[PATCH] D126534: [analyzer] Deadstore static analysis: Fix false positive on C++17 assignments

[Balázs Benics via Phabricator via cfe-commits]

steakhal added a comment.

I haven't checked the implementation; the Deadstores checker was always a nemesis for me.

Off the top of my head, I think in the analyzer we have the `elide-constructors=true/false` analyzer config option, which quotes:

> Whether elidable C++ copy-constructors and move-constructors should be actually elided during analysis. Both behaviors are allowed by the C++ standard, and the analyzer, like CodeGen, defaults to eliding. Starting with C++17 some elisions become mandatory, and in these cases the option will be ignored.

You also mentioned this copy elision, but I don't see anywhere checks for the standard version in code. I would expect that your change should apply only to C++17 and above.



================
Comment at: clang/test/Analysis/dead-stores.cpp:63-68
+void copy(int x) {
+  TestConstructor tc1 = x;                    // no-warning
+  TestConstructor tc2 = TestConstructor(x);   // no-warning
+  TestConstructor tc3 = (TestConstructor(x)); // no-warning
+  TestConstructor tc4 = (TestConstructor)(x); // no-warning
+}
----------------
So we don't warn for these since the opaque ctor call might introduce some side-effects.
Please note this after the `no-warning` tag.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D126534/new/

https://reviews.llvm.org/D126534