Merge pull request #182 from Medios-lgtm/parse-number-slice-bounds

base: don't read past the end of the slice in parse_number_{u64,f64}
This commit is contained in:
Nigel Tao authored and GitHub committed 2026-10-20 13:28:38 +11:00
commit 2fa1d20d76
4 files changed
+61 -36

No files matched your search

+12 -18
View File
@@ -268,14 +268,12 @@ wuffs_private_impl__high_prec_dec__parse(wuffs_private_impl__high_prec_dec* h,
uint8_t* q = s.ptr + s.len;
if (options & WUFFS_BASE__PARSE_NUMBER_XXX__ALLOW_UNDERSCORES) {
for (;; p++) {
if (p >= q) {
return wuffs_base__make_status(wuffs_base__error__bad_argument);
} else if (*p != '_') {
break;
}
for (; (p < q) && (*p == '_'); p++) {
}
}
if (p >= q) {
return wuffs_base__make_status(wuffs_base__error__bad_argument);
}
// Parse sign.
do {
@@ -288,14 +286,12 @@ wuffs_private_impl__high_prec_dec__parse(wuffs_private_impl__high_prec_dec* h,
break;
}
if (options & WUFFS_BASE__PARSE_NUMBER_XXX__ALLOW_UNDERSCORES) {
for (;; p++) {
if (p >= q) {
return wuffs_base__make_status(wuffs_base__error__bad_argument);
} else if (*p != '_') {
break;
}
for (; (p < q) && (*p == '_'); p++) {
}
}
if (p >= q) {
return wuffs_base__make_status(wuffs_base__error__bad_argument);
}
} while (0);
// Parse digits, up to (and including) a '.', 'E' or 'e'. Examples for each
@@ -408,14 +404,12 @@ after_sep:
after_exp:
do {
if (options & WUFFS_BASE__PARSE_NUMBER_XXX__ALLOW_UNDERSCORES) {
for (;; p++) {
if (p >= q) {
return wuffs_base__make_status(wuffs_base__error__bad_argument);
} else if (*p != '_') {
break;
}
for (; (p < q) && (*p == '_'); p++) {
}
}
if (p >= q) {
return wuffs_base__make_status(wuffs_base__error__bad_argument);
}
int32_t exp_sign = +1;
if (*p == '+') {
+2
View File
@@ -215,6 +215,7 @@ wuffs_base__parse_number_u64(wuffs_base__slice_u8 s, uint32_t options) {
if (p < q) {
goto hexadecimal;
}
goto fail_bad_argument;
} else if ((*p == 'd') || (*p == 'D')) {
p++;
@@ -225,6 +226,7 @@ wuffs_base__parse_number_u64(wuffs_base__slice_u8 s, uint32_t options) {
if (p < q) {
goto decimal;
}
goto fail_bad_argument;
}
if (options & WUFFS_BASE__PARSE_NUMBER_XXX__ALLOW_MULTIPLE_LEADING_ZEROES) {
+14 -18
View File
@@ -21668,14 +21668,12 @@ wuffs_private_impl__high_prec_dec__parse(wuffs_private_impl__high_prec_dec* h,
uint8_t* q = s.ptr + s.len;
if (options & WUFFS_BASE__PARSE_NUMBER_XXX__ALLOW_UNDERSCORES) {
for (;; p++) {
if (p >= q) {
return wuffs_base__make_status(wuffs_base__error__bad_argument);
} else if (*p != '_') {
break;
}
for (; (p < q) && (*p == '_'); p++) {
}
}
if (p >= q) {
return wuffs_base__make_status(wuffs_base__error__bad_argument);
}
// Parse sign.
do {
@@ -21688,14 +21686,12 @@ wuffs_private_impl__high_prec_dec__parse(wuffs_private_impl__high_prec_dec* h,
break;
}
if (options & WUFFS_BASE__PARSE_NUMBER_XXX__ALLOW_UNDERSCORES) {
for (;; p++) {
if (p >= q) {
return wuffs_base__make_status(wuffs_base__error__bad_argument);
} else if (*p != '_') {
break;
}
for (; (p < q) && (*p == '_'); p++) {
}
}
if (p >= q) {
return wuffs_base__make_status(wuffs_base__error__bad_argument);
}
} while (0);
// Parse digits, up to (and including) a '.', 'E' or 'e'. Examples for each
@@ -21808,14 +21804,12 @@ after_sep:
after_exp:
do {
if (options & WUFFS_BASE__PARSE_NUMBER_XXX__ALLOW_UNDERSCORES) {
for (;; p++) {
if (p >= q) {
return wuffs_base__make_status(wuffs_base__error__bad_argument);
} else if (*p != '_') {
break;
}
for (; (p < q) && (*p == '_'); p++) {
}
}
if (p >= q) {
return wuffs_base__make_status(wuffs_base__error__bad_argument);
}
int32_t exp_sign = +1;
if (*p == '+') {
@@ -23585,6 +23579,7 @@ wuffs_base__parse_number_u64(wuffs_base__slice_u8 s, uint32_t options) {
if (p < q) {
goto hexadecimal;
}
goto fail_bad_argument;
} else if ((*p == 'd') || (*p == 'D')) {
p++;
@@ -23595,6 +23590,7 @@ wuffs_base__parse_number_u64(wuffs_base__slice_u8 s, uint32_t options) {
if (p < q) {
goto decimal;
}
goto fail_bad_argument;
}
if (options & WUFFS_BASE__PARSE_NUMBER_XXX__ALLOW_MULTIPLE_LEADING_ZEROES) {
+33
View File
@@ -1082,6 +1082,22 @@ test_wuffs_strconv_parse_number_f64_regular() {
}
}
// Input that ends right after a sign or an exponent marker must be rejected
// without reading past the end of the slice. Use DEFAULT_OPTIONS, as
// ALLOW_UNDERSCORES takes a different code path.
{
const char* strs[] = {"", "+", "-", "1e", "1E", "12.5e"};
for (size_t i = 0; i < WUFFS_TESTLIB_ARRAY_SIZE(strs); i++) {
wuffs_base__result_f64 r = wuffs_base__parse_number_f64(
wuffs_base__make_slice_u8((uint8_t*)(void*)strs[i], strlen(strs[i])),
WUFFS_BASE__PARSE_NUMBER_XXX__DEFAULT_OPTIONS);
if (r.status.repr != wuffs_base__error__bad_argument) {
RETURN_FAIL("\"%s\": have \"%s\", want \"%s\"", strs[i],
r.status.repr, wuffs_base__error__bad_argument);
}
}
}
return NULL;
}
@@ -1245,6 +1261,23 @@ test_wuffs_strconv_parse_number_u64() {
}
}
// A bare "0x" or "0d" prefix is invalid, even with
// ALLOW_MULTIPLE_LEADING_ZEROES. Pass an explicit length (not strlen) so
// that the byte just past the end of the slice is a digit, which must not be
// read.
{
const char* strs[] = {"0x7", "0X7", "0d7", "0D7"};
for (size_t i = 0; i < WUFFS_TESTLIB_ARRAY_SIZE(strs); i++) {
wuffs_base__result_u64 r = wuffs_base__parse_number_u64(
wuffs_base__make_slice_u8((uint8_t*)(void*)strs[i], 2),
WUFFS_BASE__PARSE_NUMBER_XXX__ALLOW_MULTIPLE_LEADING_ZEROES);
if (r.status.repr != wuffs_base__error__bad_argument) {
RETURN_FAIL("\"%.2s\": have \"%s\", want \"%s\"", strs[i],
r.status.repr, wuffs_base__error__bad_argument);
}
}
}
// Test WUFFS_BASE__PARSE_NUMBER_XXX__ALLOW_UNDERSCORES.
{
for (int o = 0; o < 2; o++) {