Teach PasswordAutofillAgent sometimes match prefixes of usernames

components/autofill/core/common/autofill_util.cc

Index: components/autofill/core/common/autofill_util.cc
diff --git a/components/autofill/core/common/autofill_util.cc b/components/autofill/core/common/autofill_util.cc
index 50e50176d8c336e700bef8af1c7e6bfd865250fd..251522fe9cd0ab5548debb183d62dbef6db8f69b 100644
--- a/components/autofill/core/common/autofill_util.cc
+++ b/components/autofill/core/common/autofill_util.cc
@@ -115,6 +115,23 @@ bool FieldIsSuggestionSubstringStartingOnTokenBoundary(
              base::string16::npos;
 }
 
+bool IsPrefixOfEmailEndingAtSign(const base::string16& full_string,
+                                 const base::string16& prefix) {
+  auto at_sign = std::find(full_string.begin(), full_string.end(), '@');
+  auto e = full_string.end();
+  // weak check if |full_string| is a valid email address.

[vabr (Chromium), 2017/06/19 12:54:15]

nit: weak -> A weak
Or maybe better "An incomplete check".

[melandory, 2017/06/21 12:00:42]

Done.

+  bool is_email =

[vabr (Chromium), 2017/06/19 12:54:15]

Well, actually, hinting at e-mail, but only doing it just a little is confusing.
The original suggestion https://crbug.com/543435#c10 from the security team does
only request that instead of testing that the suggestion is a prefix of field
contents, we test that suggestion + "@" is a prefix of the field contents.

So I suggest replacing the whole function with just that:

base::StartsWith(full_string, prefix + "@", base::CompareCase::SENSITIVE));

with the following reasons:

* It is secure enough: this still respects origins and follows the security team
suggestion; also other incomplete e-mail address checks are not clearly better.
* Code will be significantly simpler.
* We might still tune this based on user feedback (I'm looking especially at the
CompareCase::SENSITIVE), so let's start simple.

[melandory, 2017/06/21 12:00:42]

In this case "ab@cd@g", "ab@cd" will result in match. Is it fine?

[vabr (Chromium), 2017/06/21 12:28:07]

I think it is fine. From usability perspective I don't see any issues, and for
security, this follows the guidance from security team given on the bug.

+      (at_sign != e                            // there is @
+       && std::find(at_sign, e, '.') != e      // there is . and it's after @
+       && std::find(++at_sign, e, '@') == e);  // there is at most 1 @
+  if (is_email &&
+      base::StartsWith(full_string, prefix, base::CompareCase::SENSITIVE)) {
+    return (full_string.size() > prefix.size() &&
+            full_string[prefix.size()] == '@');
+  }
+  return false;
+}
+
 size_t GetTextSelectionStart(const base::string16& suggestion,
                              const base::string16& field_contents,
                              bool case_sensitive) {