static uint8_t *data;
static size_t n;
+int version;
+
static bool
all_ascii(const uint8_t *p, size_t n)
{
else
match_u32_assert (0);
- /* This special case is really unsatisfying, but there are several
- examples in the corpus, all from one SPV file. */
- int save_pos = pos;
- if (match_u32 (0x200) && match_u32(0x1000000) && match_u32(0)
- && match_byte(0))
- return;
- pos = save_pos;
+ if (version == 1)
+ {
+ /* We only have one SPV file for this version (with many
+ tables). */
+ match_byte(0);
+ if (!match_u32(1))
+ match_u32_assert(2);
+ match_byte(0);
+ match_byte(0);
+ if (!match_u32(0) && !match_u32(1) && !match_u32(2) && !match_u32(3) && !match_u32(4) && !match_u32(5) && !match_u32(6) && !match_u32(7) && !match_u32(8) && !match_u32(9))
+ match_u32_assert(10);
+ match_byte(0);
+ match_byte(0);
+ return;
+ }
int outer_end = pos + get_u32();
int inner_end = pos + get_u32();
match_byte_assert(0x58);
if (match_byte(0x31))
{
- /* All the examples in the corpus, all from one SPV file, are
- exactly like this. */
+ /* Only two SPV files have anything like this, so it's hard to
+ generalize. */
match_u32_assert(0);
match_u32_assert(0);
match_u32_assert(0);
match_u32_assert(0);
match_byte_assert(1);
match_byte_assert(0);
- match_byte_assert(8);
+ if (!match_byte(8) && !match_byte(1))
+ match_byte_assert(2);
match_byte_assert(0);
match_byte_assert(8);
match_byte_assert(0);
else if (match_u32 (2))
{
fprintf(stream, "(special 2)");
+ if (!match_byte(0))
+ match_byte_assert(2);
match_byte_assert(0);
- match_byte_assert(0);
- if (!match_u32 (2))
- match_u32_assert(1);
+ if (!match_u32 (2) && !match_u32(1))
+ match_u32_assert(3);
dump_nested_string(); /* Our corpus doesn't contain any examples with strings though. */
}
else
}
static void
-dump_value(FILE *stream, int level, bool match1)
+dump_value(FILE *stream, int level)
{
match_byte(0);
match_byte(0);
fprintf (stream, "/>\n");
if (!match_byte (0))
match_byte_assert(1);
- if (match1)
- match_byte (1);
+
}
else if (match_byte (5))
{
char *var, *vallab;
double value;
- match_byte_assert (0x58);
+ dump_value_31 (stream);
format = get_u32 ();
value = get_double ();
var = get_string ();
value = get_double ();
fprintf (stream, "<number value=\"%.*g\" format=\"%s%d.%d\"/>\n",
DBL_DIG, value, format_to_string(format >> 16), (format >> 8) & 0xff, format & 0xff);
- if (match1)
- match_byte (1);
}
else
{
fprintf (stream, " ");
fprintf (stream, "<substitution index=\"%d\">\n", i + 1);
for (int j = 0; j < y; j++)
- dump_value (stream, level + 2, false);
+ dump_value (stream, level + 2);
for (int j = 0; j <= level + 1; j++)
fprintf (stream, " ");
fprintf (stream, "</substitution>\n");
}
static void
-dump_category(int level, int *indexes, int *n_indexes)
+dump_category(int level, int *indexes, int *n_indexes, int max_indexes)
{
for (int i = 0; i <= level; i++)
fprintf (stdout, " ");
printf ("<category>\n");
- dump_value (stdout, level + 1, true);
- match_byte(0);
- match_byte(0);
- match_byte(0);
+ dump_value (stdout, level + 1);
+ match_u32(1);
if (match_u32 (1))
match_byte (0);
fprintf(stderr, "index not -1 but subcategories\n");
exit(1);
}
+ if (*n_indexes >= max_indexes)
+ {
+ fprintf(stderr, "too many categories (increase max_indexes)\n");
+ exit(1);
+ }
indexes[(*n_indexes)++] = indx;
}
if (n_categories == 0)
fprintf (stdout, "<category-index>%d</category-index>\n", indx);
}
for (int i = 0; i < n_categories; i++)
- dump_category (level + 1, indexes, n_indexes);
+ dump_category (level + 1, indexes, n_indexes, max_indexes);
for (int i = 0; i <= level; i++)
fprintf (stdout, " ");
printf ("</category>\n");
int n_categories;
printf ("<dimension index=\"%d\">\n", indx);
- dump_value (stdout, 0, false);
+ dump_value (stdout, 0);
- /* This byte is usually 0x02 but 0x00 and 0x75 (!) have also been spotted. */
+ /* This byte is usually 0x02 but many other values have been spotted. */
pos++;
if (!match_byte(0) && !match_byte(1))
if (!match_byte(0))
match_byte_assert(1);
match_byte_assert(1);
- match_u32_assert(indx);
+ if (!match_u32(UINT32_MAX))
+ match_u32_assert(indx);
n_categories = get_u32();
- int indexes[1024];
+ int indexes[2048];
int n_indexes = 0;
for (int i = 0; i < n_categories; i++)
- dump_category (0, indexes, &n_indexes);
+ dump_category (0, indexes, &n_indexes, sizeof indexes / sizeof *indexes);
check_permutation(indexes, n_indexes, "categories");
fprintf (stdout, "</dimension>\n");
{
printf (" <datum index=\"%d\">\n", get_u32());
match_u32_assert(0);
- dump_value(stdout, 1, false);
+ if (version == 1)
+ match_byte(0);
+ dump_value(stdout, 1);
fprintf (stdout, " </datum>\n");
}
printf ("</data>\n");
{
pos = 0x27;
printf ("<title-local>\n");
- dump_value(stdout, 0, true);
+ dump_value(stdout, 0);
+ match_byte(1);
printf ("</title-local>\n");
printf ("<subtype>\n");
- dump_value(stdout, 0, true);
+ dump_value(stdout, 0);
+ match_byte(1);
printf ("</subtype>\n");
match_byte_assert(0x31);
printf ("<title-c>\n");
- dump_value(stdout, 0, true);
+ dump_value(stdout, 0);
+ match_byte(1);
printf ("</title-c>\n");
match_byte(0);
if (match_byte(0x31))
{
printf ("<caption>\n");
- dump_value(stdout, 0, false);
+ dump_value(stdout, 0);
printf ("</caption>\n");
}
else
for (int i = 0; i < n_footnotes; i++)
{
printf ("<footnote index=\"%d\">\n", i);
- dump_value(stdout, 0, false);
+ dump_value(stdout, 0);
if (match_byte (0x31))
{
/* Custom footnote marker string. */
printf(" font=\"%s\"", get_string());
match_byte_assert(0);
match_byte_assert(0);
- if (!match_byte(0x40) && !match_byte(0x20) && !match_byte(0x80) && !match_byte(0x10))
+ if (!match_byte(0x40) && !match_byte(0x20) && !match_byte(0x80) && !match_byte(0x10) && !match_byte(0x70))
match_byte_assert(0x50);
if (!match_byte(0x41))
match_byte_assert(0x51);
- if (!match_u32(0))
- match_u32_assert(1);
+ if (!match_u32(0) && !match_u32(1))
+ match_u32_assert(2);
match_byte_assert(0);
/* OK, this seems really unlikely to be totally correct, but it matches my corpus... */
match_u32_assert(0);
match_byte_assert(0);
- /* These seem unlikely to be correct too. */
- if (i != 3)
- {
- match_u32_assert(8);
- if (!match_u32(10))
- match_u32_assert(11);
- match_u32_assert(1);
- }
- else
+ if (version > 1)
{
+ /* These seem unlikely to be correct too. */
+ if (i != 3)
+ {
+ if (!match_u32(8))
+ match_u32_assert(5);
+ if (!match_u32(10) && !match_u32(11) && !match_u32(5))
+ match_u32_assert(9);
+ if (!match_u32(0))
+ match_u32_assert(1);
+ }
+ else
+ {
+ get_u32();
+ if (!match_u32(-1) && !match_u32(8))
+ match_u32_assert(24);
+ if (!match_u32(-1) && !match_u32(2))
+ match_u32_assert(3);
+ }
+
+ /* Who knows? Ranges from -1 to 8 with no obvious pattern. */
get_u32();
- if (!match_u32(-1) && !match_u32(8))
- match_u32_assert(24);
- if (!match_u32(-1) && !match_u32(2))
- match_u32_assert(3);
}
- /* Who knows? Ranges from -1 to 8 with no obvious pattern. */
- get_u32();
-
printf ("/>\n");
}
if (match_u32(117))
pos += 117;
+ else if (match_u32(142))
+ pos += 142;
+ else if (match_u32(143))
+ pos += 143;
+ else if (match_u32(150))
+ pos += 150;
else
{
- match_u32_assert(142);
- pos += 142;
+ match_u32_assert(16);
+ pos += 16;
}
int count = get_u32();
match_byte_assert(0);
if (!match_byte(0))
match_byte_assert(1);
- if (!match_byte(0x97) && !match_byte(0x98) && !match_byte(0x99))
- match_byte_assert(0x9a);
- match_byte_assert(7);
- match_byte_assert(0);
- match_byte_assert(0);
+ if (version > 1)
+ {
+ if (!match_byte(0x97) && !match_byte(0x98) && !match_byte(0x99))
+ match_byte_assert(0x9a);
+ match_byte_assert(7);
+ match_byte_assert(0);
+ match_byte_assert(0);
+ }
+ else
+ match_u32_assert(UINT32_MAX);
if (match_byte('.'))
- match_byte_assert(',');
+ {
+ if (!match_byte(','))
+ match_byte_assert(' ');
+ }
else
{
match_byte_assert(',');
- if (!match_byte('.'))
- match_byte_assert(' ');
+ if (!match_byte('.') && !match_byte(' '))
+ match_byte_assert(0);
}
- match_u32_assert(5);
- for (int i = 0; i < 5; i++)
- get_string();
- pos += get_u32();
+ if (match_u32(5))
+ {
+ for (int i = 0; i < 5; i++)
+ get_string();
+ }
+ else
+ match_u32_assert(0);
+ int skip = get_u32();
+ pos += skip;
}
int
match_byte_assert(1);
match_byte_assert(0);
- /* This might be a version number of some kind, because it seems to
- only appear in an SPV file that also required its own weird
+ /* This might be a version number of some kind, because value 1 seems
+ to only appear in an SPV file that also required its own weird
special cases in dump_value_31(). */
+ version = get_u32();
+ pos -= 4;
if (!match_u32(1))
match_u32_assert(3);