Bug 870419
| Summary: | gcc 4.7.2.2 (64bit) does not inline compile the following properly, gcc 4.6.3 (16bit) inline compiles properly (fedora 17 for 64bit) fedora 16 (for 32bit) | ||
|---|---|---|---|
| Product: | [Fedora] Fedora | Reporter: | Leslie Satenstein <lsatenstein> |
| Component: | gcc4 | Assignee: | Jakub Jelinek <jakub> |
| Status: | CLOSED NOTABUG | QA Contact: | |
| Severity: | high | Docs Contact: | |
| Priority: | unspecified | ||
| Version: | 4 | ||
| Target Milestone: | --- | ||
| Target Release: | --- | ||
| Hardware: | x86_64 | ||
| OS: | Linux | ||
| Whiteboard: | |||
| Fixed In Version: | Doc Type: | Bug Fix | |
| Doc Text: | Story Points: | --- | |
| Clone Of: | Environment: | ||
| Last Closed: | 2012-11-19 13:59:49 UTC | Type: | Bug |
| Regression: | --- | Mount Type: | --- |
| Documentation: | --- | CRM: | |
| Verified Versions: | Category: | --- | |
| oVirt Team: | --- | RHEL 7.3 requirements from Atomic Host: | |
| Cloudforms Team: | --- | Target Upstream Version: | |
| Embargoed: | |||
Please read http://gcc.gnu.org/bugs.html, this bugreport is completely useless. You haven't provided self-contained preprocessed testcase, so one can only guess you have aliasing violation (if what dest/src point to aren't actually objects of uint32 or compatible type). See if your testcase works with -fno-strict-aliasing, then it is likely user fault. I very much doubt adding a pair of {}s would change anything in that regard, but without a self-contained testcase there is nothing I can try to compile. Furthermore, there is no 4.7.2.2 gcc version. |
Description of problem: Getting bad inline code Version-Release number of selected component (if applicable): gcc 4.7.2.2 (64bit) does not compile the following properly, (SEE BELOW) gcc 4.6.3 (16bit) compiles properly fedora 17 for 64bit 4.7.2.2. fedora 16 4.6.3 for 32bit) Hardware Intel e7300 dual core 64bit for both systems. Memory available (free unused) 2.5+ gig) How reproducible: everytime Steps to Reproduce: 1. 2. 3. Actual results: Expected results: Additional info: The following code fragment indicates a bug that took me 2 days to discover. It was used for the past 2 years. It is compiled and used correctly on 32 bit system but generates incorrect results for 64 bit systems. The fix is shown below the actual code. The code works with the fix. /* * do an xor of an 8 byte block * optimized to two loops vs an 8 char loop */ static inline void memxor(unsigned char *dest, char *src) { register int i = 8 / sizeof(uint32); //uint32 is 4 bytes uint32 *d = (uint32 *) dest; uint32 *s = (uint32 *) src; while (0 < i--) //two loops if unsigned long is 4 bytes *d++ ^= *s++; } The fixed code that works /* * do an xor of an 8 byte block * optimized vs an 8 char loop */ static inline void memxor(unsigned char *dest, char *src) { register int i = 8 / sizeof(uint32); //uint32 is 4 bytes uint32 *d = (uint32 *) dest; uint32 *s = (uint32 *) src; while (0 < i--) //two loops if unsigned long is 4 bytes { //added { and } and code generation is correct. *d++ ^= *s++; } }