Add some unittests for DoubleToString.

base/string_number_conversions_unittest.cc

Index: base/string_number_conversions_unittest.cc
diff --git a/base/string_number_conversions_unittest.cc b/base/string_number_conversions_unittest.cc
index 438db077a953225adac1d4bf9c8881aca750cb5b..e243663e844e76d5deb2d248fcb7b3ae7c713777 100644
--- a/base/string_number_conversions_unittest.cc
+++ b/base/string_number_conversions_unittest.cc
@@ -344,6 +344,25 @@ TEST(StringNumberConversionsTest, StringToDouble) {
   EXPECT_DOUBLE_EQ(3.14, output);
 }
 
+TEST(StringNumberConversionsTest, DoubleToString) {
+  static const struct {
+    double input;
+    std::string expected;

[eroman, 2012/05/14 22:52:39]

I would suggest making this a |const char*| instead. But I see that other
functions in this file use std::string so I guess that is fine.

[Mark Mentovai, 2012/05/14 23:20:32]

eroman wrote:
This should be const char* to avoid an exit-time destructor.

+  } cases[] = {
+    {0.0, "0"},
+    {1.25, "1.25"},
+    {1.33518e+012, "1.33518e+12"},

[eroman, 2012/05/14 22:52:39]

Note that these are NOT the exact same values as I listed in that bug.

1.33518e+12 = 00 00 b0 df ee 6d 73 42
1.33489e+12 = 00 00 e8 4e da 6c 73 42

Whereas in bug 123157 the first two values were:

00 00 00 00 ee 6d 73 42 (1.33518e+12)
00 00 00 a0 da 6c 73 42 (1.33489e+12)

Notice that bytes 2 and 3 are different.

I suggest running at least a local test using the exact same bytes in the
underlying floating point to see if that is tickling anything.

For the purposes of this CL and adding tests, the way you have it written is
fine though.

+    {1.33489e+012, "1.33489e+12"},
+    {1.33505e+012, "1.33505e+12"},
+    {1.33545e+009, "1335450000"},
+    {1.33503e+009, "1335030000"},
+  };
+
+  for (size_t i = 0; i < ARRAYSIZE_UNSAFE(cases); ++i) {
+    EXPECT_EQ(cases[i].expected, DoubleToString(cases[i].input));
+  }
+}
+
 TEST(StringNumberConversionsTest, HexEncode) {
   std::string hex(HexEncode(NULL, 0));
   EXPECT_EQ(hex.length(), 0U);