[PATCH] New "strict_string_checks" run-time flag

[Maria Guseva]

In http://reviews.llvm.org/D7123#114229, @kcc wrote:

> Specifics please. We may want to fix the tests


Here is a list of failed tests for one configuration:

  AddressSanitizer-Unit :: Asan-i386-inline-Test/AddressSanitizer.AtoiAndFriendsOOBTest
  AddressSanitizer-Unit :: Asan-i386-inline-Test/AddressSanitizer.StrCaseCmpOOBTest
  AddressSanitizer-Unit :: Asan-i386-inline-Test/AddressSanitizer.StrCatOOBTest
  AddressSanitizer-Unit :: Asan-i386-inline-Test/AddressSanitizer.StrCmpOOBTest
  AddressSanitizer-Unit :: Asan-i386-inline-Test/AddressSanitizer.StrNCatOOBTest
  AddressSanitizer-Unit :: Asan-i386-inline-Test/AddressSanitizer.StrtolOOBTest
  AddressSanitizer-Unit :: Asan-i386-inline-Test/AddressSanitizer.StrtollOOBTest

These are test cases when string args were not correct but a result still may be calculated correctly, see:

AtoiAndFriendsOOBTest unexpected READ error: arg's last symbol is not digit, 
StrCaseCmpOOBTest unexpected READ error: s1 and s2 last bytes are not null but differs
StrCatOOBTest unexpected READ error: dest not adressable/not null-terminated, WRITE error is expected
StrCmpOOBTest unexpected READ error: s1 and s2 last bytes are not null but differs
StrNCatOOBTest unexpected READ error: dest not adressable/not null-terminated, WRITE error is expected
StrtolOOBTest unexpected READ error: arg's last symbol is not valid digit
StrtollOOBTest unexpected READ error: arg's last symbol is not valid digit

And I see only first failures but there are several test cases with such special arguments for each function. I believe they are intentional and based on expected behaviour when strings are not checked strictly.


http://reviews.llvm.org/D7123

EMAIL PREFERENCES
  http://reviews.llvm.org/settings/panel/emailpreferences/