aaron.ballman added inline comments.
================ Comment at: clang/lib/Analysis/CFG.cpp:4156 // CFG generation for unevaluated operands. - if (S && !S->isTypeDependent() && S->isPotentiallyEvaluated()) + if (!S->isTypeDependent() && S->isPotentiallyEvaluated()) return VisitChildren(S); ---------------- schittir wrote: > schittir wrote: > > erichkeane wrote: > > > schittir wrote: > > > > erichkeane wrote: > > > > > schittir wrote: > > > > > > erichkeane wrote: > > > > > > > I get that we're counting on the dereference on 4145 to have made > > > > > > > this check irrelevant, but are we sure that we KNOW that "S" is > > > > > > > non-null here otherwise? That is, is the bug actually 4145 doing > > > > > > > 'alwaysAdd' without checking vs the unnecessary check here? > > > > > > VisitCXXTypeidExpr is used only in one place - here > > > > > > https://github.com/llvm/llvm-project/blob/a89c9b35be08b665cc1a01d840bc20349ba1308f/clang/lib/Analysis/CFG.cpp#L2288 > > > > > > where S is not null. Null check for S already happens at the > > > > > > beginning of the method where VisitCXXTypeidExpr is called. > > > > > SG, an assert is perhaps not a bad idea, but LGTM. I don't know very > > > > > much about the CFG stuff, so aaron might wish to take a final look. > > > > What is SG? > > > "Sounds Good" > > Haha! > > Thanks for the review. > > SG, an assert is perhaps not a bad idea, but LGTM. I don't know very much > > about the CFG stuff, so aaron might wish to take a final look. > > Hi @aaron.ballman, could you please comment on this? Thank you! I don't think an assert is necessary; I think it's a predicate to pass non-null nodes to the Visit functions in general, and I don't think we get a whole lot out of asserting in each Visit method that its given node is nonnull. ================ Comment at: clang/lib/CodeGen/CGObjCGNU.cpp:58 /// used as arguments will necessarily be available at construction time. - LazyRuntimeFunction() - : CGM(nullptr), FunctionName(nullptr), Function(nullptr) {} + LazyRuntimeFunction() {} ---------------- schittir wrote: > erichkeane wrote: > > schittir wrote: > > > erichkeane wrote: > > > > > > > The comment above needs to be updated? > > I don't think so? It still leaves the class uninitialized. > Right! Still valid. I think the comment is stale or imprecise -- the class certainly is initialized (it was being initialized previously as well, just not fully initialized). I think it may have meant "not initialized to useful values", so we could either repair it that way or remove the comment entirely. CHANGES SINCE LAST ACTION https://reviews.llvm.org/D153589/new/ https://reviews.llvm.org/D153589 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits