Fix SafeSPrintfTest.Truncation in base_unittests

base/strings/safe_sprintf_unittest.cc

 // Copyright 2013 The Chromium Authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
 #include "base/strings/safe_sprintf.h"
 
+#include <inttypes.h>

[Mark Mentovai, 2014/06/15 12:27:26]

Does MSVC even have this header?

[Yang Gu, 2014/06/16 10:39:09]

Not quite sure about MSVC. I searched the web and seems it might not contain
this header. 
I see some other places in this file use "%llX" (switch back to original format)
as format and cast pointer to "(unsigned long long)(uintptr_t)". Looks ok
according to my test on 32-bit and 64-bit.

 #include <stdio.h>
 #include <string.h>
 
 #include <limits>
 
 #include "base/logging.h"
 #include "base/memory/scoped_ptr.h"
 #include "testing/gtest/include/gtest/gtest.h"
 
 // Death tests on Android are currently very flaky. No need to add more flaky
@@ skipping 392 matching lines @@
       break;
     }
   }
 
   // All trailing bytes are unchanged.
   for (size_t i = len+1; i < sz+2; ++i)
     EXPECT_EQ('X', tmp[i]);
 
   // The text that was generated by SafeSPrintf() should always match the
   // equivalent text generated by sprintf(). Please note that the format
   // string for sprintf() is nor complicated, as it does not have the

[Mark Mentovai, 2014/06/15 12:19:22]

This should say “not”, not “nor”.

   // benefit of getting type information from the C++ compiler.
   //
   // N.B.: It would be so much cleaner to use snprintf(). But unfortunately,
   //       Visual Studio doesn't support this function, and the work-arounds
   //       are all really awkward.
   char ref[256];
   CHECK_LE(sz, sizeof(ref));
-  sprintf(ref, "A long string: %%d 00DEADBEEF %lld 0x%llX <NULL>",
+  sprintf(ref, "A long string: %%d 00DEADBEEF %lld 0x%" PRIXPTR " <NULL>",

[Mark Mentovai, 2014/06/15 12:19:21]

Why can’t you use %p here?

[Yang Gu, 2014/06/16 10:39:09]

The actual format is in SafeSNPrintf(), where %p is formatted as upper case. But
the expected result is in sprintf, where %p is in lower case.

           static_cast<long long>(std::numeric_limits<intptr_t>::min()),

[Mark Mentovai, 2014/06/15 12:19:22]

Wouldn’t PRIdPTR (and no cast) be better here and in the format strings used
above on lines 379 and 381?

[Mark Mentovai, 2014/06/15 12:23:02]

Mark Mentovai wrote:
Never mind, I see that %d is intentionally used above, in the comment on line
370.

[Mark Mentovai, 2014/06/15 12:27:26]

Mark Mentovai wrote:
The “never mind” applies to lines 379 and 381. I still think PRIdPTR would be
better down here.

[Yang Gu, 2014/06/16 10:39:09]

PRIdPTR is in decimal format, but we actually need hexadecimal format (%p). 

-          (long long)PrintLongString);
+          (uintptr_t)PrintLongString);
   ref[sz-1] = '\000';
 
 #if defined(NDEBUG)
   const size_t kSSizeMax = std::numeric_limits<ssize_t>::max();
 #else
   const size_t kSSizeMax = internal::GetSafeSPrintfSSizeMaxForTest();
 #endif
 
   // Compare the output from SafeSPrintf() to the one from sprintf().
   EXPECT_EQ(std::string(ref).substr(0, kSSizeMax-1), std::string(tmp.get()));
@@ skipping 302 matching lines @@
   void *ptr = str;
   char buf[40];
   EXPECT_EQ(10, SafeSPrintf(buf, "%p", str));
   EXPECT_EQ("0x80000000", std::string(buf));
   EXPECT_EQ(10, SafeSPrintf(buf, "%p", ptr));
   EXPECT_EQ("0x80000000", std::string(buf));
 }
 
 }  // namespace strings
 }  // namespace base