| Summary: | Wrong code in gcc-3.3.5 and gcc-3.4.2 (4.0 not checked) | ||
|---|---|---|---|
| Product: | gcc | Reporter: | Krzysztof Strasburger <strasbur> |
| Component: | c | Assignee: | Not yet assigned to anyone <unassigned> |
| Status: | RESOLVED DUPLICATE | ||
| Severity: | critical | CC: | gcc-bugs |
| Priority: | P2 | ||
| Version: | 3.3.5 | ||
| Target Milestone: | --- | ||
| Host: | i386-pc-linux | Target: | i386-pc-linux |
| Build: | i386-pc-linux | Known to work: | |
| Known to fail: | Last reconfirmed: | ||
Wrong code indeed, but on your side: your C code violates the aliasing rules of the ISO C language. Either fix it or pass -fno-strict-aliasing to the compiler. See the entry of -fstrict-aliasing in the manual for more detailed info. Subject: Re: Wrong code in gcc-3.3.5 and gcc-3.4.2 (4.0 not checked) Thanks a lot! This error occured originally in the f2c-translated FORTRAN code. It works flawlessly with -fno-strict-aliasing. Krzysztof PS who says, there is no support with free software ??? Reopning to ... |
This code gives "0 0" at first call to printf (wrong) and "4 4" at second call (good). Works correctly with -O1, errors occur with -Os or -O2 or above. void read8(int *x) { x[0] = 4; x[1] = 4;} int main() { static double equiv[1]; double r; read8(&r); *(equiv) = r; printf("%d %d\n",((int *)equiv)[0],((int *)equiv)[1]); printf("%d %d\n",((int *)equiv)[0],((int *)equiv)[1]); return 0; }