NS_IMETHODIMP
nsFeedSniffer::GetMIMETypeFromContent(nsIRequest* request,
const uint8_t* data,
uint32_t length,
nsACString& sniffedType)
{
nsCOMPtr<nsIHttpChannel> channel(do_QueryInterface(request));
if (!channel)
return NS_ERROR_NO_INTERFACE;
// Check that this is a GET request, since you can't subscribe to a POST...
nsCAutoString method;
channel->GetRequestMethod(method);
if (!method.Equals("GET")) {
sniffedType.Truncate();
return NS_OK;
}
// We need to find out if this is a load of a view-source document. In this
// case we do not want to override the content type, since the source display
// does not need to be converted from feed format to XUL. More importantly,
// we don't want to change the content type from something
// nsContentDLF::CreateInstance knows about (e.g. application/xml, text/html
// etc) to something that only the application fe knows about (maybe.feed)
// thus deactivating syntax highlighting.
nsCOMPtr<nsIURI> originalURI;
channel->GetOriginalURI(getter_AddRefs(originalURI));
nsCAutoString scheme;
originalURI->GetScheme(scheme);
if (scheme.EqualsLiteral("view-source")) {
sniffedType.Truncate();
return NS_OK;
}
// Check the Content-Type to see if it is set correctly. If it is set to
// something specific that we think is a reliable indication of a feed, don't
// bother sniffing since we assume the site maintainer knows what they're
// doing.
nsCAutoString contentType;
channel->GetContentType(contentType);
bool noSniff = contentType.EqualsLiteral(TYPE_RSS) ||
contentType.EqualsLiteral(TYPE_ATOM);
// Check to see if this was a feed request from the location bar or from
// the feed: protocol. This is also a reliable indication.
// The value of the header doesn't matter.
if (!noSniff) {
nsCAutoString sniffHeader;
nsresult foundHeader =
channel->GetRequestHeader(NS_LITERAL_CSTRING("X-Moz-Is-Feed"),
sniffHeader);
noSniff = NS_SUCCEEDED(foundHeader);
}
if (noSniff) {
// check for an attachment after we have a likely feed.
if(HasAttachmentDisposition(channel)) {
sniffedType.Truncate();
return NS_OK;
}
// set the feed header as a response header, since we have good metadata
// telling us that the feed is supposed to be RSS or Atom
channel->SetResponseHeader(NS_LITERAL_CSTRING("X-Moz-Is-Feed"),
NS_LITERAL_CSTRING("1"), false);
sniffedType.AssignLiteral(TYPE_MAYBE_FEED);
return NS_OK;
}
// Don't sniff arbitrary types. Limit sniffing to situations that
// we think can reasonably arise.
if (!contentType.EqualsLiteral(TEXT_HTML) &&
!contentType.EqualsLiteral(APPLICATION_OCTET_STREAM) &&
// Same criterion as XMLHttpRequest. Should we be checking for "+xml"
// and check for text/xml and application/xml by hand instead?
contentType.Find("xml") == -1) {
sniffedType.Truncate();
return NS_OK;
}
// Now we need to potentially decompress data served with
// Content-Encoding: gzip
nsresult rv = ConvertEncodedData(request, data, length);
if (NS_FAILED(rv))
return rv;
// We cap the number of bytes to scan at MAX_BYTES to prevent picking up
// false positives by accidentally reading document content, e.g. a "how to
// make a feed" page.
const char* testData;
if (mDecodedData.IsEmpty()) {
testData = (const char*)data;
length = NS_MIN(length, MAX_BYTES);
} else {
testData = mDecodedData.get();
length = NS_MIN(mDecodedData.Length(), MAX_BYTES);
}
// The strategy here is based on that described in:
// http://blogs.msdn.com/rssteam/articles/PublishersGuide.aspx
// for interoperarbility purposes.
// Thus begins the actual sniffing.
nsDependentCSubstring dataString((const char*)testData, length);
bool isFeed = false;
// RSS 0.91/0.92/2.0
isFeed = ContainsTopLevelSubstring(dataString, "<rss");
// Atom 1.0
if (!isFeed)
isFeed = ContainsTopLevelSubstring(dataString, "<feed");
// RSS 1.0
if (!isFeed) {
isFeed = ContainsTopLevelSubstring(dataString, "<rdf:RDF") &&
dataString.Find(NS_RDF) != -1 &&
dataString.Find(NS_RSS) != -1;
}
// If we sniffed a feed, coerce our internal type
if (isFeed && !HasAttachmentDisposition(channel))
sniffedType.AssignLiteral(TYPE_MAYBE_FEED);
else
sniffedType.Truncate();
return NS_OK;
}
void nsUnknownDecoder::DetermineContentType(nsIRequest* aRequest)
{
NS_ASSERTION(mContentType.IsEmpty(), "Content type is already known.");
if (!mContentType.IsEmpty()) return;
const char* testData = mBuffer;
uint32_t testDataLen = mBufferLen;
// Check if data are compressed.
nsCOMPtr<nsIHttpChannel> channel(do_QueryInterface(aRequest));
if (channel) {
nsresult rv = ConvertEncodedData(aRequest, mBuffer, mBufferLen);
if (NS_SUCCEEDED(rv)) {
if (!mDecodedData.IsEmpty()) {
testData = mDecodedData.get();
testDataLen = std::min(mDecodedData.Length(), MAX_BUFFER_SIZE);
}
}
}
// First, run through all the types we can detect reliably based on
// magic numbers
uint32_t i;
for (i = 0; i < sSnifferEntryNum; ++i) {
if (testDataLen >= sSnifferEntries[i].mByteLen && // enough data
memcmp(testData, sSnifferEntries[i].mBytes, sSnifferEntries[i].mByteLen) == 0) { // and type matches
NS_ASSERTION(sSnifferEntries[i].mMimeType ||
sSnifferEntries[i].mContentTypeSniffer,
"Must have either a type string or a function to set the type");
NS_ASSERTION(!sSnifferEntries[i].mMimeType ||
!sSnifferEntries[i].mContentTypeSniffer,
"Both a type string and a type sniffing function set;"
" using type string");
if (sSnifferEntries[i].mMimeType) {
mContentType = sSnifferEntries[i].mMimeType;
NS_ASSERTION(!mContentType.IsEmpty(),
"Content type should be known by now.");
return;
}
if ((this->*(sSnifferEntries[i].mContentTypeSniffer))(aRequest)) {
NS_ASSERTION(!mContentType.IsEmpty(),
"Content type should be known by now.");
return;
}
}
}
NS_SniffContent(NS_DATA_SNIFFER_CATEGORY, aRequest,
(const uint8_t*)testData, testDataLen, mContentType);
if (!mContentType.IsEmpty()) {
return;
}
if (SniffForHTML(aRequest)) {
NS_ASSERTION(!mContentType.IsEmpty(),
"Content type should be known by now.");
return;
}
// We don't know what this is yet. Before we just give up, try
// the URI from the request.
if (SniffURI(aRequest)) {
NS_ASSERTION(!mContentType.IsEmpty(),
"Content type should be known by now.");
return;
}
LastDitchSniff(aRequest);
NS_ASSERTION(!mContentType.IsEmpty(),
"Content type should be known by now.");
}
