1557b76ed
Denys Vlasenko
2018-12-22
Subject
bc: shrink zbc_num_ulong()
Binary
No binary (build failed)
Architecture & Compiler Flags
—
Build Log
(log yok)
Commit Diff GitHub →
miscutils/bc.c+11 -10
| @@ -1442,20 +1442,21 @@ static void bc_num_copy(BcNum *d, BcNum *s) | |
| static BC_STATUS zbc_num_ulong(BcNum *n, unsigned long *result_p) | |
| { | |
| size_t i; | |
| - | unsigned long pow, result; |
| + | unsigned long result; |
| if (n->neg) RETURN_STATUS(bc_error("negative number")); | |
| - | for (result = 0, pow = 1, i = n->rdx; i < n->len; ++i) { |
| - | unsigned long prev = result, powprev = pow; |
| - | |
| - | result += ((unsigned long) n->num[i]) * pow; |
| - | pow *= 10; |
| - | |
| - | if (result < prev || pow < powprev) |
| + | result = 0; |
| + | i = n->len; |
| + | while (i > n->rdx) { |
| + | unsigned long prev = result; |
| + | result = result * 10 + n->num[--i]; |
| + | // Even overflowed N*10 can still satisfy N*10>=N. For example, |
| + | // 0x1ff00000 * 10 is 0x13f600000, |
| + | // or 0x3f600000 truncated to 32 bits. Which is larger. |
| + | // However, (N*10)/8 < N check is always correct. |
| + | if ((result / 8) < prev) |
| RETURN_STATUS(bc_error("overflow")); | |
| - | prev = result; |
| - | powprev = pow; |
| } | |
| *result_p = result; | |