAR_GLOBAL); return (DONE); } ifile = (IFile *) Lst_DeQueue (includes); free ((Address) fname); fname = ifile->fname; lineno = ifile->lineno; if (opened && curFILE) (void) fclose (curFILE); if (curPTR) { free((Address) curPTR->str); free((Address) curPTR); } curFILE = ifile->F; curPTR = ifile->p; free ((Address)ifile); /* pop the PARSEDIR/PARSEFILE variables */ ParseSetParseFile(fname); return (CONTINUE); } /*- *--------------------------------------------------------------------- * ParseReadc -- * Read a character from the current file * * Results: * The character that was read * * Side Effects: *--------------------------------------------------------------------- */ static __inline int ParseReadc() { if (curFILE) return fgetc(curFILE); if (curPTR && *curPTR->ptr) return *curPTR->ptr++; return EOF; } /*- *--------------------------------------------------------------------- * ParseUnreadc -- * Put back a character to the current file * * Results: * None. * * Side Effects: *--------------------------------------------------------------------- */ static void ParseUnreadc(c) int c; { if (curFILE) { ungetc(c, curFILE); return; } if (curPTR) { *--(curPTR->ptr) = c; return; } } /* ParseSkipLine(): * Grab the next line */ static char * ParseSkipLine(skip) int skip; /* Skip lines that don't start with . */ { char *line; int c, lastc, lineLength = 0; Buffer buf; buf = Buf_Init(MAKE_BSIZE); do { Buf_Discard(buf, lineLength); lastc = '\0'; while (((c = ParseReadc()) != '\n' || lastc == '\\') && c != EOF) { if (c == '\n') { Buf_ReplaceLastByte(buf, (Byte)' '); lineno++; while ((c = ParseReadc()) == ' ' || c == '\t'); if (c == EOF) break; } Buf_AddByte(buf, (Byte)c); lastc = c; } if (c == EOF) { Parse_Error(PARSE_FATAL, "Unclosed conditional/for loop"); Buf_Destroy(buf, TRUE); return((char *)NULL); } lineno++; Buf_AddByte(buf, (Byte)'\0'); line = (char *)Buf_GetAll(buf, &lineLength); } while (skip == 1 && line[0] != '.'); Buf_Destroy(buf, FALSE); return line; } /*- *--------------------------------------------------------------------- * ParseReadLine -- * Read an entire line from the input file. Called only by Parse_File. * To facilitate escaped newlines and what have you, a character is * buffered in 'lastc', which is '\0' when no characters have been * read. When we break out of the loop, c holds the terminating * character and lastc holds a character that should be added to * the line (unless we don't read anything but a terminator). * * Results: * A line w/o its newline * * Side Effects: * Only those associated with reading a character *--------------------------------------------------------------------- */ static char * ParseReadLine () { Buffer buf; /* Buffer for current line */ register int c; /* the current character */ register int lastc; /* The most-recent character */ Boolean semiNL; /* treat semi-colons as newlines */ Boolean ignDepOp; /* TRUE if should ignore dependency operators * for the purposes of setting semiNL */ Boolean ignComment; /* TRUE if should ignore comments (in a * shell command */ char *line; /* Result */ char *ep; /* to strip trailing blanks */ int lineLength; /* Length of result */ semiNL = FALSE; ignDepOp = FALSE; ignComment = FALSE; /* * Handle special-characters at the beginning of the line. Either a * leading tab (shell command) or pound-sign (possible conditional) * forces us to ignore comments and dependency operators and treat * semi-colons as semi-colons (by leaving semiNL FALSE). This also * discards completely blank lines. */ for (;;) { c = ParseReadc(); if (c == '\t') { ignComment = ignDepOp = TRUE; break; } else if (c == '\n') { lineno++; } else if (c == '#') { ParseUnreadc(c); break; } else { /* * Anything else breaks out without doing anything */ break; } } if (c != EOF) { lastc = c; buf = Buf_Init(MAKE_BSIZE); while (((c = ParseReadc ()) != '\n' || (lastc == '\\')) && (c != EOF)) { test_char: switch(c) { case '\n': /* * Escaped newline: read characters until a non-space or an * unescaped newline and replace them all by a single space. * This is done by storing the space over the backslash and * dropping through with the next nonspace. If it is a * semi-colon and semiNL is TRUE, it will be recognized as a * newline in the code below this... */ lineno++; lastc = ' '; while ((c = ParseReadc ()) == ' ' || c == '\t') { continue; } if (c == EOF || c == '\n') { goto line_read; } else { /* * Check for comments, semiNL's, etc. -- easier than * ParseUnreadc(c); continue; */ goto test_char; } /*NOTREACHED*/ break; case ';': /* * Semi-colon: Need to see if it should be interpreted as a * newline */ if (semiNL) { /* * To make sure the command that may be following this * semi-colon begins with a tab, we push one back into the * input stream. This will overwrite the semi-colon in the * buffer. If there is no command following, this does no * harm, since the newline remains in the buffer and the * whole line is ignored. */ ParseUnreadc('\t'); goto line_read; } break; case '=': if (!semiNL) { /* * Haven't seen a dependency operator before this, so this * must be a variable assignment -- don't pay attention to * dependency operators after this. */ ignDepOp = TRUE; } else if (lastc == ':' || lastc == '!') { /* * Well, we've seen a dependency operator already, but it * was the previous character, so this is really just an * expanded variable assignment. Revert semi-colons to * being just semi-colons again and ignore any more * dependency operators. * * XXX: Note that a line like "foo : a:=b" will blow up, * but who'd write a line like that anyway? */ ignDepOp = TRUE; semiNL = FALSE; } break; case '#': if (!ignComment) { if ( #if 0 compatMake && #endif (lastc != '\\')) { /* * If the character is a hash mark and it isn't escaped * (or we're being compatible), the thing is a comment. * Skip to the end of the line. */ do { c = ParseReadc(); } while ((c != '\n') && (c != EOF)); goto line_read; } else { /* * Don't add the backslash. Just let the # get copied * over. */ lastc = c; continue; } } break; case ':': case '!': if (!ignDepOp && (c == ':' || c == '!')) { /* * A semi-colon is recognized as a newline only on * dependency lines. Dependency lines are lines with a * colon or an exclamation point. Ergo... */ semiNL = TRUE; } break; } /* * Copy in the previous character and save this one in lastc. */ Buf_AddByte (buf, (Byte)lastc); lastc = c; } line_read: lineno++; if (lastc != '\0') { Buf_AddByte (buf, (Byte)lastc); } Buf_AddByte (buf, (Byte)'\0'); line = (char *)Buf_GetAll (buf, &lineLength); Buf_Destroy (buf, FALSE); /* * Strip trailing blanks and tabs from the line. * Do not strip a blank or tab that is preceded by * a '\' */ ep = line; while (*ep) ++ep; while (ep > line + 1 && (ep[-1] == ' ' || ep[-1] == '\t')) { if (ep > line + 1 && ep[-2] == '\\') break; --ep; } *ep = 0; if (line[0] == '.') { /* * The line might be a conditional. Ask the conditional module * about it and act accordingly */ switch (Cond_Eval (line)) { case COND_SKIP: /* * Skip to next conditional that evaluates to COND_PARSE. */ do { free (line); line = ParseSkipLine(1); } while (line && Cond_Eval(line) != COND_PARSE); if (line == NULL) break; /*FALLTHRU*/ case COND_PARSE: free ((Address) line); line = ParseReadLine(); break; case COND_INVALID: if (For_Eval(line)) { int ok; free(line); do { /* * Skip after the matching end */ line = ParseSkipLine(0); if (line == NULL) { Parse_Error (PARSE_FATAL, "Unexpected end of file in for loop.\n"); break; } ok = For_Eval(line); free(line); } while (ok); if (line != NULL) For_Run(); line = ParseReadLine(); } break; } } return (line); } else { /* * Hit end-of-file, so return a NULL line to indicate this. */ return((char *)NULL); } } /*- *----------------------------------------------------------------------- * ParseFinishLine -- * Handle the end of a dependency group. * * Results: * Nothing. * * Side Effects: * inLine set FALSE. 'targets' list destroyed. * *----------------------------------------------------------------------- */ static void ParseFinishLine() { if (inLine) { Lst_ForEach(targets, Suff_EndTransform, (ClientData)NULL); Lst_Destroy (targets, ParseHasCommands); targets = NULL; inLine = FALSE; } } /*- *--------------------------------------------------------------------- * Parse_File -- * Parse a file into its component parts, incorporating it into the * current dependency graph. This is the main function and controls * almost every other function in this module * * Results: * None * * Side Effects: * Loads. Nodes are added to the list of all targets, nodes and links * are added to the dependency graph. etc. etc. etc. *--------------------------------------------------------------------- */ void Parse_File(name, stream) char *name; /* the name of the file being read */ FILE * stream; /* Stream open to makefile to parse */ { register char *cp, /* pointer into the line */ *line; /* the line we're working on */ inLine = FALSE; fname = name; curFILE = stream; lineno = 0; fatals = 0; ParseSetParseFile(fname); do { while ((line = ParseReadLine ()) != NULL) { if (*line == '.') { /* * Lines that begin with the special character are either * include or undef directives. */ for (cp = line + 1; isspace ((unsigned char)*cp); cp++) { continue; } if (strncmp(cp, "include", 7) == 0 || ((cp[0] == 's' || cp[0] == '-') && strncmp(&cp[1], "include", 7) == 0)) { ParseDoInclude (cp); goto nextLine; } else if (strncmp(cp, "undef", 5) == 0) { char *cp2; for (cp += 5; isspace((unsigned char) *cp); cp++) { continue; } for (cp2 = cp; !isspace((unsigned char) *cp2) && (*cp2 != '\0'); cp2++) { continue; } *cp2 = '\0'; Var_Delete(cp, VAR_GLOBAL); goto nextLine; } } if (*line == '#') { /* If we're this far, the line must be a comment. */ goto nextLine; } if (*line == '\t') { /* * If a line starts with a tab, it can only hope to be * a creation command. */ #ifndef POSIX shellCommand: #endif for (cp = line + 1; isspace ((unsigned char)*cp); cp++) { continue; } if (*cp) { if (inLine) { /* * So long as it's not a blank line and we're actually * in a dependency spec, add the command to the list of * commands of all targets in the dependency spec */ Lst_ForEach (targets, ParseAddCmd, cp); #ifdef CLEANUP Lst_AtEnd(targCmds, (ClientData) line); #endif continue; } else { Parse_Error (PARSE_FATAL, "Unassociated shell command \"%s\"", cp); } } #ifdef SYSVINCLUDE } else if (((strncmp(line, "include", 7) == 0 && isspace((unsigned char) line[7])) || ((line[0] == 's' || line[0] == '-') && strncmp(&line[1], "include", 7) == 0 && isspace((unsigned char) line[8]))) && strchr(line, ':') == NULL) { /* * It's an S3/S5-style "include". */ ParseTraditionalInclude (line); goto nextLine; #endif } else if (Parse_IsVar (line)) { ParseFinishLine(); Parse_DoVar (line, VAR_GLOBAL); } else { /* * We now know it's a dependency line so it needs to have all * variables expanded before being parsed. Tell the variable * module to complain if some variable is undefined... * To make life easier on novices, if the line is indented we * first make sure the line has a dependency operator in it. * If it doesn't have an operator and we're in a dependency * line's script, we assume it's actually a shell command * and add it to the current list of targets. */ #ifndef POSIX Boolean nonSpace = FALSE; #endif cp = line; if (isspace((unsigned char) line[0])) { while ((*cp != '\0') && isspace((unsigned char) *cp)) { cp++; } if (*cp == '\0') { goto nextLine; } #ifndef POSIX while ((*cp != ':') && (*cp != '!') && (*cp != '\0')) { nonSpace = TRUE; cp++; } #endif } #ifndef POSIX if (*cp == '\0') { if (inLine) { Parse_Error (PARSE_WARNING, "Shell command needs a leading tab"); goto shellCommand; } else if (nonSpace) { Parse_Error (PARSE_FATAL, "Missing operator"); } } else { #endif ParseFinishLine(); cp = Var_Subst (NULL, line, VAR_CMD, TRUE); free (line); line = cp; /* * Need a non-circular list for the target nodes */ if (targets) Lst_Destroy(targets, NOFREE); targets = Lst_Init (FALSE); inLine = TRUE; ParseDoDependency (line); #ifndef POSIX } #endif } nextLine: free (line); } /* * Reached EOF, but it may be just EOF of an include file... */ } while (ParseEOF(1) == CONTINUE); /* * Make sure conditionals are clean */ Cond_End(); if (fatals) { (void)fprintf(stderr, "%s: Fatal errors encountered -- cannot continue\n", progname); PrintOnError(NULL); exit (1); } } /*- *--------------------------------------------------------------------- * Parse_Init -- * initialize the parsing module * * Results: * none * * Side Effects: * the parseIncPath list is initialized... *--------------------------------------------------------------------- */ void Parse_Init () { mainNode = NILGNODE; parseIncPath = Lst_Init (FALSE); sysIncPath = Lst_Init (FALSE); includes = Lst_Init (FALSE); #ifdef CLEANUP targCmds = Lst_Init (FALSE); #endif } void Parse_End() { #ifdef CLEANUP Lst_Destroy(targCmds, (void (*) __P((ClientData))) free); if (targets) Lst_Destroy(targets, NOFREE); Lst_Destroy(sysIncPath, Dir_Destroy); Lst_Destroy(parseIncPath, Dir_Destroy); Lst_Destroy(includes, NOFREE); /* Should be empty now */ #endif } /*- *----------------------------------------------------------------------- * Parse_MainName -- * Return a Lst of the main target to create for main()'s sake. If * no such target exists, we Punt with an obnoxious error message. * * Results: * A Lst of the single node to create. * * Side Effects: * None. * *----------------------------------------------------------------------- */ Lst Parse_MainName() { Lst mainList; /* result list */ mainList = Lst_Init (FALSE); if (mainNode == NILGNODE) { Punt ("no target to make."); /*NOTREACHED*/ } else if (mainNode->type & OP_DOUBLEDEP) { (void) Lst_AtEnd (mainList, (ClientData)mainNode); Lst_Concat(mainList, mainNode->cohorts, LST_CONCNEW); } else (void) Lst_AtEnd (mainList, (ClientData)mainNode); return (mainList); } /*- *----------------------------------------------------------------------- * ParseMark -- * Add the filename and lineno to the GNode so that we remember * where it was first defined. * * Side Effects: * None. * *----------------------------------------------------------------------- */ static void ParseMark(gn) GNode *gn; { gn->fname = strdup(fname); gn->lineno = lineno; }