Add support for CSS unicod encoding to the text codec.

chrome/browser/download/save_page_browsertest.cc

Index: chrome/browser/download/save_page_browsertest.cc
diff --git a/chrome/browser/download/save_page_browsertest.cc b/chrome/browser/download/save_page_browsertest.cc
index 0c497e9d93967e1cbcad7abedddfb9b4ce26ab69..7bff37c69e358f5a28d532288c1ff3cba62e3449 100644
--- a/chrome/browser/download/save_page_browsertest.cc
+++ b/chrome/browser/download/save_page_browsertest.cc
@@ -735,6 +735,11 @@ IN_PROC_BROWSER_TEST_F(SavePageAsMHTMLBrowserTest, SavePageAsMHTML) {
   int64_t actual_file_size = -1;
   EXPECT_TRUE(base::GetFileSize(full_file_name, &actual_file_size));
   EXPECT_LE(kFileSizeMin, actual_file_size);
+
+  std::string contents;
+  EXPECT_TRUE(base::ReadFileToString(full_file_name, &contents));
+  // Test for a CSS encoded character.  This used to use HTML encoding.
+  EXPECT_THAT(contents, HasSubstr("content: \"\\e003\""));
 }

[Łukasz Anforowicz, 2016/04/27 04:28:12]

I am happy with the test above, but could you please try to see how much trouble
it would be to also test via SavePageOriginalVsSavedComparisonTest?  This would
1) cover both MHTML and CompleteHTML modes + 2) do more of an end-to-end
verification (via find-in-page in original-vs-saved page) and 3) be a little bit
resilient to implementation details (i.e. verification would pass regardless of
whether ' vs " is used and/or e vs E).  To do this you could:
1. Add some content (span?) to b.html that is affected by the .icon:before that
you've added to 1.css
2. Add a search string to one of the test cases
   - I think we can just pick an arbitrary existing testcase that uses b.html -
i.e. the new search string can be added to |arr| in NestedFrames
   - For how to encode the string, you can see Encoding testcase (basically -
have to encode unicode characters as utf8, using \x escapes as we've found out
in https://codereview.chromium.org/1407663004#msg56).

[bburns, 2016/04/27 20:24:21]

I tried adding this test, but it passes both with and without my change.  I
think this is due to the way that the serialization works for the complete test.

I appreciate the desire to have more end to end coverage, but at this point, I
don't think the effort is worthwhile.

[Łukasz Anforowicz, 2016/04/27 20:25:47]

Acknowledged.

 
 IN_PROC_BROWSER_TEST_F(SavePageBrowserTest, SavePageBrowserTest_NonMHTML) {