# include "base64.h"
# include "hex.h"
#include "dstr.h"
+#include "maths.h"
#include "tvec.h"
/*----- Preliminary utilities ---------------------------------------------*/
-#ifdef isnan
-# define NANP(x) isnan(x)
-#else
-# define NANP(x) (!((x) == (x)))
-#endif
-
-#ifdef isinf
-# define INFP(x) isinf(x)
-#else
-# define INFP(x) ((x) > DBL_MAX || (x) < DBL_MIN)
-#endif
-
-#ifdef signbit
-# define NEGP(x) signbit(x)
-#else
-# define NEGP(x) ((x) < 0.0)
-#endif
+/* --- @trivial_release@ --- *
+ *
+ * Arguments: @union tvec_regval *rv@ = a register value
+ * @const struct tvec_regdef@ = the register definition
+ *
+ * Returns: ---
+ *
+ * Use: Does nothing. Used for register values which don't retain
+ * resources.
+ */
static void trivial_release(union tvec_regval *rv,
const struct tvec_regdef *rd)
{ ; }
+/*----- Integer utilities -------------------------------------------------*/
+
+/* --- @unsigned_to_buf@, @signed_to_buf@ --- *
+ *
+ * Arguments: @buf *b@ = buffer to write on
+ * @unsigned long u@ or @long i@ = integer to write
+ *
+ * Returns: Zero on success, @-1@ on failure.
+ *
+ * Use: Write @i@ to the buffer, in big-endian (two's-complement, it
+ * signed) format.
+ */
+
+static int unsigned_to_buf(buf *b, unsigned long u)
+ { kludge64 k; ASSIGN64(k, u); return (buf_putk64l(b, k)); }
+
static int signed_to_buf(buf *b, long i)
{
kludge64 k;
return (buf_putk64l(b, k));
}
-static int signed_from_buf(buf *b, long *i_out)
-{
- kludge64 k, lmax, not_lmin;
-
- ASSIGN64(lmax, LONG_MAX); ASSIGN64(not_lmin, ~(unsigned long)LONG_MIN);
- if (buf_getk64l(b, &k)) return (-1);
- if (CMP64(k, <=, lmax)) *i_out = (long)GET64(unsigned long, k);
- else {
- CPL64(k, k);
- if (CMP64(k, <=, not_lmin)) *i_out = -(long)GET64(unsigned long, k) - 1;
- else return (-1);
- }
- return (0);
-}
-
-static int unsigned_to_buf(buf *b, unsigned long u)
- { kludge64 k; ASSIGN64(k, u); return (buf_putk64l(b, k)); }
+/* --- @unsigned_from_buf@, @signed_from_buf@ --- *
+ *
+ * Arguments: @buf *b@ = buffer to write on
+ * @unsigned long *u_out@ or @long *i_out@ = where to put the
+ * result
+ *
+ * Returns: Zero on success, @-1@ on failure.
+ *
+ * Use: Read an integer, in big-endian (two's-complement, if signed)
+ * format, from the buffer.
+ */
static int unsigned_from_buf(buf *b, unsigned long *u_out)
{
*u_out = GET64(unsigned long, k); return (0);
}
+/* --- @hex_width@ --- *
+ *
+ * Arguments: @unsigned long u@ = an integer
+ *
+ * Returns: A suitable number of digits to use in order to display @u@ in
+ * hex. Currently, we select a power of two sufficient to show
+ * the value, but at least 2.
+ */
+
static int hex_width(unsigned long u)
{
int wd;
return (wd/4);
}
+/* --- @format_unsigned_hex@, @format_signed_hex@ --- *
+ *
+ * Arguments: @const struct gprintf_ops *gops@ = print operations
+ * @void *go@ = print destination
+ * @unsigned long u@ or @long i@ = integer to print
+ *
+ * Returns: ---
+ *
+ * Use: Print an unsigned or signed integer in hexadecimal.
+ */
+
+static void format_unsigned_hex(const struct gprintf_ops *gops, void *go,
+ unsigned long u)
+ { gprintf(gops, go, "0x%0*lx", hex_width(u), u); }
+
+static void format_signed_hex(const struct gprintf_ops *gops, void *go,
+ long i)
+{
+ unsigned long u = i >= 0 ? i : -(unsigned long)i;
+ gprintf(gops, go, "%s0x%0*lx", i < 0 ? "-" : "", hex_width(u), u);
+}
+
+static int signed_from_buf(buf *b, long *i_out)
+{
+ kludge64 k, lmax, not_lmin;
+
+ ASSIGN64(lmax, LONG_MAX); ASSIGN64(not_lmin, ~(unsigned long)LONG_MIN);
+ if (buf_getk64l(b, &k)) return (-1);
+ if (CMP64(k, <=, lmax)) *i_out = (long)GET64(unsigned long, k);
+ else {
+ CPL64(k, k);
+ if (CMP64(k, <=, not_lmin)) *i_out = -(long)GET64(unsigned long, k) - 1;
+ else return (-1);
+ }
+ return (0);
+}
+
+/* --- @check_unsigned_range@, @check_signed_range@ --- *
+ *
+ * Arguments: @unsigned long u@ or @long i@ = an integer
+ * @const struct tvec_urange *ur@ or
+ * @const struct tvec_irange *ir@ = range specification,
+ * or null
+ * @struct tvec_state *tv@ = test vector state
+ *
+ * Returns: Zero on success, or @-1@ on error.
+ *
+ * Use: Check that the integer is within bounds. If not, report a
+ * suitable error and return a failure indication.
+ */
+
static int check_signed_range(long i,
const struct tvec_irange *ir,
struct tvec_state *tv)
return (0);
}
+/* --- @chtodig@ --- *
+ *
+ * Arguments: @int ch@ = a character
+ *
+ * Returns: The numeric value of the character as a digit, or @-1@ if
+ * it's not a digit. Letters count as extended digits starting
+ * with value 10; case is not significant.
+ */
+
static int chtodig(int ch)
{
if ('0' <= ch && ch <= '9') return (ch - '0');
else return (-1);
}
+/* --- @parse_unsigned_integer@, @parse_signed_integer@ --- *
+ *
+ * Arguments: @unsigned long *u_out@, @long *i_out@ = where to put the
+ * result
+ * @const char **q_out@ = where to put the end position
+ * @const char *p@ = pointer to the string to parse
+ *
+ * Returns: Zero on success, @-1@ on error.
+ *
+ * Use: Parse an integer from a string in the test-vector format.
+ * This is mostly extension of the traditional C @strtoul@
+ * format: supported inputs include:
+ *
+ * * NNN -- a decimal number (even if it starts with `0');
+ * * 0xNNN -- hexadecimal;
+ * * 0oNNN -- octal;
+ * * 0bNNN -- binary;
+ * * NNrNNN -- base NN.
+ *
+ * Furthermore, single underscores are permitted internally as
+ * an insignificant digit separator.
+ */
+
static int parse_unsigned_integer(unsigned long *u_out, const char **q_out,
const char *p)
{
int ch, d, r;
const char *q;
unsigned f = 0;
-#define f_implicit 1u
-#define f_digit 2u
-#define f_uscore 4u
-
+#define f_implicit 1u /* implicitly reading base 10 */
+#define f_digit 2u /* read a real digit */
+#define f_uscore 4u /* found an underscore */
+
+ /* Initial setup
+ *
+ * This will deal with the traditional `0[box]...' prefixes. We'll leave
+ * our new `NNr...' syntax for later.
+ */
if (p[0] != '0' || !p[1]) {
d = chtodig(*p); if (0 > d || d >= 10) return (-1);
r = 10; u = d; p++; f |= f_implicit | f_digit;
q = p;
for (;;) {
- ch = *p;
- if (ch == '_') {
- if (f&f_uscore) break;
- p++; f = (f&~f_implicit) | f_uscore;
- }
- else if (ch == 'r' || ch == 'R') {
- if (!(f&f_implicit) || !u || u >= 36) break;
- d = chtodig(p[1]); if (0 > d || d >= u) break;
- r = u; u = d; f = (f&~f_implicit) | f_digit; p += 2; q = p;
- } else {
- d = chtodig(ch);
- if (d < 0 || d >= r) break;
- if (u > ULONG_MAX/r) return (-1);
- u *= r; if (u > ULONG_MAX - d) return (-1);
- u += d; f = (f&~f_uscore) | f_digit; p++; q = p;
+ /* Work through the string a character at a time. */
+
+ ch = *p; switch (ch) {
+
+ case '_':
+ /* An underscore is OK if we haven't just seen one. */
+
+ if (f&f_uscore) goto done;
+ p++; f = (f&~f_implicit) | f_uscore;
+ break;
+
+ case 'r': case 'R':
+ /* An `r' is OK if the number so far is small enough to be a sensible
+ * base, and we're scanning decimal implicitly.
+ */
+
+ if (!(f&f_implicit) || !u || u >= 36) goto done;
+ d = chtodig(p[1]); if (0 > d || d >= u) goto done;
+ r = u; u = d; f = (f&~f_implicit) | f_digit; p += 2; q = p;
+ break;
+
+ default:
+ /* Otherwise we expect a valid digit and accumulate it. */
+ d = chtodig(ch); if (d < 0 || d >= r) goto done;
+ if (u > ULONG_MAX/r) return (-1);
+ u *= r; if (u > ULONG_MAX - d) return (-1);
+ u += d; f = (f&~f_uscore) | f_digit; p++; q = p;
+ break;
}
}
+done:
if (!(f&f_digit)) return (-1);
*u_out = u; *q_out = q; return (0);
unsigned f = 0;
#define f_neg 1u
+ /* Read an initial sign. */
if (*p == '+') p++;
else if (*p == '-') { f |= f_neg; p++; }
+ /* Scan an unsigned number. */
if (parse_unsigned_integer(&u, q_out, p)) return (-1);
+ /* Check for signed overflow and apply the sign. */
if (!(f&f_neg)) {
if (u > LONG_MAX) return (-1);
*i_out = u;
#undef f_neg
}
+/* --- @parse_unsigned@, @parse_signed@ --- *
+ *
+ * Arguments: @unsigned long *u_out@ or @long *i_out@ = where to put the
+ * result
+ * @const char *p@ = string to parse
+ * @const struct tvec_urange *ur@ or
+ * @const struct tvec_irange *ir@ = range specification,
+ * or null
+ * @struct tvec_state *tv@ = test vector state
+ *
+ * Returns: Zero on success, @-1@ on error.
+ *
+ * Use: Parse and range-check an integer. Unlike @parse_(un)signed_
+ * integer@, these functions check that there's no cruft
+ * following the final digit, and report errors as they find
+ * them rather than leaving that to the caller.
+ */
+
+static int parse_unsigned(unsigned long *u_out, const char *p,
+ const struct tvec_urange *ur,
+ struct tvec_state *tv)
+{
+ unsigned long u;
+ const char *q;
+
+ if (parse_unsigned_integer(&u, &q, p))
+ return (tvec_error(tv, "invalid unsigned integer `%s'", p));
+ if (*q) return (tvec_syntax(tv, *q, "end-of-line"));
+ if (check_unsigned_range(u, ur, tv)) return (-1);
+ *u_out = u; return (0);
+}
+
static int parse_signed(long *i_out, const char *p,
const struct tvec_irange *ir,
struct tvec_state *tv)
*i_out = i; return (0);
}
-static int parse_unsigned(unsigned long *u_out, const char *p,
- const struct tvec_urange *ur,
- struct tvec_state *tv)
-{
- unsigned long u;
- const char *q;
+/*----- Floating-point utilities ------------------------------------------*/
- if (parse_unsigned_integer(&u, &q, p))
- return (tvec_error(tv, "invalid unsigned integer `%s'", p));
- if (*q) return (tvec_syntax(tv, *q, "end-of-line"));
- if (check_unsigned_range(u, ur, tv)) return (-1);
- *u_out = u; return (0);
-}
+/* --- @eqish_floating_p@ --- *
+ *
+ * Arguments: @double x, y@ = two numbers to compare
+ * @const struct tvec_floatinfo *fi@ = floating-point info
+ *
+ * Returns: Nonzero if the comparand @y@ is sufficiently close to the
+ * reference @x@, or zero if it's definitely different.
+ */
-static void format_signed_hex(const struct gprintf_ops *gops, void *go,
- long i)
+static int eqish_floating_p(double x, double y,
+ const struct tvec_floatinfo *fi)
{
- unsigned long u = i >= 0 ? i : -(unsigned long)i;
- gprintf(gops, go, "%s0x%0*lx", i < 0 ? "-" : "", hex_width(u), u);
+ double t;
+
+ if (NANP(x)) return (NANP(y)); else if (NANP(y)) return (0);
+ if (INFP(x)) return (x == y); else if (INFP(y)) return (0);
+
+ switch (fi ? fi->f&TVFF_EQMASK : TVFF_EXACT) {
+ case TVFF_EXACT:
+ return (x == y && NEGP(x) == NEGP(y));
+ case TVFF_ABSDELTA:
+ t = x - y; if (t < 0) t = -t; return (t < fi->delta);
+ case TVFF_RELDELTA:
+ t = 1.0 - y/x; if (t < 0) t = -t; return (t < fi->delta);
+ default:
+ abort();
+ }
}
-static void format_unsigned_hex(const struct gprintf_ops *gops, void *go,
- unsigned long u)
- { gprintf(gops, go, "0x%0*lx", hex_width(u), u); }
+/* --- @format_floating@ --- *
+ *
+ * Arguments: @const struct gprintf_ops *gops@ = print operations
+ * @void *go@ = print destination
+ * @double x@ = number to print
+ *
+ * Returns: ---
+ *
+ * Use: Print a floating-point number, accurately.
+ */
static void format_floating(const struct gprintf_ops *gops, void *go,
double x)
}
}
-static int eqish_floating_p(double x, double y,
- const struct tvec_floatinfo *fi)
-{
- double xx, yy, t;
-
- if (NANP(x)) return (NANP(y)); else if (NANP(y)) return (0);
- if (INFP(x)) return (x == y); else if (INFP(y)) return (0);
-
- switch (fi ? fi->f&TVFF_EQMASK : TVFF_EXACT) {
- case TVFF_EXACT:
- return (x == y && NEGP(x) == NEGP(y));
- case TVFF_ABSDELTA:
- t = x - y; if (t < 0) t = -t; return (t < fi->delta);
- case TVFF_RELDELTA:
- t = 1.0 - y/x; if (t < 0) t = -t; return (t < fi->delta);
- default:
- abort();
- }
-}
+/* --- @parse_floating@ --- *
+ *
+ * Arguments: @double *x_out@ = where to put the result
+ * @const char *p@ = string to parse
+ * @const struct tvec_floatinfo *fi@ = floating-point info
+ * @struct tvec_state *tv@ = test vector state
+ *
+ * Returns: Zero on success, @-1@ on error.
+ *
+ * Use: Parse a floating-point number from a string. Reports any
+ * necessary errors.
+ */
static int parse_floating(double *x_out, const char *p,
const struct tvec_floatinfo *fi,
double x;
int olderr, rc;
+ /* Check for special tokens. */
if (STRCMP(p, ==, "#nan")) {
#ifdef NAN
x = NAN; rc = 0;
tvec_error(tv, "NaN not supported on this system");
rc = -1; goto end;
#endif
- } else if (STRCMP(p, ==, "#inf") ||
- STRCMP(p, ==, "#+inf") ||
- STRCMP(p, ==, "+#inf")) {
+ }
+
+ else if (STRCMP(p, ==, "#inf") ||
+ STRCMP(p, ==, "#+inf") || STRCMP(p, ==, "+#inf")) {
#ifdef INFINITY
x = INFINITY; rc = 0;
#else
tvec_error(tv, "infinity not supported on this system");
rc = -1; goto end;
#endif
- } else if (STRCMP(p, ==, "#-inf") ||
- STRCMP(p, ==, "-#inf")) {
+ }
+
+ else if (STRCMP(p, ==, "#-inf") || STRCMP(p, ==, "-#inf")) {
#ifdef INFINITY
x = -INFINITY; rc = 0;
#else
tvec_error(tv, "infinity not supported on this system");
rc = -1; goto end;
#endif
- } else {
+ }
+
+ /* Check that this looks like a number, so we can exclude `strtod'
+ * recognizing its own non-finite number tokens.
+ */
+ else {
pp = p;
if (*pp == '+' || *pp == '-') pp++;
if (*pp == '.') pp++;
tvec_syntax(tv, *p ? *p : fgetc(tv->fp), "floating-point number");
rc = -1; goto end;
}
+
+ /* Parse the number using the system parser. */
olderr = errno; errno = 0;
x = strtod(p, &q);
if (*q) {
errno = olderr;
}
+ /* Check that the number is acceptable. */
if (NANP(x) && fi && !(fi->f&TVFF_NANOK)) {
tvec_error(tv, "#nan not allowed here");
rc = -1; goto end;
}
+
if (fi && ((!(fi->f&TVFF_NOMIN) && x < fi->min) ||
(!(fi->f&TVFF_NOMAX) && x > fi->max))) {
dstr_puts(&d, "floating-point number ");
tvec_error(tv, "%s", d.buf); rc = -1; goto end;
}
- *x_out = x; rc = 0;
-end:
- dstr_destroy(&d);
- return (rc);
+ /* All done. */
+ *x_out = x; rc = 0;
+end:
+ dstr_destroy(&d);
+ return (rc);
+}
+
+/*----- String utilities --------------------------------------------------*/
+
+/* Special character name table. */
+static const struct chartab {
+ const char *name; /* character name */
+ int ch; /* character value */
+ unsigned f; /* flags: */
+#define CTF_PREFER 1u /* preferred name */
+#define CTF_SHORT 2u /* short name (compact style) */
+} chartab[] = {
+ { "#eof", EOF, CTF_PREFER | CTF_SHORT },
+ { "#nul", '\0', CTF_PREFER },
+ { "#bell", '\a', CTF_PREFER },
+ { "#ding", '\a', 0 },
+ { "#bel", '\a', CTF_SHORT },
+ { "#backspace", '\b', CTF_PREFER },
+ { "#bs", '\b', CTF_SHORT },
+ { "#escape", '\x1b', CTF_PREFER },
+ { "#esc", '\x1b', CTF_SHORT },
+ { "#formfeed", '\f', CTF_PREFER },
+ { "#ff", '\f', CTF_SHORT },
+ { "#newline", '\n', CTF_PREFER },
+ { "#linefeed", '\n', 0 },
+ { "#lf", '\n', CTF_SHORT },
+ { "#nl", '\n', 0 },
+ { "#return", '\r', CTF_PREFER },
+ { "#carriage-return", '\r', 0 },
+ { "#cr", '\r', CTF_SHORT },
+ { "#tab", '\t', CTF_PREFER | CTF_SHORT },
+ { "#horizontal-tab", '\t', 0 },
+ { "#ht", '\t', 0 },
+ { "#vertical-tab", '\v', CTF_PREFER },
+ { "#vt", '\v', CTF_SHORT },
+ { "#space", ' ', 0 },
+ { "#spc", ' ', CTF_SHORT },
+ { "#delete", '\x7f', CTF_PREFER },
+ { "#del", '\x7f', CTF_SHORT },
+ { 0, 0, 0 }
+};
+
+/* --- @find_charname@ --- *
+ *
+ * Arguments: @int ch@ = character to match
+ * @unsigned f@ = flags (@CTF_...@) to match
+ *
+ * Returns: The name of the character, or null if no match is found.
+ *
+ * Use: Looks up a name for a character. Specifically, it returns
+ * the first entry in the @chartab@ table which matches @ch@ and
+ * which has one of the flags @f@ set.
+ */
+
+static const char *find_charname(int ch, unsigned f)
+{
+ const struct chartab *ct;
+
+ for (ct = chartab; ct->name; ct++)
+ if (ct->ch == ch && (ct->f&f)) return (ct->name);
+ return (0);
+}
+
+/* --- @read_charname@ --- *
+ *
+ * Arguments: @int *ch_out@ = where to put the character
+ * @const char *p@ = character name
+ * @unsigned f@ = flags (@TCF_...@)
+ *
+ * Returns: Zero if a match was found, @-1@ if not.
+ *
+ * Use: Looks up a character by name. If @RCF_EOFOK@ is set in @f@,
+ * then the @EOF@ marker can be matched; otherwise it can't.
+ */
+
+#define RCF_EOFOK 1u
+static int read_charname(int *ch_out, const char *p, unsigned f)
+{
+ const struct chartab *ct;
+
+ for (ct = chartab; ct->name; ct++)
+ if (STRCMP(p, ==, ct->name) && ((f&RCF_EOFOK) || ct->ch >= 0))
+ { *ch_out = ct->ch; return (0); }
+ return (-1);
+}
+
+/* --- @format_charesc@ --- *
+ *
+ * Arguments: @const struct gprintf_ops *gops@ = print operations
+ * @void *go@ = print destination
+ * @int ch@ = character to format
+ * @unsigned f@ = flags (@FCF_...@)
+ *
+ * Returns: ---
+ *
+ * Use: Format a character as an escape sequence, possibly as part of
+ * a larger string. If @FCF_BRACE@ is set in @f@, then put
+ * braces around a `\x...' code, so that it's suitable for use
+ * in a longer string.
+ */
+
+#define FCF_BRACE 1u
+static void format_charesc(const struct gprintf_ops *gops, void *go,
+ int ch, unsigned f)
+{
+ switch (ch) {
+ case '\a': gprintf(gops, go, "\\a"); break;
+ case '\b': gprintf(gops, go, "\\b"); break;
+ case '\x1b': gprintf(gops, go, "\\e"); break;
+ case '\f': gprintf(gops, go, "\\f"); break;
+ case '\r': gprintf(gops, go, "\\r"); break;
+ case '\n': gprintf(gops, go, "\\n"); break;
+ case '\t': gprintf(gops, go, "\\t"); break;
+ case '\v': gprintf(gops, go, "\\v"); break;
+ case '\\': gprintf(gops, go, "\\\\"); break;
+ case '\'': gprintf(gops, go, "\\'"); break;
+ case '\0':
+ if (f&FCF_BRACE) gprintf(gops, go, "\\{0}");
+ else gprintf(gops, go, "\\0");
+ break;
+ default:
+ if (f&FCF_BRACE)
+ gprintf(gops, go, "\\x{%0*x}", hex_width(UCHAR_MAX), ch);
+ else
+ gprintf(gops, go, "\\x%0*x", hex_width(UCHAR_MAX), ch);
+ break;
+ }
+}
+
+/* --- @format_char@ --- *
+ *
+ * Arguments: @const struct gprintf_ops *gops@ = print operations
+ * @void *go@ = print destination
+ * @int ch@ = character to format
+ *
+ * Returns: ---
+ *
+ * Use: Format a single character.
+ */
+
+static void format_char(const struct gprintf_ops *gops, void *go, int ch)
+{
+ switch (ch) {
+ case '\\': case '\'': escape:
+ gprintf(gops, go, "'");
+ format_charesc(gops, go, ch, 0);
+ gprintf(gops, go, "'");
+ break;
+ default:
+ if (!isprint(ch)) goto escape;
+ gprintf(gops, go, "'%c'", ch);
+ break;
+ }
+}
+
+/* --- @maybe_format_unsigned_char@, @maybe_format_signed_char@ --- *
+ *
+ * Arguments: @const struct gprintf_ops *gops@ = print operations
+ * @void *go@ = print destination
+ * @unsigned long u@ or @long i@ = an integer
+ *
+ * Returns: ---
+ *
+ * Use: Format a (signed or unsigned) integer as a character, if it's
+ * in range, printing something like `= 'q''. It's assumed that
+ * a comment marker has already been output.
+ */
+
+static void maybe_format_unsigned_char
+ (const struct gprintf_ops *gops, void *go, unsigned long u)
+{
+ const char *p;
+
+ p = find_charname(u, CTF_PREFER);
+ if (p) gprintf(gops, go, " = %s", p);
+ if (u < UCHAR_MAX)
+ { gprintf(gops, go, " = "); format_char(gops, go, u); }
}
-static int convert_hex(char ch, int *v_out)
+static void maybe_format_signed_char
+ (const struct gprintf_ops *gops, void *go, long i)
{
- if ('0' <= ch && ch <= '9') { *v_out = ch - '0'; return (0); }
- else if ('a' <= ch && ch <= 'f') { *v_out = ch - 'a' + 10; return (0); }
- else if ('A' <= ch && ch <= 'F') { *v_out = ch - 'A' + 10; return (0); }
- else return (-1);
+ const char *p;
+
+ p = find_charname(i, CTF_PREFER);
+ if (p) gprintf(gops, go, " = %s", p);
+ if (0 <= i && i < UCHAR_MAX)
+ { gprintf(gops, go, " = "); format_char(gops, go, i); }
}
-static int read_escape(int *ch_out, struct tvec_state *tv)
+/* --- @read_charesc@ --- *
+ *
+ * Arguments: @int *ch_out@ = where to put the result
+ * @struct tvec_state *tv@ = test vector state
+ *
+ * Returns: Zero on success, @-1@ on error.
+ *
+ * Use: Parse and convert an escape sequence from @tv@'s input
+ * stream, assuming that the initial `\' has already been read.
+ * Reports errors as appropriate.
+ */
+
+static int read_charesc(int *ch_out, struct tvec_state *tv)
{
int ch, i, esc;
unsigned f = 0;
ch = getc(tv->fp);
switch (ch) {
- case EOF: case '\n': tvec_syntax(tv, ch, "string escape");
+
+ /* Things we shouldn't find. */
+ case EOF: case '\n': return (tvec_syntax(tv, ch, "string escape"));
+
+ /* Single-character escapes. */
case '\'': *ch_out = '\''; break;
case '\\': *ch_out = '\\'; break;
case '"': *ch_out = '"'; break;
case 't': *ch_out = '\t'; break;
case 'v': *ch_out = '\v'; break;
+ /* Hex escapes, with and without braces. */
case 'x':
ch = getc(tv->fp);
if (ch == '{') { f |= f_brace; ch = getc(tv->fp); }
else f &= ~f_brace;
- if (convert_hex(ch, &esc))
- return (tvec_syntax(tv, ch, "hex digit"));
+ esc = chtodig(ch);
+ if (esc < 0 || esc >= 16) return (tvec_syntax(tv, ch, "hex digit"));
for (;;) {
- ch = getc(tv->fp); if (convert_hex(ch, &i)) break;
- esc = 8*esc + i;
+ ch = getc(tv->fp); i = chtodig(ch); if (i < 0 || i >= 16) break;
+ esc = 16*esc + i;
if (esc > UCHAR_MAX)
return (tvec_error(tv,
"character code %d out of range", esc));
*ch_out = esc;
break;
+ /* Other things, primarily octal escapes. */
+ case '{':
+ f |= f_brace; ch = getc(tv->fp);
+ /* fall through */
default:
if ('0' <= ch && ch < '8') {
i = 1; esc = ch - '0';
esc = 8*esc + ch - '0';
i++; if (i >= 3) break;
}
+ if (f&f_brace) {
+ ch = getc(tv->fp);
+ if (ch != '}') return (tvec_syntax(tv, ch, "`}'"));
+ }
if (esc > UCHAR_MAX)
return (tvec_error(tv,
"character code %d out of range", esc));
- *ch_out = esc;
+ *ch_out = esc; break;
} else
return (tvec_syntax(tv, ch, "string escape"));
}
+ /* Done. */
return (0);
#undef f_brace
}
+/* --- @read_quoted_string@ --- *
+ *
+ * Arguments: @dstr *d@ = string to write to
+ * @int quote@ = initial quote, `'' or `"'
+ * @struct tvec_state *tv@ = test vector state
+ *
+ * Returns: Zero on success, @-1@ on error.
+ *
+ * Use: Read the rest of a quoted string into @d@, reporting errors
+ * as appropriate.
+ *
+ * A single-quoted string is entirely literal. A double-quoted
+ * string may contain C-like escapes.
+ */
+
static int read_quoted_string(dstr *d, int quote, struct tvec_state *tv)
{
int ch;
case '\\':
if (quote == '\'') goto ordinary;
ch = getc(tv->fp); if (ch == '\n') { tv->lno++; break; }
- ungetc(ch, tv->fp); if (read_escape(&ch, tv)) return (-1);
+ ungetc(ch, tv->fp); if (read_charesc(&ch, tv)) return (-1);
goto ordinary;
default:
if (ch == quote) goto end;
return (0);
}
-#define FCF_BRACE 1u
-static void format_charesc(const struct gprintf_ops *gops, void *go,
- int ch, unsigned f)
-{
- switch (ch) {
- case '\a': gprintf(gops, go, "\\a"); break;
- case '\b': gprintf(gops, go, "\\b"); break;
- case '\x1b': gprintf(gops, go, "\\e"); break;
- case '\f': gprintf(gops, go, "\\f"); break;
- case '\r': gprintf(gops, go, "\\r"); break;
- case '\n': gprintf(gops, go, "\\n"); break;
- case '\t': gprintf(gops, go, "\\t"); break;
- case '\v': gprintf(gops, go, "\\v"); break;
- case '\'': gprintf(gops, go, "\\'"); break;
- case '"': gprintf(gops, go, "\\\""); break;
- default:
- if (f&FCF_BRACE)
- gprintf(gops, go, "\\x{%0*x}", hex_width(UCHAR_MAX), ch);
- else
- gprintf(gops, go, "\\x%0*x", hex_width(UCHAR_MAX), ch);
- break;
- }
-}
-
-static void format_char(const struct gprintf_ops *gops, void *go, int ch)
-{
- if (ch == EOF)
- gprintf(gops, go, "#eof");
- else if (isprint(ch) && ch != '\'')
- gprintf(gops, go, "'%c'", ch);
- else {
- gprintf(gops, go, "'");
- format_charesc(gops, go, ch, 0);
- gprintf(gops, go, "'");
- }
-}
-
-static void maybe_format_signed_char
- (const struct gprintf_ops *gops, void *go, long i)
-{
- if (i == EOF || (0 <= i && i < UCHAR_MAX))
- { gprintf(gops, go, " = "); format_char(gops, go, i); }
-}
-
-static void maybe_format_unsigned_char
- (const struct gprintf_ops *gops, void *go, unsigned long u)
-{
- if (u < UCHAR_MAX)
- { gprintf(gops, go, " = "); format_char(gops, go, u); }
-}
-
-enum { TVCODE_BARE, TVCODE_HEX, TVCODE_BASE64, TVCODE_BASE32 };
+/* --- @collect_bare@ --- *
+ *
+ * Arguments: @dstr *d@ = string to write to
+ * @struct tvec_state *tv@ = test vector state
+ *
+ * Returns: Zero on success, @-1@ on error.
+ *
+ * Use: Read barewords and the whitespace between them. Stop when we
+ * encounter something which can't start a bareword.
+ */
static int collect_bare(dstr *d, struct tvec_state *tv)
{
ungetc(ch, tv->fp); if (tvec_nexttoken(tv)) { rc = -1; goto end; }
DPUTC(d, ' '); s = SPACE;
break;
- case '"': case '\'': case '!':
+ case '"': case '\'': case '!': case '#': case ')': case '}': case ']':
if (s == SPACE) { ungetc(ch, tv->fp); goto done; }
goto addch;
case '\\':
return (rc);
}
+/* --- @set_up_encoding@ --- *
+ *
+ * Arguments: @const codec_class **ccl_out@ = where to put the class
+ * @unsigned *f_out@ = where to put the flags
+ * @unsigned code@ = the coding scheme to use (@TVEC_...@)
+ *
+ * Returns: ---
+ *
+ * Use: Helper for @read_compound_string@ below.
+ *
+ * Return the appropriate codec class and flags for @code@.
+ * Leaves @*ccl_out@ null if the coding scheme doesn't have a
+ * backing codec class (e.g., @TVCODE_BARE@).
+ */
+
+enum { TVCODE_BARE, TVCODE_HEX, TVCODE_BASE64, TVCODE_BASE32 };
static void set_up_encoding(const codec_class **ccl_out, unsigned *f_out,
unsigned code)
{
}
}
+/* --- @flush_codec@ --- *
+ *
+ * Arguments: @codec *cdc@ = a codec, or null
+ * @dstr *d@ = output string
+ * @struct tvec_state *tv@ = test vector state
+ *
+ * Returns: Zero on success, @-1@ on error.
+ *
+ * Use: Helper for @read_compound_string@ below.
+ *
+ * Flush out any final buffered material from @cdc@, and check
+ * that it's in a good state. Frees the codec on success. Does
+ * nothing if @cdc@ is null.
+ */
+
+static int flush_codec(codec *cdc, dstr *d, struct tvec_state *tv)
+{
+ int err;
+
+ if (cdc) {
+ err = cdc->ops->code(cdc, 0, 0, d);
+ if (err)
+ return (tvec_error(tv, "invalid %s sequence end: %s",
+ cdc->ops->c->name, codec_strerror(err)));
+ cdc->ops->destroy(cdc);
+ }
+ return (0);
+}
+
+/* --- @read_compound_string@ --- *
+ *
+ * Arguments: @void **p_inout@ = address of output buffer pointer
+ * @size_t *sz_inout@ = address of buffer size
+ * @unsigned code@ = initial interpretation of barewords
+ * @unsigned f@ = other flags (@RCSF_...@)
+ * @struct tvec_state *tv@ = test vector state
+ *
+ * Returns: Zero on success, @-1@ on error.
+ *
+ * Use: Parse a compound string, i.e., a sequence of stringish pieces
+ * which might be quoted strings, character names, or barewords
+ * to be decoded accoding to @code@, interspersed with
+ * additional directives.
+ *
+ * If the initial buffer pointer is non-null and sufficiently
+ * large, then it will be reused; otherwise, it is freed and a
+ * fresh, sufficiently large buffer is allocated and returned.
+ */
+
+#define RCSF_NESTED 1u
static int read_compound_string(void **p_inout, size_t *sz_inout,
- unsigned code, struct tvec_state *tv)
+ unsigned code, unsigned f,
+ struct tvec_state *tv)
{
- const codec_class *ccl; unsigned f;
+ const codec_class *ccl; unsigned cdf;
codec *cdc;
dstr d = DSTR_INIT, w = DSTR_INIT;
char *p;
+ const char *q;
+ void *pp = 0; size_t sz;
+ unsigned long n;
int ch, err, rc;
- set_up_encoding(&ccl, &f, code);
- if (tvec_nexttoken(tv)) tvec_syntax(tv, fgetc(tv->fp), "string");
+ set_up_encoding(&ccl, &cdf, code); cdc = 0;
+
+ if (tvec_nexttoken(tv)) return (tvec_syntax(tv, fgetc(tv->fp), "string"));
do {
ch = getc(tv->fp);
- if (ch == '"' || ch == '\'')
- { if (read_quoted_string(&d, ch, tv)) { rc = -1; goto end; } }
- else if (ch == '!') {
- ungetc(ch, tv->fp);
- DRESET(&w); tvec_readword(tv, &w, ";", "`!'-keyword");
- if (STRCMP(w.buf, ==, "!bare")) code = TVCODE_BARE;
- else if (STRCMP(w.buf, ==, "!hex")) code = TVCODE_HEX;
- else if (STRCMP(w.buf, ==, "!base32")) code = TVCODE_BASE32;
- else if (STRCMP(w.buf, ==, "!base64")) code = TVCODE_BASE64;
- else {
- tvec_error(tv, "unknown string keyword `%s'", w.buf);
- rc = -1; goto end;
- }
- set_up_encoding(&ccl, &f, code);
- } else if (ccl) {
- ungetc(ch, tv->fp);
- DRESET(&w);
- if (tvec_readword(tv, &w, ";", "%s-encoded fragment", ccl->name))
- { rc = -1; goto end; }
- cdc = ccl->decoder(f);
- err = cdc->ops->code(cdc, w.buf, w.len, &d);
- if (!err) err = cdc->ops->code(cdc, 0, 0, &d);
- if (err) {
- tvec_error(tv, "invalid %s fragment `%s': %s",
- ccl->name, w.buf, codec_strerror(err));
- rc = -1; goto end;
- }
- cdc->ops->destroy(cdc);
- } else switch (code) {
- case TVCODE_BARE:
+ switch (ch) {
+
+ case ')': case ']': case '}':
+ /* Close brackets. Leave these for recursive caller if there is one,
+ * or just complain.
+ */
+
+ if (!(f&RCSF_NESTED))
+ { rc = tvec_syntax(tv, ch, "string"); goto end; }
+ ungetc(ch, tv->fp); goto done;
+
+ case '"': case '\'':
+ /* Quotes. Read a quoted string. */
+
+ if (cdc && flush_codec(cdc, &d, tv)) { rc = -1; goto end; }
+ cdc = 0;
+ if (read_quoted_string(&d, ch, tv)) { rc = -1; goto end; }
+ break;
+
+ case '#':
+ /* A named character. */
+
+ ungetc(ch, tv->fp);
+ if (cdc && flush_codec(cdc, &d, tv)) { rc = -1; goto end; }
+ cdc = 0;
+ DRESET(&w); tvec_readword(tv, &w, ";", "character name");
+ if (read_charname(&ch, w.buf, RCF_EOFOK)) {
+ rc = tvec_error(tv, "unknown character name `%s'", d.buf);
+ goto end;
+ }
+ DPUTC(&d, ch); break;
+
+ case '!':
+ /* A magic keyword. */
+
+ if (cdc && flush_codec(cdc, &d, tv)) { rc = -1; goto end; }
+ cdc = 0;
ungetc(ch, tv->fp);
- if (collect_bare(&d, tv)) goto done;
+ DRESET(&w); tvec_readword(tv, &w, ";", "`!'-keyword");
+
+ /* Change bareword coding system. */
+ if (STRCMP(w.buf, ==, "!bare"))
+ { code = TVCODE_BARE; set_up_encoding(&ccl, &cdf, code); }
+ else if (STRCMP(w.buf, ==, "!hex"))
+ { code = TVCODE_HEX; set_up_encoding(&ccl, &cdf, code); }
+ else if (STRCMP(w.buf, ==, "!base32"))
+ { code = TVCODE_BASE32; set_up_encoding(&ccl, &cdf, code); }
+ else if (STRCMP(w.buf, ==, "!base64"))
+ { code = TVCODE_BASE64; set_up_encoding(&ccl, &cdf, code); }
+
+ /* Repeated substrings. */
+ else if (STRCMP(w.buf, ==, "!repeat")) {
+ if (tvec_nexttoken(tv)) {
+ rc = tvec_syntax(tv, fgetc(tv->fp), "repeat count");
+ goto end;
+ }
+ DRESET(&w);
+ if (tvec_readword(tv, &w, ";{", "repeat count"))
+ { rc = -1; goto end; }
+ if (parse_unsigned_integer(&n, &q, w.buf)) {
+ rc = tvec_error(tv, "invalid repeat count `%s'", w.buf);
+ goto end;
+ }
+ if (*q) { rc = tvec_syntax(tv, *q, "`{'"); goto end; }
+ if (tvec_nexttoken(tv))
+ { rc = tvec_syntax(tv, fgetc(tv->fp), "`{'"); goto end; }
+ ch = getc(tv->fp); if (ch != '{')
+ { rc = tvec_syntax(tv, ch, "`{'"); goto end; }
+ sz = 0;
+ if (read_compound_string(&pp, &sz, code, f | RCSF_NESTED, tv))
+ { rc = -1; goto end; }
+ ch = getc(tv->fp); if (ch != '}')
+ { rc = tvec_syntax(tv, ch, "`}'"); goto end; }
+ if (sz) {
+ if (n > (size_t)-1/sz)
+ { rc = tvec_error(tv, "repeat size out of range"); goto end; }
+ dstr_ensure(&d, n*sz);
+ if (sz == 1)
+ { memset(d.buf + d.len, *(unsigned char *)pp, n); d.len += n; }
+ else
+ for (; n--; d.len += sz) memcpy(d.buf + d.len, pp, sz);
+ }
+ xfree(pp); pp = 0;
+ }
+
+ /* Anything else is an error. */
+ else {
+ tvec_error(tv, "unknown string keyword `%s'", w.buf);
+ rc = -1; goto end;
+ }
break;
+
default:
- abort();
+ /* A bareword. Process it according to the current coding system. */
+
+ switch (code) {
+ case TVCODE_BARE:
+ ungetc(ch, tv->fp);
+ if (collect_bare(&d, tv)) goto done;
+ break;
+ default:
+ assert(ccl);
+ ungetc(ch, tv->fp); DRESET(&w);
+ if (tvec_readword(tv, &w, ";", "%s-encoded fragment", ccl->name))
+ { rc = -1; goto end; }
+ if (!cdc) cdc = ccl->decoder(cdf);
+ err = cdc->ops->code(cdc, w.buf, w.len, &d);
+ if (err) {
+ tvec_error(tv, "invalid %s fragment `%s': %s",
+ ccl->name, w.buf, codec_strerror(err));
+ rc = -1; goto end;
+ }
+ break;
+ }
+ break;
}
} while (!tvec_nexttoken(tv));
done:
+ /* Wrap things up. */
+ if (cdc && flush_codec(cdc, &d, tv)) { rc = -1; goto end; }
+ cdc = 0;
if (*sz_inout <= d.len)
{ xfree(*p_inout); *p_inout = xmalloc(d.len + 1); }
p = *p_inout; memcpy(p, d.buf, d.len); p[d.len] = 0; *sz_inout = d.len;
rc = 0;
+
end:
+ /* Clean up any debris. */
+ if (cdc) cdc->ops->destroy(cdc);
+ if (pp) xfree(pp);
dstr_destroy(&d); dstr_destroy(&w);
return (rc);
}
tvrange_u16 = { 0, 65535 },
tvrange_u32 = { 0, 4294967296 };
+/* --- @tvec_claimeq_int@ --- *
+ *
+ * Arguments: @struct tvec_state *tv@ = test-vector state
+ * @long i0, i1@ = two signed integers
+ * @const char *file@, @unsigned @lno@ = calling file and line
+ * @const char *expr@ = the expression to quote on failure
+ *
+ * Returns: Nonzero if @i0@ and @i1@ are equal, otherwise zero.
+ *
+ * Use: Check that values of @i0@ and @i1@ are equal. As for
+ * @tvec_claim@ above, a test case is automatically begun and
+ * ended if none is already underway. If the values are
+ * unequal, then @tvec_fail@ is called, quoting @expr@, and the
+ * mismatched values are dumped: @i0@ is printed as the output
+ * value and @i1@ is printed as the input reference.
+ */
+
int tvec_claimeq_int(struct tvec_state *tv, long i0, long i1,
const char *file, unsigned lno, const char *expr)
{
return (tvec_claimeq(tv, &tvty_int, 0, file, lno, expr));
}
+/* --- @tvec_claimeq_uint@ --- *
+ *
+ * Arguments: @struct tvec_state *tv@ = test-vector state
+ * @unsigned long u0, u1@ = two unsigned integers
+ * @const char *file@, @unsigned @lno@ = calling file and line
+ * @const char *expr@ = the expression to quote on failure
+ *
+ * Returns: Nonzero if @u0@ and @u1@ are equal, otherwise zero.
+ *
+ * Use: Check that values of @u0@ and @u1@ are equal. As for
+ * @tvec_claim@ above, a test case is automatically begun and
+ * ended if none is already underway. If the values are
+ * unequal, then @tvec_fail@ is called, quoting @expr@, and the
+ * mismatched values are dumped: @u0@ is printed as the output
+ * value and @u1@ is printed as the input reference.
+ */
+
int tvec_claimeq_uint(struct tvec_state *tv,
unsigned long u0, unsigned long u1,
const char *file, unsigned lno, const char *expr)
parse_float, dump_float
};
-int tvec_claimeq_float(struct tvec_state *tv,
- double f0, double f1,
- const char *file, unsigned lno,
- const char *expr)
-{
- return (tvec_claimeqish_float(tv, f0, f1, TVFF_EXACT, 0.0,
- file, lno, expr));
-}
+/* --- @tvec_claimeqish_float@ --- *
+ *
+ * Arguments: @struct tvec_state *tv@ = test-vector state
+ * @double f0, f1@ = two floating-point numbers
+ * @unsigned f@ = flags (@TVFF_...@)
+ * @double delta@ = maximum tolerable difference
+ * @const char *file@, @unsigned @lno@ = calling file and line
+ * @const char *expr@ = the expression to quote on failure
+ *
+ * Returns: Nonzero if @f0@ and @u1@ are sufficiently close, otherwise
+ * zero.
+ *
+ * Use: Check that values of @f0@ and @f1@ are sufficiently close.
+ * As for @tvec_claim@ above, a test case is automatically begun
+ * and ended if none is already underway. If the values are
+ * too far apart, then @tvec_fail@ is called, quoting @expr@,
+ * and the mismatched values are dumped: @f0@ is printed as the
+ * output value and @f1@ is printed as the input reference.
+ *
+ * The details for the comparison are as follows.
+ *
+ * * A NaN value matches any other NaN, and nothing else.
+ *
+ * * An infinity matches another infinity of the same sign,
+ * and nothing else.
+ *
+ * * If @f&TVFF_EQMASK@ is @TVFF_EXACT@, then any
+ * representable number matches only itself: in particular,
+ * positive and negative zero are considered distinct.
+ * (This allows tests to check that they land on the correct
+ * side of branch cuts, for example.)
+ *
+ * * If @f&TVFF_EQMASK@ is @TVFF_ABSDELTA@, then %$x$% matches
+ * %$y$% when %$|x - y| < \delta$%.
+ *
+ * * If @f&TVFF_EQMASK@ is @TVFF_RELDELTA@, then %$x$% matches
+ * %$y$% when %$|1 - y/x| < \delta$%. (Note that this
+ * criterion is asymmetric FIXME
+ */
+
int tvec_claimeqish_float(struct tvec_state *tv,
double f0, double f1, unsigned f, double delta,
const char *file, unsigned lno,
return (tvec_claimeq(tv, &tvty_float, &arg, file, lno, expr));
}
+/* --- @tvec_claimeq_float@ --- *
+ *
+ * Arguments: @struct tvec_state *tv@ = test-vector state
+ * @double f0, f1@ = two floating-point numbers
+ * @const char *file@, @unsigned @lno@ = calling file and line
+ * @const char *expr@ = the expression to quote on failure
+ *
+ * Returns: Nonzero if @f0@ and @u1@ are identical, otherwise zero.
+ *
+ * Use: Check that values of @f0@ and @f1@ are identical. The
+ * function is exactly equivalent to @tvec_claimeqish_float@
+ * with @f == TVFF_EXACT@.
+ */
+
+int tvec_claimeq_float(struct tvec_state *tv,
+ double f0, double f1,
+ const char *file, unsigned lno,
+ const char *expr)
+{
+ return (tvec_claimeqish_float(tv, f0, f1, TVFF_EXACT, 0.0,
+ file, lno, expr));
+}
+
/*----- Enumerations ------------------------------------------------------*/
#define init_ienum init_int
const struct tvec_ienuminfo tvenum_bool =
{ "bool", bool_assoc, &tvrange_int };
+/* --- @tvec_claimeq_tenum@ --- *
+ *
+ * Arguments: @struct tvec_state *tv@ = test-vector state
+ * @const struct tvec_typeenuminfo *ei@ = enumeration type info
+ * @ty t0, t1@ = two values
+ * @const char *file@, @unsigned @lno@ = calling file and line
+ * @const char *expr@ = the expression to quote on failure
+ *
+ * Returns: Nonzero if @t0@ and @t1@ are equal, otherwise zero.
+ *
+ * Use: Check that values of @t0@ and @t1@ are equal. As for
+ * @tvec_claim@ above, a test case is automatically begun and
+ * ended if none is already underway. If the values are
+ * unequal, then @tvec_fail@ is called, quoting @expr@, and the
+ * mismatched values are dumped: @t0@ is printed as the output
+ * value and @t1@ is printed as the input reference.
+ */
+
#define DEFCLAIM(tag, ty, slot) \
int tvec_claimeq_##slot##enum \
(struct tvec_state *tv, \
parse_flags, dump_flags
};
+/* --- @tvec_claimeq_flags@ --- *
+ *
+ * Arguments: @struct tvec_state *tv@ = test-vector state
+ * @const struct tvec_flaginfo *fi@ = flags type info
+ * @unsigned long f0, f1@ = two values
+ * @const char *file@, @unsigned @lno@ = calling file and line
+ * @const char *expr@ = the expression to quote on failure
+ *
+ * Returns: Nonzero if @f0@ and @f1@ are equal, otherwise zero.
+ *
+ * Use: Check that values of @f0@ and @f1@ are equal. As for
+ * @tvec_claim@ above, a test case is automatically begun and
+ * ended if none is already underway. If the values are
+ * unequal, then @tvec_fail@ is called, quoting @expr@, and the
+ * mismatched values are dumped: @f0@ is printed as the output
+ * value and @f1@ is printed as the input reference.
+ */
+
int tvec_claimeq_flags(struct tvec_state *tv,
const struct tvec_flaginfo *fi,
unsigned long f0, unsigned long f1,
/*----- Characters --------------------------------------------------------*/
static int tobuf_char(buf *b, const union tvec_regval *rv,
- const struct tvec_regdef *rd)
+ const struct tvec_regdef *rd)
{
uint32 u;
if (0 <= rv->i && rv->i <= UCHAR_MAX) u = rv->i;
}
static int frombuf_char(buf *b, union tvec_regval *rv,
- const struct tvec_regdef *rd)
+ const struct tvec_regdef *rd)
{
uint32 u;
if (ch == '#') {
ungetc(ch, tv->fp);
if (tvec_readword(tv, &d, ";", "character name")) { rc = -1; goto end; }
- if (STRCMP(d.buf, ==, "#eof"))
- rv->i = EOF;
- else {
+ if (read_charname(&ch, d.buf, RCF_EOFOK)) {
rc = tvec_error(tv, "unknown character name `%s'", d.buf);
goto end;
}
- rc = 0; goto end;
+ if (tvec_flushtoeol(tv, 0)) { rc = -1; goto end; }
+ rv->i = ch; rc = 0; goto end;
}
if (ch == '\'') { f |= f_quote; ch = getc(tv->fp); }
switch (ch) {
+ case ';':
+ if (!(f&f_quote)) { rc = tvec_syntax(tv, ch, "character"); goto end; }
+ goto plain;
+ case '\n':
+ if (f&f_quote)
+ { f &= ~f_quote; ungetc(ch, tv->fp); ch = '\''; goto plain; }
+ case EOF:
+ if (f&f_quote) { f &= ~f_quote; ch = '\''; goto plain; }
+ /* fall through */
case '\'':
- if (!(f&f_quote)) goto plain;
- case EOF: case '\n':
rc = tvec_syntax(tv, ch, "character"); goto end;
case '\\':
- if (read_escape(&ch, tv)) return (-1);
+ if (read_charesc(&ch, tv)) return (-1);
default: plain:
rv->i = ch; break;
}
unsigned style,
const struct gprintf_ops *gops, void *go)
{
- if ((style&TVSF_COMPACT) && isprint(rv->i) && rv->i != '\'')
- gprintf(gops, go, "%c", (int)rv->i);
- else
- format_char(gops, go, rv->i);
+ const char *p;
+ unsigned f = 0;
+#define f_semi 1u
+
+ p = find_charname(rv->i, (style&TVSF_COMPACT) ? CTF_SHORT : CTF_PREFER);
+ if (p) {
+ gprintf(gops, go, "%s", p);
+ if (style&TVSF_COMPACT) return;
+ else { gprintf(gops, go, " ;"); f |= f_semi; }
+ }
+
+ if (rv->i >= 0) {
+ if (f&f_semi) gprintf(gops, go, " = ");
+ switch (rv->i) {
+ case ' ': case '\\': case '\'': quote:
+ format_char(gops, go, rv->i);
+ break;
+ default:
+ if (!(style&TVSF_COMPACT) || !isprint(rv->i)) goto quote;
+ gprintf(gops, go, "%c", (int)rv->i);
+ return;
+ }
+ }
if (!(style&TVSF_COMPACT)) {
- gprintf(gops, go, " ; = %ld = ", rv->i);
+ if (!(f&f_semi)) gprintf(gops, go, " ;");
+ gprintf(gops, go, " = %ld = ", rv->i);
format_signed_hex(gops, go, rv->i);
}
+
+#undef f_semi
}
const struct tvec_regty tvty_char = {
parse_char, dump_char
};
+/* --- @tvec_claimeq_char@ --- *
+ *
+ * Arguments: @struct tvec_state *tv@ = test-vector state
+ * @int ch0, ch1@ = two character codes
+ * @const char *file@, @unsigned @lno@ = calling file and line
+ * @const char *expr@ = the expression to quote on failure
+ *
+ * Returns: Nonzero if @ch0@ and @ch1@ are equal, otherwise zero.
+ *
+ * Use: Check that values of @ch0@ and @ch1@ are equal. As for
+ * @tvec_claim@ above, a test case is automatically begun and
+ * ended if none is already underway. If the values are
+ * unequal, then @tvec_fail@ is called, quoting @expr@, and the
+ * mismatched values are dumped: @ch0@ is printed as the output
+ * value and @ch1@ is printed as the input reference.
+ */
+
int tvec_claimeq_char(struct tvec_state *tv, int c0, int c1,
const char *file, unsigned lno, const char *expr)
{
/*----- Text and byte strings ---------------------------------------------*/
-void tvec_allocstring(union tvec_regval *rv, size_t sz)
-{
- if (rv->str.sz < sz) { xfree(rv->str.p); rv->str.p = xmalloc(sz); }
- rv->str.sz = sz;
-}
-
-void tvec_allocbytes(union tvec_regval *rv, size_t sz)
-{
- if (rv->bytes.sz < sz) { xfree(rv->bytes.p); rv->bytes.p = xmalloc(sz); }
- rv->bytes.sz = sz;
-}
-
static void init_string(union tvec_regval *rv, const struct tvec_regdef *rd)
{ rv->str.p = 0; rv->str.sz = 0; }
size_t sz;
p = buf_getmem32l(b, &sz); if (!p) return (-1);
- tvec_allocstring(rv, sz); memcpy(rv->str.p, p, sz);
+ tvec_allocstring(rv, sz); memcpy(rv->str.p, p, sz); rv->str.p[sz] = 0;
return (0);
}
{
if (ur && (ur->min > sz || sz > ur->max))
return (tvec_error(tv,
- "invalid string length %lu; must be in [%lu..%lu]",
+ "invalid string length %lu; must be in [%lu .. %lu]",
(unsigned long)sz, ur->min, ur->max));
return (0);
}
{
void *p = rv->str.p;
- if (read_compound_string(&p, &rv->str.sz, TVCODE_BARE, tv)) return (-1);
+ if (read_compound_string(&p, &rv->str.sz, TVCODE_BARE, 0, tv))
+ return (-1);
rv->str.p = p;
if (check_string_length(rv->str.sz, rd->arg.p, tv)) return (-1);
return (0);
{
void *p = rv->bytes.p;
- if (read_compound_string(&p, &rv->bytes.sz, TVCODE_HEX, tv)) return (-1);
+ if (read_compound_string(&p, &rv->bytes.sz, TVCODE_HEX, 0, tv))
+ return (-1);
rv->bytes.p = p;
if (check_string_length(rv->bytes.sz, rd->arg.p, tv)) return (-1);
return (0);
if (!rv->str.sz) { gprintf(gops, go, "\"\""); return; }
p = (const unsigned char *)rv->str.p; l = p + rv->str.sz;
- if (*p == '!' || *p == ';' || *p == '"' || *p == '\'') goto quote;
+ switch (*p) {
+ case '!': case '#': case ';': case '"': case '\'':
+ case '(': case '{': case '[': case ']': case '}': case ')':
+ f |= f_nonword; break;
+ }
for (q = p; q < l; q++)
if (*q == '\n' && q != l - 1) f |= f_newline;
else if (!*q || !isgraph(*q) || *q == '\\') f |= f_nonword;
if (f&f_newline) { gprintf(gops, go, "\n\t"); goto quote; }
else if (f&f_nonword) goto quote;
- gops->putm(go, (const char *)p, rv->str.sz); return;
+
+ gops->putm(go, (const char *)p, rv->str.sz);
+ return;
quote:
gprintf(gops, go, "\"");
for (q = p; q < l; q++)
if (!isprint(*q) || *q == '"') {
if (p < q) gops->putm(go, (const char *)p, q - p);
- if (*q == '\n' && !(style&TVSF_COMPACT))
- gprintf(gops, go, "\\n\"\t\"");
- else
+ if (*q != '\n' || (style&TVSF_COMPACT))
format_charesc(gops, go, *q, FCF_BRACE);
+ else {
+ if (q + 1 == l) { gprintf(gops, go, "\\n\""); return; }
+ else gprintf(gops, go, "\\n\"\n\t\"");
+ }
+ p = q + 1;
}
if (p < q) gops->putm(go, (const char *)p, q - p);
gprintf(gops, go, "\"");
if (l - p < 16) n = l - p;
else n = 16;
- for (i = 0; i < 16; i++) {
+ for (i = 0; i < n; i++) {
if (i < n) gprintf(gops, go, "%02x", p[i]);
else gprintf(gops, go, " ");
- if (i%4 == 3) gprintf(gops, go, " ");
+ if (i < n - 1 && i%4 == 3) gprintf(gops, go, " ");
}
gprintf(gops, go, " ; ");
if (sz > 16) gprintf(gops, go, "[%0*lx] ", wd, (unsigned long)off);
parse_bytes, dump_bytes
};
+/* --- @tvec_claimeq_string@ --- *
+ *
+ * Arguments: @struct tvec_state *tv@ = test-vector state
+ * @const char *p0@, @size_t sz0@ = first string with length
+ * @const char *p1@, @size_t sz1@ = second string with length
+ * @const char *file@, @unsigned @lno@ = calling file and line
+ * @const char *expr@ = the expression to quote on failure
+ *
+ * Returns: Nonzero if the strings at @p0@ and @p1@ are equal, otherwise
+ * zero.
+ *
+ * Use: Check that strings at @p0@ and @p1@ are equal. As for
+ * @tvec_claim@ above, a test case is automatically begun and
+ * ended if none is already underway. If the values are
+ * unequal, then @tvec_fail@ is called, quoting @expr@, and the
+ * mismatched values are dumped: @p0@ is printed as the output
+ * value and @p1@ is printed as the input reference.
+ */
+
int tvec_claimeq_string(struct tvec_state *tv,
const char *p0, size_t sz0,
const char *p1, size_t sz1,
return (tvec_claimeq(tv, &tvty_string, 0, file, lno, expr));
}
+/* --- @tvec_claimeq_strz@ --- *
+ *
+ * Arguments: @struct tvec_state *tv@ = test-vector state
+ * @const char *p0, *p1@ = two strings to compare
+ * @const char *file@, @unsigned @lno@ = calling file and line
+ * @const char *expr@ = the expression to quote on failure
+ *
+ * Returns: Nonzero if the strings at @p0@ and @p1@ are equal, otherwise
+ * zero.
+ *
+ * Use: Check that strings at @p0@ and @p1@ are equal, as for
+ * @tvec_claimeq_string@, except that the strings are assumed
+ * null-terminated, so their lengths don't need to be supplied
+ * explicitly.
+ */
+
int tvec_claimeq_strz(struct tvec_state *tv,
const char *p0, const char *p1,
const char *file, unsigned lno, const char *expr)
return (tvec_claimeq(tv, &tvty_string, 0, file, lno, expr));
}
+/* --- @tvec_claimeq_bytes@ --- *
+ *
+ * Arguments: @struct tvec_state *tv@ = test-vector state
+ * @const void *p0@, @size_t sz0@ = first string with length
+ * @const void *p1@, @size_t sz1@ = second string with length
+ * @const char *file@, @unsigned @lno@ = calling file and line
+ * @const char *expr@ = the expression to quote on failure
+ *
+ * Returns: Nonzero if the strings at @p0@ and @p1@ are equal, otherwise
+ * zero.
+ *
+ * Use: Check that binary strings at @p0@ and @p1@ are equal. As for
+ * @tvec_claim@ above, a test case is automatically begun and
+ * ended if none is already underway. If the values are
+ * unequal, then @tvec_fail@ is called, quoting @expr@, and the
+ * mismatched values are dumped: @p0@ is printed as the output
+ * value and @p1@ is printed as the input reference.
+ */
+
int tvec_claimeq_bytes(struct tvec_state *tv,
const void *p0, size_t sz0,
const void *p1, size_t sz1,
return (tvec_claimeq(tv, &tvty_bytes, 0, file, lno, expr));
}
+/* --- @tvec_allocstring@, @tvec_allocbytes@ --- *
+ *
+ * Arguments: @union tvec_regval *rv@ = register value
+ * @size_t sz@ = required size
+ *
+ * Returns: ---
+ *
+ * Use: Allocated space in a text or binary string register. If the
+ * current register size is sufficient, its buffer is left
+ * alone; otherwise, the old buffer, if any, is freed and a
+ * fresh buffer allocated. These functions are not intended to
+ * be used to adjust a buffer repeatedly, e.g., while building
+ * output incrementally: (a) they will perform badly, and (b)
+ * the old buffer contents are simply discarded if reallocation
+ * is necessary. Instead, use a @dbuf@ or @dstr@.
+ *
+ * The @tvec_allocstring@ function sneakily allocates an extra
+ * byte for a terminating zero. The @tvec_allocbytes@ function
+ * doesn't do this.
+ */
+
+void tvec_allocstring(union tvec_regval *rv, size_t sz)
+{
+ if (rv->str.sz <= sz) { xfree(rv->str.p); rv->str.p = xmalloc(sz + 1); }
+ rv->str.sz = sz;
+}
+
+void tvec_allocbytes(union tvec_regval *rv, size_t sz)
+{
+ if (rv->bytes.sz < sz) { xfree(rv->bytes.p); rv->bytes.p = xmalloc(sz); }
+ rv->bytes.sz = sz;
+}
+
/*----- Buffer type -------------------------------------------------------*/
static int eq_buffer(const union tvec_regval *rv0,
for (t = u, unit = units; *unit; unit++) {
if (t > (size_t)-1/1024) f |= f_range;
else t *= 1024;
- if (*q == *unit && (!q[1] || q[1] == 'B')) {
+ if (*q == *unit) {
if (f&f_range) goto rangerr;
- u = t; q += 2; break;
+ u = t; q++; break;
}
}
- if (*q && *q != ';') goto bad;
+ if (*q == 'B') q++;
+ if (*q) goto bad;
if (check_string_length(u, rd->arg.p, tv)) { rc = -1; goto end; }
if (tvec_flushtoeol(tv, 0)) { rc = -1; goto end; }
~mdw / mLib / blobdiff