base/test/launcher: don't trim leading whitespace from test output

base/strings/string_split.cc

Index: base/strings/string_split.cc
diff --git a/base/strings/string_split.cc b/base/strings/string_split.cc
index 692f3008e7827cf9b57a25a58d8e74509f92b373..88a623664fcc1c256de60c110c3186ecc0034885 100644
--- a/base/strings/string_split.cc
+++ b/base/strings/string_split.cc
@@ -6,7 +6,6 @@
 
 #include "base/logging.h"
 #include "base/strings/string_util.h"
-#include "base/strings/utf_string_conversions.h"
 #include "base/third_party/icu/icu_utf.h"
 
 namespace base {
@@ -193,7 +192,6 @@ void SplitStringDontTrim(const string16& str,
 void SplitStringDontTrim(const std::string& str,
                          char c,
                          std::vector<std::string>* r) {
-  DCHECK(IsStringUTF8(str));

[Nico, 2015/04/16 17:05:24]

This change looks both unrelated and incorrect. (`c` is a single byte, so
this'll only work for utf-8)

[ncarter (slow), 2015/04/16 18:08:00]

"only work for UTF-8" as opposed to what?

Did you see my message where I explained that it's both related and correct?
Here's what I wrote:

===

I've removed the IsUTF8 check here after determining it to be unnecessary.
Whether it's extended ascii or UTF-8, we can strip whitespace and split along
newlines all the same. SplitString and SplitStringDontTrim share an
implementation, so what's valid input for the former should be valid for the
latter.

At least test on the linux asan bot generates stdio which is not valid UTF-8,
though I don't know why this is.

Archaeological research shows this DCHECK to have been introduced during a
file-move of this function, as part of https://codereview.chromium.org/3366011 .

===

[ncarter (slow), 2015/04/16 18:08:50]

And oh, to understand the relationship: see the failed tryjob on patchset 1.

 #if CHAR_MIN < 0
   DCHECK_GE(c, 0);
 #endif