According to the comment at line 239:
if (LoadedSuccessfully.count(*victim) == 0) {
// Before removing the module file, check if it was validated in an
// ancestor thread, if yes, throw a hard error instead of causing
// use-after-free in the ancestor thread.
bool IsSystem;
if (FileMgr.getPCMCache()->isValidatedByAncestor((*victim)->FileName,
IsSystem))
ValidationConflicts.push_back((*victim)->FileName);
else
FileMgr.invalidateCache((*victim)->File);
FileMgr.getPCMCache()->removeFromConsistentBuffer((*victim)->FileName);
}
that looks like it would cause a use-after-free in the parent thread. What am
I missing?
> On 2017-Jan-24, at 17:56, Bruno Cardoso Lopes via Phabricator
> <revi...@reviews.llvm.org> wrote:
>
> bruno added inline comments.
>
>
> ================
> Comment at: lib/Serialization/ASTReader.cpp:3692
> + ValidationConflicts);
> + for (auto N : ValidationConflicts)
> + Diag(diag::err_module_ancestor_conflict) << N;
> ----------------
> This should honor `ARR_OutOfDate`, so that the module can be rebuilt in case
> a user module mismatch via a different diagnostics
> (lib/Serialization/ASTReader.cpp:4076) and returns OutOfDate instead of
> Success:
>
> if ((ClientLoadCapabilities & ARR_OutOfDate) == 0)
> for (auto N : ValidationConflicts)
> Diag(diag::err_module_ancestor_conflict) << N;
>
>
> https://reviews.llvm.org/D28299
>
>
>
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
