Fetch API: Stop lowercasing header names.

third_party/WebKit/Source/modules/cachestorage/Cache.cpp

Index: third_party/WebKit/Source/modules/cachestorage/Cache.cpp
diff --git a/third_party/WebKit/Source/modules/cachestorage/Cache.cpp b/third_party/WebKit/Source/modules/cachestorage/Cache.cpp
index 2c89c8a3f81507533e3dfee6e60354d80f4347f7..6250d8c40d14121b0603b2181d1e9447e8beb973 100644
--- a/third_party/WebKit/Source/modules/cachestorage/Cache.cpp
+++ b/third_party/WebKit/Source/modules/cachestorage/Cache.cpp
@@ -205,9 +205,8 @@ void RecordResponseTypeForAdd(const Member<Response>& response) {
 
 bool varyHeaderContainsAsterisk(const Response* response) {
   const FetchHeaderList* headers = response->headers()->headerList();
-  for (size_t i = 0; i < headers->size(); ++i) {
-    const FetchHeaderList::Header& header = headers->entry(i);
-    if (header.first == "vary") {
+  for (const auto& header : headers->list()) {
+    if (equalIgnoringCase(header.first, "vary")) {

[haraken, 2017/03/31 00:06:28]

Should we probably use equalIgnoringASCIICase instead?

[Raphael Kubo da Costa (rakuco), 2017/03/31 10:22:21]

Sure. I've done that in patch v3.

This actually reminds me if the new implementation of
FetchHeaderList::sortAndCombine() looks OK: there's no WTFString::lowerASCII(),
so I end up calling WTFString::lower() even though the spec says the header
names should only be "byte-lowercased".

       Vector<String> fields;
       header.second.split(',', fields);
       for (size_t j = 0; j < fields.size(); ++j) {