[PATCH] Sanitize printf functions

[Yury Gribov]

================
Comment at: lib/sanitizer_common/sanitizer_common_interceptors.inc:697
@@ +696,3 @@
+      va_copy(ar, ap);                                                         \
+      int size = REAL(vname2)(str, 0, format, ar);                             \
+      if (size >= 0) {                                                         \
----------------
Evgeniy Stepanov wrote:
> Unless I'm missing something, you could use REAL(strlen)(str) here instead of calling *printf twice.
Hm but str isn't necessarily a string - it's just plain array of chars. I don't think we can call strlen on it.

================
Comment at: lib/sanitizer_common/sanitizer_common_interceptors_scanf.inc:161
@@ +160,3 @@
+  if (convSpecifier == 'p') {
+    if (lengthModifier[1] != 0)
+      return FSS_INVALID;
----------------
Evgeniy Stepanov wrote:
> Should it be lengthModifier[0]?
> Looks like a bug in the original code.
> 
1) Bug or some weird GNU scanf (mis)feature. Who is the author of this piece?
2) How do you guys normally treat situations when both change code and move it to a different place for clarity? Do you make two separate commits to make reviewers life easier?


http://llvm-reviews.chandlerc.com/D2480