wlei added a comment.
In D109386#2990504 <https://reviews.llvm.org/D109386#2990504>, @rnk wrote:
> This seems incorrect: a weak VH won't have the same semantics, it will just
> drop the comdat associativity when the underlying global changes type.
@rnk Thanks. Here I tried to use `WeakTrackingVH` working together with
`replaceAllUsesWith` which is already there (in CodeGenModule.cpp:3904), my
debugging dump showed like:
GlobalCtors = [..,OldPtr];
OldPtr->replaceAllUsesWith(NewPtr);
GlobalCtors = [..,NewPtr];
I think this is what we desire. Do you think this can cause other issues or
anything I missed?
> BTW, can this be tested, or is this only possible when clang is being used as
> a library, perhaps as in swift?
Not swift, it seems this use-after-free needs a large program to repro, I tried
to simplify it but failed to expose it by small program.
================
Comment at: clang/lib/CodeGen/CodeGenModule.cpp:3904
llvm::ConstantExpr::getBitCast(GV, Entry->getType());
Entry->replaceAllUsesWith(NewPtrForOldDecl);
}
----------------
Here is the `replaceAllUsesWith` .
Repository:
rG LLVM Github Monorepo
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D109386/new/
https://reviews.llvm.org/D109386
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
