[cfe-commits] [PATCH] Add three new sanitizers depending on ASan.

Alexey Samsonov samsonov at google.com
Tue Nov 27 22:28:53 PST 2012


  Ok. Kostya, what is your opinion?


================
Comment at: lib/Driver/Tools.cpp:1518-1523
@@ -1517,1 +1517,8 @@
+
+  // If -fsanitize contains extra features of ASan, it should also
+  // explicitly contain -fsanitize=address. 
+  if (NeedsAsan && ((Kind & Address) == 0))
+    D.Diag(diag::err_drv_argument_only_allowed_with)
+      << describeSanitizeArg(Args, AsanArg, NeedsAsanRt)
+      << "-fsanitize=address";
 }
----------------
Richard Smith wrote:
> For argument lists like "-fsanitize=use-after-return -fsanitize=address -fno-sanitize=address", we'll say "-fsanitize=address is only allowed with -fsanitize=address". The existing diagnostic has a similar issue for "-fsanitize=address -fsanitize=alignment -fsanitize=vptr -fno-sanitize=vptr", where it says "-fsanitize=vptr not allowed with -fsanitize=address". I think we'd need to teach describeSanitizerArg to re-parse the argument list to handle this properly.
RIght, the logic is untrivial here. Mailed D143 to fix this.

================
Comment at: lib/Driver/SanitizerArgs.h:59-71
@@ -58,1 +58,15 @@
+  
+    // Add args for LLVM backend.
+    if (Kind & InitOrder) {
+      CmdArgs.push_back("-mllvm");
+      CmdArgs.push_back("-asan-initialization-order");
+    }
+    if (Kind & UseAfterReturn) {
+      CmdArgs.push_back("-mllvm");
+      CmdArgs.push_back("-asan-use-after-return");
+    }
+    if (Kind & UseAfterScope) {
+      CmdArgs.push_back("-mllvm");
+      CmdArgs.push_back("-asan-use-lifetime");
+    }
   }
----------------
Richard Smith wrote:
> I would prefer this to be handled by the frontend instead of by the driver (the frontend is responsible for adding all the other IR instrumentation, including adding the ASan passes).
> 
> Have you considered passing these flags to ASan when creating the passes in addAddressSanitizerPass, rather than as command-line options?
Hm, passing arguments to createAddressSanitizerPass() certainly seems a better (though, more intrusive) solution than playing with -mllvm flags. I'll work on that.


http://llvm-reviews.chandlerc.com/D142



More information about the cfe-commits mailing list