kikairoya wrote:

> As you say, there are different outputs between -O0 and -O1, but is that 
> because of a behaviour related to this feature, or is it just the normal 
> behaviour of the compiler? If it's normal, we should remove the -O1 coverage. 
> If there's something important (the presence / absence of the external 
> symbol?) then we should narrow down what pass causes that behaviour and test 
> that, rather than a full invocation of clang.

I've noticed a difference in behavior based on the optimization level, which 
changes the linkage of the vtables.

https://github.com/llvm/llvm-project/blob/c0562ff27f4de086e4687c5ad5aaea6102faf0b8/clang/lib/CodeGen/CGVTables.cpp#L1120-L1121

We definitely need both of `-O0` and `-O1`, but the latter can be `-O1 
-disable-llvm-passes`. Does that make sense?

https://github.com/llvm/llvm-project/pull/151818
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

Reply via email to