/**
* Alters the current line number and/or file name. You may wish to
* use this directive if you extract definition source from other files.
* @command{getdefs} uses this mechanism so AutoGen will report the correct
* file and approximate line number of any errors found in extracted
* definitions.
*/
char *
doDir_line(directive_enum_t id, char const * dir, char * scan_next)
{
(void)id;
/*
* The sequence must be: #line <number> "file-name-string"
*
* Start by scanning up to and extracting the line number.
*/
dir = SPN_WHITESPACE_CHARS(dir);
if (! IS_DEC_DIGIT_CHAR(*dir))
return scan_next;
cctx->scx_line = (int)strtol(dir, (char **)&dir, 0);
/*
* Now extract the quoted file name string.
* We dup the string so it won't disappear on us.
*/
dir = SPN_WHITESPACE_CHARS(dir);
if (*(dir++) != '"')
return scan_next;
{
char * pz = strchr(dir, '"');
if (pz == NULL)
return scan_next;
*pz = NUL;
}
AGDUPSTR(cctx->scx_fname, dir, "#line");
return scan_next;
}
/**
* Configure a dependency option.
* Handles any of these letters: MFQTPGD as the first part of the option
* argument.
*
* @param opts the autogen options data structure
* @param pOptDesc the option descriptor for this option.
*/
LOCAL void
config_dep(tOptions * opts, tOptDesc * od)
{
char const * opt_arg = od->optArg.argString;
(void)opts;
/*
* The option argument is optional. Make sure we have one.
*/
if (opt_arg == NULL)
return;
while (*opt_arg == 'M') opt_arg++;
opt_arg = SPN_WHITESPACE_CHARS(opt_arg);
switch (*opt_arg) {
case 'F':
if (dep_file != NULL)
dep_usage(CFGDEP_DUP_TARGET_MSG);
opt_arg = SPN_WHITESPACE_CHARS(opt_arg + 1);
AGDUPSTR(dep_file, opt_arg, "f name");
break;
case 'Q':
case 'T':
{
bool quote_it = (*opt_arg == 'Q');
if (dep_target != NULL)
dep_usage(CFGDEP_DUP_TARGET_MSG);
opt_arg = SPN_WHITESPACE_CHARS(opt_arg + 1);
if (quote_it)
dep_target = make_quote_str(opt_arg);
else
AGDUPSTR(dep_target, opt_arg, "t name");
break;
}
case 'P':
dep_phonies = true;
break;
case 'D':
case 'G':
case NUL:
/*
* 'D' and 'G' make sense to GCC, not us. Ignore 'em. If we
* found a NUL byte, then act like we found -MM on the command line.
*/
break;
default:
dep_usage(CFGDEP_UNKNOWN_DEP_FMT, opt_arg);
}
}
/**
* Parse the option usage flags string. Any parsing problems yield
* a zero (no flags set) result. This function is internal to
* set_usage_flags().
*
* @param[in] fnt Flag Name Table - maps a name to a mask
* @param[in] txt the text to process. If NULL, then
* getenv("AUTOOPTS_USAGE") is used.
* @returns a bit mask indicating which \a fnt entries were found.
*/
static unsigned int
parse_usage_flags(ao_flag_names_t const * fnt, char const * txt)
{
unsigned int res = 0;
/*
* The text may be passed in. If not, use the environment variable.
*/
if (txt == NULL) {
txt = getenv("AUTOOPTS_USAGE");
if (txt == NULL)
return 0;
}
txt = SPN_WHITESPACE_CHARS(txt);
if (*txt == NUL)
return 0;
/*
* search the string for table entries. We must understand everything
* we see in the string, or we give up on it.
*/
for (;;) {
int ix = 0;
for (;;) {
if (strneqvcmp(txt, fnt[ix].fnm_name, (int)fnt[ix].fnm_len) == 0)
break;
if (++ix >= AOUF_COUNT)
return 0;
}
/*
* Make sure we have a full match. Look for whitespace,
* a comma, or a NUL byte.
*/
if (! IS_END_LIST_ENTRY_CHAR(txt[fnt[ix].fnm_len]))
return 0;
res |= 1U << ix;
txt = SPN_WHITESPACE_CHARS(txt + fnt[ix].fnm_len);
switch (*txt) {
case NUL:
return res;
case ',':
txt = SPN_WHITESPACE_CHARS(txt + 1);
/* Something must follow the comma */
default:
continue;
}
}
}
/**
* We've found a closing '>' without a preceding '/', thus we must search
* the text for '<name/>' where "name" is the name of the XML element.
*
* @param[in] name the start of the name in the element header
* @param[in] nm_len the length of that name
* @param[out] len the length of the value (string between header and
* the trailer/tail.
* @returns the character after the trailer, or NULL if not found.
*/
static char const *
find_end_xml(char const * src, size_t nm_len, char const * val, size_t * len)
{
char z[72] = "</";
char * dst = z + 2;
do {
*(dst++) = *(src++);
} while (--nm_len > 0); /* nm_len is known to be 64 or less */
*(dst++) = '>';
*dst = NUL;
{
char const * res = strstr(val, z);
if (res != NULL) {
char const * end = (option_load_mode != OPTION_LOAD_KEEP)
? SPN_WHITESPACE_BACK(val, res)
: res;
*len = (size_t)(end - val); /* includes trailing white space */
res = SPN_WHITESPACE_CHARS(res + (dst - z));
}
return res;
}
}
/**
* Scan off the xml element name, and the rest of the header, too.
* Set the value type to NONE if it ends with "/>".
*
* @param[in] name the first name character (alphabetic)
* @param[out] nm_len the length of the name
* @param[out] val set valType field to STRING or NONE.
*
* @returns the scan resumption point, or NULL on error
*/
static char const *
scan_xml_name(char const * name, size_t * nm_len, tOptionValue * val)
{
char const * scan = SPN_VALUE_NAME_CHARS(name + 1);
*nm_len = (size_t)(scan - name);
if (*nm_len > 64)
return NULL;
val->valType = OPARG_TYPE_STRING;
if (IS_WHITESPACE_CHAR(*scan)) {
/*
* There are attributes following the name. Parse 'em.
*/
scan = SPN_WHITESPACE_CHARS(scan);
scan = parse_attrs(NULL, scan, &option_load_mode, val);
if (scan == NULL)
return NULL; /* oops */
}
if (! IS_END_XML_TOKEN_CHAR(*scan))
return NULL; /* oops */
if (*scan == '/') {
/*
* Single element XML entries get inserted as an empty string.
*/
if (*++scan != '>')
return NULL;
val->valType = OPARG_TYPE_NONE;
}
return scan+1;
}
static void
check_assert_str(char const * pz, char const * arg)
{
pz = SPN_WHITESPACE_CHARS(pz);
if (IS_FALSE_TYPE_CHAR(*pz))
AG_ABEND(aprf(DIRECT_ASSERT_FMT, pz, arg));
}
/**
* Check membership start conditions. An equal character (@samp{=}) says to
* clear the result and not carry over any residual value. A carat
* (@samp{^}), which may follow the equal character, says to invert the
* result. The scanning pointer is advanced past these characters and any
* leading white space. Invalid sequences are indicated by setting the
* scanning pointer to NULL.
*
* @param od the set membership option description
* @param argp a pointer to the string scanning pointer
* @param invert a pointer to the boolean inversion indicator
*
* @returns either zero or the original value for the optCookie.
*/
static uintptr_t
check_membership_start(tOptDesc * od, char const ** argp, bool * invert)
{
uintptr_t res = (uintptr_t)od->optCookie;
char const * arg = SPN_WHITESPACE_CHARS(od->optArg.argString);
if ((arg == NULL) || (*arg == NUL))
goto member_start_fail;
*invert = false;
switch (*arg) {
case '=':
res = 0UL;
arg = SPN_WHITESPACE_CHARS(arg + 1);
switch (*arg) {
case '=': case ',':
goto member_start_fail;
case '^':
goto inversion;
default:
break;
}
break;
case '^':
inversion:
*invert = true;
arg = SPN_WHITESPACE_CHARS(arg + 1);
if (*arg != ',')
break;
/* FALLTHROUGH */
case ',':
goto member_start_fail;
default:
break;
}
*argp = arg;
return res;
member_start_fail:
*argp = NULL;
return 0UL;
}
/**
* This directive will pass the option name and associated text to the
* AutoOpts optionLoadLine routine (@pxref{libopts-optionLoadLine}). The
* option text may span multiple lines by continuing them with a backslash.
* The backslash/newline pair will be replaced with two space characters.
* This directive may be used to set a search path for locating template files
* For example, this:
*
* @example
* #option templ-dirs $ENVVAR/dirname
* @end example
* @noindent
* will direct autogen to use the @code{ENVVAR} environment variable to find
* a directory named @code{dirname} that (may) contain templates. Since these
* directories are searched in most recently supplied first order, search
* directories supplied in this way will be searched before any supplied on
* the command line.
*/
char *
doDir_option(directive_enum_t id, char const * dir, char * scan_next)
{
dir = SPN_WHITESPACE_CHARS(dir);
optionLoadLine(&autogenOptions, dir);
(void)id;
return scan_next;
}
/**
* Will remove any entries from the define list
* that match the undef name pattern.
*/
char *
doDir_undef(directive_enum_t id, char const * dir, char * scan_next)
{
dir = SPN_WHITESPACE_CHARS(dir);
SET_OPT_UNDEFINE(dir);
(void)id;
return scan_next;
}
static token_list_t *
alloc_token_list(char const * str)
{
token_list_t * res;
int max_token_ct = 2; /* allow for trailing NULL pointer & NUL on string */
if (str == NULL) goto enoent_res;
/*
* Trim leading white space. Use "ENOENT" and a NULL return to indicate
* an empty string was passed.
*/
str = SPN_WHITESPACE_CHARS(str);
if (*str == NUL) goto enoent_res;
/*
* Take an approximate count of tokens. If no quoted strings are used,
* it will be accurate. If quoted strings are used, it will be a little
* high and we'll squander the space for a few extra pointers.
*/
{
char const * pz = str;
do {
max_token_ct++;
pz = BRK_WHITESPACE_CHARS(pz+1);
pz = SPN_WHITESPACE_CHARS(pz);
} while (*pz != NUL);
res = malloc(sizeof(*res) + (size_t)(pz - str)
+ ((size_t)max_token_ct * sizeof(ch_t*)));
}
if (res == NULL)
errno = ENOMEM;
else res->tkn_list[0] = (ch_t*)(res->tkn_list + (max_token_ct - 1));
return res;
enoent_res:
errno = ENOENT;
return NULL;
}
/**
* This directive will cause AutoGen to stop processing
* and exit with a status of EXIT_FAILURE.
*/
char *
doDir_error(directive_enum_t id, char const * arg, char * scan_next)
{
arg = SPN_WHITESPACE_CHARS(arg);
AG_ABEND(aprf(DIRECT_ERROR_FMT, cctx->scx_fname, cctx->scx_line, arg));
/* NOTREACHED */
(void)scan_next;
(void)id;
return NULL;
}
/**
* The definitions that follow, up to the matching @code{#endif} will be
* processed only if there is a corresponding @code{-Dname} command line
* option or if a @code{#define} of that name has been previously encountered.
*/
char *
doDir_ifdef(directive_enum_t id, char const * dir, char * scan_next)
{
char const * defstr = get_define_str(SPN_WHITESPACE_CHARS(dir), false);
bool ok = (id == DIR_IFDEF) ? (defstr != NULL) : (defstr == NULL);
if (! ok)
return skip_to_else_end(scan_next);
ifdef_lvl++;
return scan_next;
}
/**
* Print the usage information for a single vendor option.
*
* @param[in] opts the program option descriptor
* @param[in] od the option descriptor
* @param[in] argtp names of the option argument types
* @param[in] usefmt format for primary usage line
*/
static void
prt_one_vendor(tOptions * opts, tOptDesc * od,
arg_types_t * argtp, char const * usefmt)
{
prt_preamble(opts, od, argtp);
{
char z[ 80 ];
char const * pzArgType;
/*
* Determine the argument type string first on its usage, then,
* when the option argument is required, base the type string on the
* argument type.
*/
if (od->fOptState & OPTST_ARG_OPTIONAL) {
pzArgType = argtp->pzOpt;
} else switch (OPTST_GET_ARGTYPE(od->fOptState)) {
case OPARG_TYPE_NONE: pzArgType = argtp->pzNo; break;
case OPARG_TYPE_ENUMERATION: pzArgType = argtp->pzKey; break;
case OPARG_TYPE_FILE: pzArgType = argtp->pzFile; break;
case OPARG_TYPE_MEMBERSHIP: pzArgType = argtp->pzKeyL; break;
case OPARG_TYPE_BOOLEAN: pzArgType = argtp->pzBool; break;
case OPARG_TYPE_NUMERIC: pzArgType = argtp->pzNum; break;
case OPARG_TYPE_HIERARCHY: pzArgType = argtp->pzNest; break;
case OPARG_TYPE_STRING: pzArgType = argtp->pzStr; break;
case OPARG_TYPE_TIME: pzArgType = argtp->pzTime; break;
default: goto bogus_desc;
}
pzArgType = SPN_WHITESPACE_CHARS(pzArgType);
if (*pzArgType == NUL)
snprintf(z, sizeof(z), "%s", od->pz_Name);
else
snprintf(z, sizeof(z), "%s=%s", od->pz_Name, pzArgType);
fprintf(option_usage_fp, usefmt, z, od->pzText);
switch (OPTST_GET_ARGTYPE(od->fOptState)) {
case OPARG_TYPE_ENUMERATION:
case OPARG_TYPE_MEMBERSHIP:
displayEnum = (od->pOptProc != NULL) ? true : displayEnum;
}
}
return;
bogus_desc:
fprintf(stderr, zbad_od, opts->pzProgName, od->pz_Name);
ao_bug(zbad_arg_type_msg);
}
static char *
assemble_arg_val(char * txt, tOptionLoadMode mode)
{
char* pzEnd = strpbrk(txt, ARG_BREAK_STR);
int space_break;
/*
* Not having an argument to a configurable name is okay.
*/
if (pzEnd == NULL)
return txt + strlen(txt);
/*
* If we are keeping all whitespace, then the modevalue starts with the
* character that follows the end of the configurable name, regardless
* of which character caused it.
*/
if (mode == OPTION_LOAD_KEEP) {
*(pzEnd++) = NUL;
return pzEnd;
}
/*
* If the name ended on a white space character, remember that
* because we'll have to skip over an immediately following ':' or '='
* (and the white space following *that*).
*/
space_break = IS_WHITESPACE_CHAR(*pzEnd);
*(pzEnd++) = NUL;
pzEnd = SPN_WHITESPACE_CHARS(pzEnd);
if (space_break && ((*pzEnd == ':') || (*pzEnd == '=')))
pzEnd = SPN_WHITESPACE_CHARS(pzEnd+1);
return pzEnd;
}
static char *
next_directive(char * scan)
{
if (*scan == '#')
scan++;
else {
char * pz = strstr(scan, DIRECT_CK_LIST_MARK);
if (pz == NULL)
AG_ABEND(aprf(DIRECT_NOENDIF_FMT, cctx->scx_fname,
cctx->scx_line));
scan = pz + 2;
}
return SPN_WHITESPACE_CHARS(scan);
}
/* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
*
* processDirective
*
* THIS IS THE ONLY EXTERNAL ENTRY POINT
*
* A directive character has been found.
* Decide what to do and return a pointer to the character
* where scanning is to resume.
*/
LOCAL char *
processDirective(char * scan)
{
char * eodir = end_of_directive(scan);
/*
* Ignore '#!' as a comment, enabling a definition file to behave
* as a script that gets interpreted by autogen. :-)
*/
if (*scan == '!')
return eodir;
scan = SPN_WHITESPACE_CHARS(scan);
if (! IS_ALPHABETIC_CHAR(*scan))
return doDir_invalid(DIR_INVALID, scan, eodir);
return doDir_directive_disp(scan, eodir);
}
/**
* handle AutoOpts mode flags.
*
* @param[in,out] opts program option descriptor
* @param[in] txt scanning pointer
* @returns the next character to look at
*/
static char *
aoflags_directive(tOptions * opts, char * txt)
{
char * pz;
pz = SPN_WHITESPACE_CHARS(txt+1);
txt = strchr(pz, '>');
if (txt != NULL) {
size_t len = (unsigned)(txt - pz);
char * ftxt = AGALOC(len + 1, "aoflags");
memcpy(ftxt, pz, len);
ftxt[len] = NUL;
set_usage_flags(opts, ftxt);
AGFREE(ftxt);
txt++;
}
return txt;
}
LOCAL void
mungeString(char * txt, tOptionLoadMode mode)
{
char * pzE;
if (mode == OPTION_LOAD_KEEP)
return;
if (IS_WHITESPACE_CHAR(*txt)) {
char * pzS = SPN_WHITESPACE_CHARS(txt+1);
size_t l = strlen(pzS) + 1;
memmove(txt, pzS, l);
pzE = txt + l - 1;
} else
pzE = txt + strlen(txt);
pzE = SPN_WHITESPACE_BACK(txt, pzE);
*pzE = NUL;
if (mode == OPTION_LOAD_UNCOOKED)
return;
switch (*txt) {
default: return;
case '"':
case '\'': break;
}
switch (pzE[-1]) {
default: return;
case '"':
case '\'': break;
}
(void)ao_string_cook(txt, NULL);
}
/**
* handle program segmentation of config file.
*
* @param[in,out] opts program option descriptor
* @param[in] txt scanning pointer
* @returns the next character to look at
*/
static char *
program_directive(tOptions * opts, char * txt)
{
static char const ttlfmt[] = "<?";
size_t ttl_len = sizeof(ttlfmt) + strlen(zCfgProg);
char * ttl = AGALOC(ttl_len, "prog title");
size_t name_len = strlen(opts->pzProgName);
memcpy(ttl, ttlfmt, sizeof(ttlfmt) - 1);
memcpy(ttl + sizeof(ttlfmt) - 1, zCfgProg, ttl_len - (sizeof(ttlfmt) - 1));
do {
txt = SPN_WHITESPACE_CHARS(txt+1);
if ( (strneqvcmp(txt, opts->pzProgName, (int)name_len) == 0)
&& (IS_END_XML_TOKEN_CHAR(txt[name_len])) ) {
txt += name_len;
break;
}
txt = strstr(txt, ttl);
} while (txt != NULL);
AGFREE(ttl);
if (txt != NULL)
for (;;) {
if (*txt == NUL) {
txt = NULL;
break;
}
if (*(txt++) == '>')
break;
}
return txt;
}
/**
* This directive @i{is} processed, but only if the expression begins with
* either a back quote (@code{`}) or an open parenthesis (@code{(}).
* Text within the back quotes are handed off to the shell for processing
* and parenthesized text is handed off to Guile. Multiple line expressions
* must be joined with backslashes.
*
* If the @code{shell-script} or @code{scheme-expr} do not yield @code{true}
* valued results, autogen will be aborted. If @code{<anything else>} or
* nothing at all is provided, then this directive is ignored.
*
* The result is @code{false} (and fails) if the result is empty, the
* number zero, or a string that starts with the letters 'n' or 'f' ("no"
* or "false").
*/
char *
doDir_assert(directive_enum_t id, char const * dir, char * scan_next)
{
(void)id;
dir = SPN_WHITESPACE_CHARS(dir);
switch (*dir) {
case '`':
{
char * pzS = (char *)dir+1;
char * pzR = SPN_WHITESPACE_BACK(pzS, NULL);
if (*(--pzR) != '`')
break; /* not a valid script */
*pzR = NUL;
pzS = shell_cmd((char const *)pzS);
check_assert_str(pzS, dir);
AGFREE(pzS);
break;
}
case '(':
{
SCM res = ag_scm_c_eval_string_from_file_line(
dir, cctx->scx_fname, cctx->scx_line);
check_assert_str(scm2display(res), dir);
break;
}
default:
break;
}
return scan_next;
}
/**
* We've found a '<' character. We ignore this if it is a comment or a
* directive. If it is something else, then whatever it is we are looking
* at is bogus. Returning NULL stops processing.
*
* @param[in] xml_name the name of an xml bracket (usually)
* @param[in,out] res_val the option data derived from the XML element
*
* @returns the place to resume scanning input
*/
static char const *
scan_xml(char const * xml_name, tOptionValue * res_val)
{
size_t nm_len, v_len;
char const * scan;
char const * val_str;
tOptionValue valu;
tOptionLoadMode save_mode = option_load_mode;
if (! IS_VAR_FIRST_CHAR(*++xml_name))
return unnamed_xml(xml_name);
/*
* "scan_xml_name()" may change "option_load_mode".
*/
val_str = scan_xml_name(xml_name, &nm_len, &valu);
if (val_str == NULL)
goto bail_scan_xml;
if (valu.valType == OPARG_TYPE_NONE)
scan = val_str;
else {
if (option_load_mode != OPTION_LOAD_KEEP)
val_str = SPN_WHITESPACE_CHARS(val_str);
scan = find_end_xml(xml_name, nm_len, val_str, &v_len);
if (scan == NULL)
goto bail_scan_xml;
}
/*
* "scan" now points to where the scan is to resume after returning.
* It either points after "/>" at the end of the XML element header,
* or it points after the "</name>" tail based on the name in the header.
*/
switch (valu.valType) {
case OPARG_TYPE_NONE:
add_string(&(res_val->v.nestVal), xml_name, nm_len, NULL, 0);
break;
case OPARG_TYPE_STRING:
{
tOptionValue * new_val = add_string(
&(res_val->v.nestVal), xml_name, nm_len, val_str, v_len);
if (option_load_mode != OPTION_LOAD_KEEP)
munge_str(new_val->v.strVal, option_load_mode);
break;
}
case OPARG_TYPE_BOOLEAN:
add_bool(&(res_val->v.nestVal), xml_name, nm_len, val_str, v_len);
break;
case OPARG_TYPE_NUMERIC:
add_number(&(res_val->v.nestVal), xml_name, nm_len, val_str, v_len);
break;
case OPARG_TYPE_HIERARCHY:
{
char * pz = AGALOC(v_len+1, "h scan");
memcpy(pz, val_str, v_len);
pz[v_len] = NUL;
add_nested(&(res_val->v.nestVal), xml_name, nm_len, pz, v_len);
AGFREE(pz);
break;
}
case OPARG_TYPE_ENUMERATION:
case OPARG_TYPE_MEMBERSHIP:
default:
break;
}
option_load_mode = save_mode;
return scan;
bail_scan_xml:
option_load_mode = save_mode;
return NULL;
}
/**
* "txt" points to a '<' character, followed by an alpha.
* The end of the entry is either the "/>" following the name, or else a
* "</name>" string.
*/
static char *
handle_struct(tOptions * opts, tOptState * ost, char * txt, int dir)
{
tOptionLoadMode mode = option_load_mode;
tOptionValue valu;
char* pzName = ++txt;
char* pzData;
char* pcNulPoint;
txt = SPN_VALUE_NAME_CHARS(txt);
pcNulPoint = txt;
valu.valType = OPARG_TYPE_STRING;
switch (*txt) {
case ' ':
case '\t':
txt = (void *)parse_attrs(
opts, SPN_WHITESPACE_CHARS(txt), &mode, &valu);
if (txt == NULL)
return txt;
if (*txt == '>')
break;
if (*txt != '/')
return NULL;
/* FALLTHROUGH */
case '/':
if (txt[1] != '>')
return NULL;
*txt = NUL;
txt += 2;
loadOptionLine(opts, ost, pzName, dir, mode);
return txt;
case '>':
break;
default:
txt = strchr(txt, '>');
if (txt != NULL)
txt++;
return txt;
}
/*
* If we are here, we have a value. "txt" points to a closing angle
* bracket. Separate the name from the value for a moment.
*/
*pcNulPoint = NUL;
pzData = ++txt;
txt = trim_xml_text(txt, pzName, mode);
if (txt == NULL)
return txt;
/*
* Rejoin the name and value for parsing by "loadOptionLine()".
* Erase any attributes parsed by "parse_attrs()".
*/
memset(pcNulPoint, ' ', (size_t)(pzData - pcNulPoint));
/*
* If we are getting a "string" value that is to be cooked,
* then process the XML-ish &xx; XML-ish and %XX hex characters.
*/
if ( (valu.valType == OPARG_TYPE_STRING)
&& (mode == OPTION_LOAD_COOKED))
cook_xml_text(pzData);
/*
* "pzName" points to what looks like text for one option/configurable.
* It is NUL terminated. Process it.
*/
loadOptionLine(opts, ost, pzName, dir, mode);
return txt;
}
/**
* "txt" points to the start of some value name.
* The end of the entry is the end of the line that is not preceded by
* a backslash escape character. The string value is always processed
* in "cooked" mode.
*/
static char *
handle_cfg(tOptions * opts, tOptState * ost, char * txt, int dir)
{
char* pzName = txt++;
char* pzEnd = strchr(txt, NL);
if (pzEnd == NULL)
return txt + strlen(txt);
txt = SPN_VALUE_NAME_CHARS(txt);
txt = SPN_WHITESPACE_CHARS(txt);
if (txt > pzEnd) {
name_only:
*pzEnd++ = NUL;
loadOptionLine(opts, ost, pzName, dir, OPTION_LOAD_UNCOOKED);
return pzEnd;
}
/*
* Either the first character after the name is a ':' or '=',
* or else we must have skipped over white space. Anything else
* is an invalid format and we give up parsing the text.
*/
if ((*txt == '=') || (*txt == ':')) {
txt = SPN_WHITESPACE_CHARS(txt+1);
if (txt > pzEnd)
goto name_only;
} else if (! IS_WHITESPACE_CHAR(txt[-1]))
return NULL;
/*
* IF the value is continued, remove the backslash escape and push "pzEnd"
* on to a newline *not* preceded by a backslash.
*/
if (pzEnd[-1] == '\\') {
char* pcD = pzEnd-1;
char* pcS = pzEnd;
for (;;) {
char ch = *(pcS++);
switch (ch) {
case NUL:
pcS = NULL;
/* FALLTHROUGH */
case NL:
*pcD = NUL;
pzEnd = pcS;
goto copy_done;
case '\\':
if (*pcS == NL)
ch = *(pcS++);
/* FALLTHROUGH */
default:
*(pcD++) = ch;
}
} copy_done:;
} else {
/*
* The newline was not preceded by a backslash. NUL it out
*/
*(pzEnd++) = NUL;
}
/*
* "pzName" points to what looks like text for one option/configurable.
* It is NUL terminated. Process it.
*/
loadOptionLine(opts, ost, pzName, dir, OPTION_LOAD_UNCOOKED);
return pzEnd;
}
/**
* Load a file containing presetting information (a configuration file).
*/
static void
file_preset(tOptions * opts, char const * fname, int dir)
{
tmap_info_t cfgfile;
tOptState optst = OPTSTATE_INITIALIZER(PRESET);
opt_state_mask_t st_flags = optst.flags;
char * ftext =
text_mmap(fname, PROT_READ|PROT_WRITE, MAP_PRIVATE, &cfgfile);
if (TEXT_MMAP_FAILED_ADDR(ftext))
return;
if (dir == DIRECTION_CALLED) {
st_flags = OPTST_DEFINED;
dir = DIRECTION_PROCESS;
}
/*
* IF this is called via "optionProcess", then we are presetting.
* This is the default and the PRESETTING bit will be set.
* If this is called via "optionFileLoad", then the bit is not set
* and we consider stuff set herein to be "set" by the client program.
*/
if ((opts->fOptSet & OPTPROC_PRESETTING) == 0)
st_flags = OPTST_SET;
do {
optst.flags = st_flags;
ftext = SPN_WHITESPACE_CHARS(ftext);
if (IS_VAR_FIRST_CHAR(*ftext)) {
ftext = handle_cfg(opts, &optst, ftext, dir);
} else switch (*ftext) {
case '<':
if (IS_VAR_FIRST_CHAR(ftext[1]))
ftext = handle_struct(opts, &optst, ftext, dir);
else switch (ftext[1]) {
case '?':
ftext = handle_directive(opts, ftext);
break;
case '!':
ftext = handle_comment(ftext);
break;
case '/':
ftext = strchr(ftext + 2, '>');
if (ftext++ != NULL)
break;
default:
ftext = NULL;
}
if (ftext == NULL)
goto all_done;
break;
case '[':
ftext = handle_section(opts, ftext);
break;
case '#':
ftext = strchr(ftext + 1, NL);
break;
default:
goto all_done; /* invalid format */
}
} while (ftext != NULL);
all_done:
text_munmap(&cfgfile);
}
/**
* Parse the various attributes of an XML-styled config file entry
*
* @returns NULL on failure, otherwise the scan point
*/
LOCAL char const *
parse_attrs(tOptions * opts, char const * txt, tOptionLoadMode * pMode,
tOptionValue * pType)
{
size_t len = 0;
for (;;) {
len = (size_t)(SPN_LOWER_CASE_CHARS(txt) - txt);
/*
* The enumeration used in this switch is derived from this switch
* statement itself. The "find_option_xat_attribute_cmd" function
* will return XAT_CMD_MEMBERS for the "txt" string value
* "members", etc.
*/
switch (find_option_xat_attribute_cmd(txt, len)) {
case XAT_CMD_TYPE:
txt = parse_value(txt+len, pType);
break;
case XAT_CMD_WORDS:
txt = parse_keyword(opts, txt+len, pType);
break;
case XAT_CMD_MEMBERS:
txt = parse_set_mem(opts, txt+len, pType);
break;
case XAT_CMD_COOKED:
txt += len;
if (! IS_END_XML_TOKEN_CHAR(*txt))
goto invalid_kwd;
*pMode = OPTION_LOAD_COOKED;
break;
case XAT_CMD_UNCOOKED:
txt += len;
if (! IS_END_XML_TOKEN_CHAR(*txt))
goto invalid_kwd;
*pMode = OPTION_LOAD_UNCOOKED;
break;
case XAT_CMD_KEEP:
txt += len;
if (! IS_END_XML_TOKEN_CHAR(*txt))
goto invalid_kwd;
*pMode = OPTION_LOAD_KEEP;
break;
default:
case XAT_INVALID_CMD:
invalid_kwd:
pType->valType = OPARG_TYPE_NONE;
return skip_unkn(txt);
}
if (txt == NULL)
return NULL;
txt = SPN_WHITESPACE_CHARS(txt);
switch (*txt) {
case '/': pType->valType = OPARG_TYPE_NONE;
/* FALLTHROUGH */
case '>': return txt;
}
if (! IS_LOWER_CASE_CHAR(*txt))
return NULL;
}
}
/*=
* private:
*
* what: parse a hierarchical option argument
* arg: + char const * + pzTxt + the text to scan +
* arg: + char const * + pzName + the name for the text +
* arg: + size_t + nm_len + the length of "name" +
*
* ret_type: tOptionValue *
* ret_desc: An allocated, compound value structure
*
* doc:
* A block of text represents a series of values. It may be an
* entire configuration file, or it may be an argument to an
* option that takes a hierarchical value.
*
* If NULL is returned, errno will be set:
* @itemize @bullet
* @item
* @code{EINVAL} the input text was NULL.
* @item
* @code{ENOMEM} the storage structures could not be allocated
* @item
* @code{ENOMSG} no configuration values were found
* @end itemize
=*/
LOCAL tOptionValue *
optionLoadNested(char const * text, char const * name, size_t nm_len)
{
tOptionValue * res_val;
/*
* Make sure we have some data and we have space to put what we find.
*/
if (text == NULL) {
errno = EINVAL;
return NULL;
}
text = SPN_WHITESPACE_CHARS(text);
if (*text == NUL) {
errno = ENOMSG;
return NULL;
}
res_val = AGALOC(sizeof(*res_val) + nm_len + 1, "nest args");
res_val->valType = OPARG_TYPE_HIERARCHY;
res_val->pzName = (char *)(res_val + 1);
memcpy(res_val->pzName, name, nm_len);
res_val->pzName[nm_len] = NUL;
{
tArgList * arg_list = AGALOC(sizeof(*arg_list), "nest arg l");
res_val->v.nestVal = arg_list;
arg_list->useCt = 0;
arg_list->allocCt = MIN_ARG_ALLOC_CT;
}
/*
* Scan until we hit a NUL.
*/
do {
text = SPN_WHITESPACE_CHARS(text);
if (IS_VAR_FIRST_CHAR(*text))
text = scan_name(text, res_val);
else switch (*text) {
case NUL: goto scan_done;
case '<': text = scan_xml(text, res_val);
if (text == NULL) goto woops;
if (*text == ',') text++; break;
case '#': text = strchr(text, NL); break;
default: goto woops;
}
} while (text != NULL); scan_done:;
{
tArgList * al = res_val->v.nestVal;
if (al->useCt == 0) {
errno = ENOMSG;
goto woops;
}
if (al->useCt > 1)
sort_list(al);
}
return res_val;
woops:
AGFREE(res_val->v.nestVal);
AGFREE(res_val);
return NULL;
}
/*=export_func optionSetMembers
* what: Convert between bit flag values and strings
* private:
*
* arg: tOptions *, opts, the program options descriptor
* arg: tOptDesc *, od, the set membership option description
* arg: char const * const *,
* nm_list, list of enumeration names
* arg: unsigned int, nm_ct, number of names in list
*
* doc: This converts the optArg.argString string from the option description
* into the index corresponding to an entry in the name list.
* This will match the generated enumeration value.
* Full matches are always accepted. Partial matches are accepted
* if there is only one partial match.
=*/
void
optionSetMembers(tOptions * opts, tOptDesc * od,
char const * const * nm_list, unsigned int nm_ct)
{
/*
* IF the program option descriptor pointer is invalid,
* then it is some sort of special request.
*/
switch ((uintptr_t)opts) {
case (uintptr_t)OPTPROC_EMIT_USAGE:
enum_err(OPTPROC_EMIT_USAGE, od, nm_list, nm_ct);
return;
case (uintptr_t)OPTPROC_EMIT_SHELL:
set_memb_shell(opts, od, nm_list, nm_ct);
return;
case (uintptr_t)OPTPROC_RETURN_VALNAME:
set_memb_names(opts, od, nm_list, nm_ct);
return;
default:
break;
}
if ((od->fOptState & OPTST_RESET) != 0)
return;
{
char const * arg;
bool invert;
uintptr_t res = check_membership_start(od, &arg, &invert);
if (arg == NULL)
goto fail_return;
while (*arg != NUL) {
bool inv_val = false;
int len;
switch (*arg) {
case ',':
arg = SPN_WHITESPACE_CHARS(arg+1);
if ((*arg == ',') || (*arg == '|'))
goto fail_return;
continue;
case '-':
case '!':
inv_val = true;
/* FALLTHROUGH */
case '+':
case '|':
arg = SPN_WHITESPACE_CHARS(arg+1);
}
len = (int)(BRK_SET_SEPARATOR_CHARS(arg) - arg);
if (len == 0)
break;
if ((len == 3) && (strncmp(arg, zAll, 3) == 0)) {
if (inv_val)
res = 0;
else res = ~0UL;
}
else if ((len == 4) && (strncmp(arg, zNone, 4) == 0)) {
if (! inv_val)
res = 0;
}
else do {
char * pz;
uintptr_t bit = strtoul(arg, &pz, 0);
if (pz != arg + len) {
bit = find_member_bit(opts, od, pz, len, nm_list, nm_ct);
if (bit == 0UL)
goto fail_return;
}
if (inv_val)
res &= ~bit;
else res |= bit;
} while (false);
arg = SPN_WHITESPACE_CHARS(arg + len);
}
if (invert)
res ^= ~0UL;
if (nm_ct < (8 * sizeof(uintptr_t)))
res &= (1UL << nm_ct) - 1UL;
od->optCookie = VOIDP(res);
}
return;
fail_return:
od->optCookie = VOIDP(0);
}
/**
* Will add the name to the define list as if it were a DEFINE program
* argument. Its value will be the first non-whitespace token following
* the name. Quotes are @strong{not} processed.
*
* After the definitions file has been processed, any remaining entries
* in the define list will be added to the environment.
*/
char *
doDir_define(directive_enum_t id, char const * dir, char * scan_next)
{
char * def_name = SPN_WHITESPACE_CHARS(dir);
(void)id;
/*
* Skip any #defines that do not look reasonable
*/
if (! IS_VAR_FIRST_CHAR(*def_name))
return scan_next;
dir = SPN_VARIABLE_NAME_CHARS(def_name);
/*
* IF this is a macro definition (rather than a value def),
* THEN we will ignore it.
*/
if (*dir == '(')
return scan_next;
/*
* We have found the end of the name.
* IF there is no more data on the line,
* THEN we do not have space for the '=' required by PUTENV.
* Therefore, move the name back over the "#define"
* directive itself, giving us the space needed.
*/
if (! IS_WHITESPACE_CHAR(*dir)) {
char * pzS = (char *)dir;
char * pzD = (def_name - 6);
*pzS = NUL; // whatever it used to be, it is a NUL now.
pzS = def_name;
while ((*(pzD++) = *(pzS++)) != NUL) ;
pzD[-1] = '=';
pzD[ 0] = NUL;
def_name -= 6;
} else {
/*
* Otherwise, insert the '=' and move any data up against it.
* We only accept one name-type, space separated token.
* We are not ANSI-C. ;-)
*/
char * pz = (char *)(dir++);
*pz++ = '=';
dir = SPN_WHITESPACE_CHARS(dir);
/*
* Copy chars for so long as it is not NUL and does not require quoting
*/
for (;;) {
if ((*pz++ = *dir++) == NUL)
break;
if (! IS_UNQUOTABLE_CHAR(*dir)) {
*pz = NUL;
break;
}
}
}
SET_OPT_DEFINE(def_name);
return scan_next;
}
/*=export_func ao_string_tokenize
*
* what: tokenize an input string
*
* arg: + char const* + string + string to be tokenized +
*
* ret_type: token_list_t*
* ret_desc: pointer to a structure that lists each token
*
* doc:
*
* This function will convert one input string into a list of strings.
* The list of strings is derived by separating the input based on
* white space separation. However, if the input contains either single
* or double quote characters, then the text after that character up to
* a matching quote will become the string in the list.
*
* The returned pointer should be deallocated with @code{free(3C)} when
* are done using the data. The data are placed in a single block of
* allocated memory. Do not deallocate individual token/strings.
*
* The structure pointed to will contain at least these two fields:
* @table @samp
* @item tkn_ct
* The number of tokens found in the input string.
* @item tok_list
* An array of @code{tkn_ct + 1} pointers to substring tokens, with
* the last pointer set to NULL.
* @end table
*
* There are two types of quoted strings: single quoted (@code{'}) and
* double quoted (@code{"}). Singly quoted strings are fairly raw in that
* escape characters (@code{\\}) are simply another character, except when
* preceding the following characters:
* @example
* @code{\\} double backslashes reduce to one
* @code{'} incorporates the single quote into the string
* @code{\n} suppresses both the backslash and newline character
* @end example
*
* Double quote strings are formed according to the rules of string
* constants in ANSI-C programs.
*
* example:
* @example
* #include <stdlib.h>
* int ix;
* token_list_t* ptl = ao_string_tokenize(some_string)
* for (ix = 0; ix < ptl->tkn_ct; ix++)
* do_something_with_tkn(ptl->tkn_list[ix]);
* free(ptl);
* @end example
* Note that everything is freed with the one call to @code{free(3C)}.
*
* err:
* NULL is returned and @code{errno} will be set to indicate the problem:
* @itemize @bullet
* @item
* @code{EINVAL} - There was an unterminated quoted string.
* @item
* @code{ENOENT} - The input string was empty.
* @item
* @code{ENOMEM} - There is not enough memory.
* @end itemize
=*/
token_list_t*
ao_string_tokenize(char const* str)
{
token_list_t* res = alloc_token_list(str);
ch_t* pzDest;
/*
* Now copy each token into the output buffer.
*/
if (res == NULL)
return res;
pzDest = (ch_t*)(res->tkn_list[0]);
res->tkn_ct = 0;
do {
res->tkn_list[ res->tkn_ct++ ] = pzDest;
for (;;) {
int ch = (ch_t)*str;
if (IS_WHITESPACE_CHAR(ch)) {
found_white_space:
str = SPN_WHITESPACE_CHARS(str+1);
break;
}
switch (ch) {
case '"':
copy_cooked(&pzDest, &str);
if (str == NULL) {
free(res);
errno = EINVAL;
return NULL;
}
if (IS_WHITESPACE_CHAR(*str))
goto found_white_space;
break;
case '\'':
copy_raw(&pzDest, &str);
if (str == NULL) {
free(res);
errno = EINVAL;
return NULL;
}
if (IS_WHITESPACE_CHAR(*str))
goto found_white_space;
break;
case NUL:
goto copy_done;
default:
str++;
*(pzDest++) = (unsigned char)ch;
}
} copy_done:;
/*
* NUL terminate the last token and see if we have any more tokens.
*/
*(pzDest++) = NUL;
} while (*str != NUL);
res->tkn_list[ res->tkn_ct ] = NULL;
return res;
}
/**
* This directive will insert definitions from another file into
* the current collection. If the file name is adorned with
* double quotes or angle brackets (as in a C program), then the
* include is ignored.
*/
char *
doDir_include(directive_enum_t id, char const * dir, char * scan_next)
{
static char const * const apzSfx[] = { DIRECT_INC_DEF_SFX, NULL };
scan_ctx_t * new_ctx;
size_t inc_sz;
char full_name[ AG_PATH_MAX + 1 ];
(void)id;
dir = SPN_WHITESPACE_CHARS(dir);
/*
* Ignore C-style includes. This allows "C" files to be processed
* for their "#define"s.
*/
if ((*dir == '"') || (*dir == '<'))
return scan_next;
if (! SUCCESSFUL(
find_file(dir, full_name, apzSfx, cctx->scx_fname))) {
errno = ENOENT;
fswarn("search for", cctx->scx_fname);
return scan_next;
}
/*
* Make sure the specified file is a regular file and we can get
* the correct size for it.
*/
inc_sz = file_size(full_name);
if (inc_sz == 0)
return scan_next;
/*
* Get the space for the output data and for context overhead.
* This is an extra allocation and copy, but easier than rewriting
* 'loadData()' for this special context.
*/
{
size_t sz = sizeof(scan_ctx_t) + 4 + inc_sz;
new_ctx = (scan_ctx_t *)AGALOC(sz, "inc def head");
memset(VOIDP(new_ctx), 0, sz);
new_ctx->scx_line = 1;
}
/*
* Link it into the context stack
*/
cctx->scx_scan = scan_next;
new_ctx->scx_next = cctx;
cctx = new_ctx;
AGDUPSTR(new_ctx->scx_fname, full_name, "def file");
new_ctx->scx_scan =
new_ctx->scx_data =
scan_next = (char *)(new_ctx + 1);
/*
* Read all the data. Usually in a single read, but loop
* in case multiple passes are required.
*/
{
FILE * fp = fopen(full_name, "r" FOPEN_TEXT_FLAG);
char * pz = scan_next;
if (fp == NULL)
AG_CANT(DIRECT_INC_CANNOT_OPEN, full_name);
if (dep_fp != NULL)
add_source_file(full_name);
do {
size_t rdct = fread(VOIDP(pz), (size_t)1, inc_sz, fp);
if (rdct == 0)
AG_CANT(DIRECT_INC_CANNOT_READ, full_name);
pz += rdct;
inc_sz -= rdct;
} while (inc_sz > 0);
fclose(fp);
*pz = NUL;
}
return scan_next;
}
