Many hyperlinks are disabled.
Use anonymous login
to enable hyperlinks.
Overview
| Comment: | Off-by-one error in the CAST to NUMERIC logic. |
|---|---|
| Downloads: | Tarball | ZIP archive |
| Timelines: | family | ancestors | descendants | both | trunk |
| Files: | files | file ages | folders |
| SHA3-256: |
25b89630206fdbb41c756e98f99baa71 |
| User & Date: | drh 2019-06-11 16:06:05.646 |
Context
|
2019-06-11
| ||
| 16:07 | New test case in fuzzdata8.db. (check-in: e96641f0be user: drh tags: trunk) | |
| 16:06 | Off-by-one error in the CAST to NUMERIC logic. (check-in: 25b8963020 user: drh tags: trunk) | |
| 12:03 | Have the ALTER TABLE code handle the case where an entire expression that includes a sub-select is excluded from a view, trigger or index. (check-in: f2c8179f3c user: dan tags: trunk) | |
Changes
Changes to src/vdbemem.c.
| ︙ | ︙ | |||
696 697 698 699 700 701 702 |
** For some versions of GCC on 32-bit machines, if you do the more obvious
** comparison of "r1==(double)i" you sometimes get an answer of false even
** though the r1 and (double)i values are bit-for-bit the same.
*/
int sqlite3RealSameAsInt(double r1, sqlite3_int64 i){
double r2 = (double)i;
return memcmp(&r1, &r2, sizeof(r1))==0
| | | 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 |
** For some versions of GCC on 32-bit machines, if you do the more obvious
** comparison of "r1==(double)i" you sometimes get an answer of false even
** though the r1 and (double)i values are bit-for-bit the same.
*/
int sqlite3RealSameAsInt(double r1, sqlite3_int64 i){
double r2 = (double)i;
return memcmp(&r1, &r2, sizeof(r1))==0
&& i >= -2251799813685248 && i < 2251799813685248;
}
/*
** Convert pMem so that it has type MEM_Real or MEM_Int.
** Invalidate any prior representations.
**
** Every effort is made to force the conversion, even if the input
|
| ︙ | ︙ |