[LLVMbugs] [Bug 21102] New: Warn/static-analysis diagnostic on passing reference to objects that may be invalidated
bugzilla-daemon at llvm.org
bugzilla-daemon at llvm.org
Tue Sep 30 14:41:22 PDT 2014
http://llvm.org/bugs/show_bug.cgi?id=21102
Bug ID: 21102
Summary: Warn/static-analysis diagnostic on passing reference
to objects that may be invalidated
Product: clang
Version: trunk
Hardware: All
OS: All
Status: NEW
Severity: enhancement
Priority: P
Component: Static Analyzer
Assignee: kremenek at apple.com
Reporter: seth.cantrell at gmail.com
CC: llvmbugs at cs.uiuc.edu
Classification: Unclassified
In his presentation 'Back to the Basics! Essentials of Modern C++ Style'[1]
Herb Sutter calls out one case of using smart pointers that can cause object
lifetime errors:
#include <memory>
#include <cassert>
#include <vector>
struct widget { std::vector<int> v {1,2,3}; };
std::shared_ptr<widget> g_p = std::make_shared<widget>();
void g() {
g_p = nullptr;
}
void use(widget &w) {
assert(w.v.size() == 3);
}
void f(widget &w) {
g();
use(w);
}
int main() {
f(*g_p);
}
The above program invokes undefined behavior by accessing an object through an
invalidated reference.
Herb argues that the code `f(*g_p)` should not pass code reviews and that this
problem should be statically detectable (at 25 to 27 minutes into his
presentation [2]). It would be nice to have this automated check, as a compiler
warning if possible, but at least as a static analyzer check.
Furthermore, the same error can happen with non-smart-pointer managed objects.
Here's an example that the analyzer doesn't currently catch but which might be
usefully caught:
#include <cassert>
int *p;
void g() {
*p = 0;
delete p;
p = nullptr;
}
void use(int &i) {
assert(i == 10);
}
void f(int &i) {
g();
use(i);
}
int main() {
p = new int{10}
f(*p);
}
[1]:
https://github.com/CppCon/CppCon2014/tree/master/Presentations/Back%20to%20the%20Basics!%20Essentials%20of%20Modern%20C%2B%2B%20Style
[2]: https://www.youtube.com/watch?v=xnqTKD8uD64
--
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20140930/1aeaa9e1/attachment.html>
More information about the llvm-bugs
mailing list