[PATCH] D102835: [analyzer] Correctly propagate ConstructionContextLayer thru ParenExpr

[Kristóf Umann via Phabricator via cfe-commits]

Szelethus added a comment.

This looks great!



================
Comment at: clang/test/Analysis/NewDelete-checker-test.cpp:41-52
+// RUN: %clang_analyze_cc1 -std=c++17 -fblocks %s \
+// RUN:   -verify=expected,newdelete \
+// RUN:   -analyzer-checker=core \
+// RUN:   -analyzer-checker=cplusplus.NewDelete \
+// RUN:   -analyzer-config c++-allocator-inlining=true
+//
+// RUN: %clang_analyze_cc1 -std=c++17 -fblocks -verify %s \
----------------
This mail adds some insight into the usage of `c++-allocator-inlining`:

https://lists.llvm.org/pipermail/cfe-dev/2020-February/064754.html

About the code in `MallocChecker` that handles this flag (`NewDelete` is a part of this as well):

> Yes, we should remove the old code for c++-allocator-inlining=false. The worry we've had back then as that in the new mode we've disabled aggressive behavior of MallocChecker in which it reacted to some overloaded operator new invocations but i think this was the right thing to do and also nobody complained; also nothing prevents us from bringing back the old behavior in a much less confusing way.

As a response, D75432 removed all code that handed `c++-allocator-inlinging=false`. So these RUN lines can be removed I think.


================
Comment at: clang/test/Analysis/NewDelete-checker-test.cpp:54-57
+// RUN: %clang_analyze_cc1 -std=c++17 -fblocks -verify %s \
+// RUN:   -verify=expected,leak \
+// RUN:   -analyzer-checker=core \
+// RUN:   -analyzer-checker=cplusplus.NewDeleteLeaks
----------------
This last RUN seems to be the same as the first.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D102835/new/

https://reviews.llvm.org/D102835