hubert-reinterpretcast wrote:
@AaronBallman, considering the concern of potential performance regressions
("unnecessary" initialization of automatic-storage duration objects), should it
be the case that there is no option control?
It was also the case that the prior Clang behaviour (with
`-ftrivial-auto-var-init=pattern` [1]: https://godbolt.org/z/x994WMT6h) had
some utility for identifying potential portability issues with a program.
Option control would make me much happier about this change.
[1]:
A string of `0xaa` is observed using `-ftrivial-auto-var-init=pattern` using
Clang 19.1:
```c
typedef union { int x; long long y; } U;
int printf(const char *, ...);
int main(void) {
U u = { 0 };
typedef unsigned char *ucp;
for (ucp p = (ucp)&u; p < (ucp)&1[&u]; ++p) {
printf("%02x", (unsigned)*p);
}
printf("\n");
}
```
https://github.com/llvm/llvm-project/pull/97121
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
