[PATCH] D25199: [ubsan] Sanitize deleted pointers

[Filipe Cabecinhas via cfe-commits]

filcab added a comment.

In https://reviews.llvm.org/D25199#560061, @vsk wrote:

> My question was about whether it's possible to resume normal program execution after printing the stack trace from the segv handler. I had assumed this is not possible, and (mistakenly) thought that you were suggesting this approach.


I guess we can eventually add a warning if you have this check + trap-function. If there's really a need for it.



> UndefinedBehaviorSanitizer.rst:122
> +  -  ``-fsanitize=value-after-delete``: Set the value of the pointer
> +     passed in a delete expression to 0xDEADBEEF.
>    -  ``-fsanitize=vla-bound``: A variable-length array whose bound

Why just `delete` and not `free()`?

> CGExprScalar.cpp:416
> +      if (arg->IgnoreImplicit()->isLValue() &&
> +          !arg->HasSideEffects(CGF.getContext())) {
> +        LValue LHS = EmitLValue(arg);

Missing a test for this condition.

> sanitize-value-after-delete.cpp:2
> +// Test -fsanitize-value-after-delete
> +// RUN: %clang_cc1 -O3 -fsanitize=value-after-delete -disable-llvm-optzns -std=c++11 -triple=x86_64-pc-linux -emit-llvm -o - %s | FileCheck %s
> +

Please keep the test simple. You don't even need C++11 (in addition to the flags vsk mentioned).

> sanitize-value-after-delete.cpp:22
> +// CHECK: store {{.*}} inttoptr (i64 -2401053088876216593 {{.*}} %p2
> +// CHECK-NOT: store {{.*}} inttoptr (i64 -2401053088876216593 {{.*}} %p2
> +// CHECK-LABEL: DO_NOT_MODIFY

Why?

https://reviews.llvm.org/D25199