[superseded] Issue 1265 - Improve detection of mime type of the expected response

src/plugin/PluginWbPassThrough.cpp

Index: src/plugin/PluginWbPassThrough.cpp
===================================================================
--- a/src/plugin/PluginWbPassThrough.cpp
+++ b/src/plugin/PluginWbPassThrough.cpp
@@ -9,7 +9,7 @@
 #include "PluginSettings.h"
 #include "PluginClass.h"
 #include "PluginSystem.h"
-
+#include <WinInet.h>
 #include "wtypes.h"
 
 namespace
@@ -22,6 +22,12 @@
     "</html>";
 }
 
+WBPassthruSink::WBPassthruSink()
+  : m_currentPositionOfSentPage(0)
+  , m_contentType(CFilter::EContentType::contentTypeAny)
+{
+}
+
 int WBPassthruSink::GetContentTypeFromMimeType(const CString& mimeType)
 {
   if (mimeType.Find(L"image/") >= 0)
@@ -44,12 +50,6 @@
   {
     return CFilter::contentTypeSubdocument;
   }
-  // It is important to have this check last, since it is rather generic, and might overlay text/html, for example
-  if (mimeType.Find(L"xml") >= 0)
-  {
-    return CFilter::contentTypeXmlHttpRequest;
-  }
-
   return CFilter::contentTypeAny;
 }
 
@@ -79,10 +79,6 @@
   {
     return CFilter::contentTypeScript;
   }
-  else if (ext == L".xml")
-  {
-    return CFilter::contentTypeXmlHttpRequest;
-  }
   else if (ext == L".swf")
   {
     return CFilter::contentTypeObject;
@@ -91,8 +87,7 @@
   {
     return CFilter::contentTypeSubdocument;
   }
-
-  return CFilter::contentTypeAny & ~CFilter::contentTypeSubdocument;
+  return CFilter::contentTypeAny;
 }
 
 int WBPassthruSink::GetContentType(const CString& mimeType, const CString& domain, const CString& src)
@@ -122,35 +117,32 @@
 {
   m_pTargetProtocol = pTargetProtocol;
   bool isBlocked = false;
-  m_currentPositionOfSentPage = 0;
   CString src = szUrl;
   DEBUG_GENERAL(src);
   CPluginClient::UnescapeUrl(src);
 
-  CString boundDomain;
   CString mimeType;
-  LPOLESTR mime[10];
   if (pOIBindInfo)
   {
     ULONG resLen = 0;
 
     // Apparently IE will report random mime type if there's more then 1 in the list.
     // So we get the whole list and just use the first one (top priority one)
+    LPOLESTR mime[10];
     pOIBindInfo->GetBindString(BINDSTRING_ACCEPT_MIMES, mime, 10, &resLen);
     if (mime && resLen > 0)
     {
       mimeType.SetString(mime[0]);
     }
-    LPOLESTR bindToObject = 0;
+    LPOLESTR bindToObject = nullptr;
     pOIBindInfo->GetBindString(BINDSTRING_FLAG_BIND_TO_OBJECT, &bindToObject, 1, &resLen);
-    LPOLESTR domainRetrieved = 0;
+    LPOLESTR domainRetrieved = nullptr;
     if (resLen == 0 || wcscmp(bindToObject, L"FALSE") == 0)
-    {   
+    {
       HRESULT hr = pOIBindInfo->GetBindString(BINDSTRING_XDR_ORIGIN, &domainRetrieved, 1, &resLen);
-      
       if ((hr == S_OK) && domainRetrieved && (resLen > 0))
       {
-        boundDomain.SetString(domainRetrieved);
+        m_boundDomain = domainRetrieved;
       }
     }
   }
@@ -160,12 +152,9 @@
   ULONG len2 = 2048;
 
 #ifdef SUPPORT_FILTER
-  int contentType = CFilter::contentTypeAny;
-
   CPluginTab* tab = CPluginClass::GetTab(::GetCurrentThreadId());
   CPluginClient* client = CPluginClient::GetInstance();
 
-
   if (tab && client)
   {
     CString documentUrl = tab->GetDocumentUrl();
@@ -176,104 +165,68 @@
     }
     else if (CPluginSettings::GetInstance()->IsPluginEnabled() && !client->IsWhitelistedUrl(std::wstring(documentUrl)))
     {
-      boundDomain = tab->GetDocumentUrl();
-
-      contentType = CFilter::contentTypeAny;
-
+      m_boundDomain = tab->GetDocumentUrl();
+      m_contentType = CFilter::contentTypeAny;
 #ifdef SUPPORT_FRAME_CACHING
-      if ((tab != 0) && (tab->IsFrameCached(src)))
+      if (nullptr != tab && tab->IsFrameCached(src))
       {
-        contentType = CFilter::contentTypeSubdocument;
+        m_contentType = CFilter::contentTypeSubdocument;
       }
       else
 #endif // SUPPORT_FRAME_CACHING
-      contentType = GetContentType(mimeType, boundDomain, src);
-      if (client->ShouldBlock(src, contentType, boundDomain, true))
       {
-        isBlocked = true;
-
-        DEBUG_BLOCKER("Blocker::Blocking Http-request:" + src);
+        m_contentType = GetContentType(mimeType, m_boundDomain, src);
       }
     }
-    if (!isBlocked)
-    {
-      DEBUG_BLOCKER("Blocker::Ignoring Http-request:" + src)
-    }
+    
   }
 
-
-  if (tab == NULL)
+  if (nullptr == tab)
   {
-    contentType = GetContentType(mimeType, boundDomain, src);
-    if (client->ShouldBlock(src, contentType, boundDomain, true))
-    {
-      isBlocked = true;
-    }
+    m_contentType = GetContentType(mimeType, m_boundDomain, src);
   }
 
-#ifdef _DEBUG
-  CString type;
+  if (nullptr != client
+    && CFilter::EContentType::contentTypeAny != m_contentType

[Oleksandr, 2014/09/11 08:58:54]

This line is not clear to me. We can in theory block requests that have
contentTypeAny as well.

[sergei, 2014/09/11 09:53:25]

If the type is `contentTypeAny` then the decision is made in
`BeginningTransaction`, it will be blocked there.

+    && client->ShouldBlock(src, m_contentType, m_boundDomain, true))
+  {
+    isBlocked = true;
+    DEBUG_BLOCKER("Blocker::Blocking Http-request:" + src);

[Oleksandr, 2014/09/11 08:58:54]

DEBUG_BLOCKER is broken here. Either use CPluginDebug::DebugResultBlocking
directly (prefered), or fix the DEBUG_BLOCKER. The debug build should write the
output to debug_result.txt

[sergei, 2014/09/11 09:53:25]

`client->ShouldBlock` already puts into the log. `DEBUG_BLOCKER` is removed.

+  }
+  if (!isBlocked)
+  {
+    DEBUG_BLOCKER("Blocker::Ignoring Http-request:" + src)

[Oleksandr, 2014/09/11 08:58:54]

CPluginDebug::DebugResultIgnoring here accordingly

[sergei, 2014/09/11 09:53:25]

The same as above, deleted.

+  }
 
-  if (contentType == CFilter::contentTypeDocument) type = "DOCUMENT";
-  else if (contentType == CFilter::contentTypeObject) type = "OBJECT";
-  else if (contentType == CFilter::contentTypeImage) type = "IMAGE";
-  else if (contentType == CFilter::contentTypeScript) type = "SCRIPT";
-  else if (contentType == CFilter::contentTypeOther) type = "OTHER";
-  else if (contentType == CFilter::contentTypeUnknown) type = "OTHER";
-  else if (contentType == CFilter::contentTypeSubdocument) type = "SUBDOCUMENT";
-  else if (contentType == CFilter::contentTypeStyleSheet) type = "STYLESHEET";
-  else type = "OTHER";

[Oleksandr, 2014/09/11 08:58:54]

We will still need a way to convert contentType to text representation for
debugging purposes.

[sergei, 2014/09/11 09:53:25]

It's done in `CPluginFilter`

-
-  if (isBlocked)
+  // For IE6 and earlier there is iframe back button issue, so avoid it.
+  if (isBlocked && client->GetIEVersion() > 6)
   {
-    CPluginDebug::DebugResultBlocking(type, src, boundDomain);
-  }
-  else
-  {
-    CPluginDebug::DebugResultIgnoring(type, src, boundDomain);
-  }
-#endif
-
-  //Fixes the iframe back button issue
-  if (client->GetIEVersion() > 6)
-  {
-    if (contentType == CFilter::contentTypeImage && isBlocked)
+    handled = true;
+    if (CFilter::EContentType::contentTypeImage == m_contentType)
     {
       BaseClass::OnStart(szUrl, pOIProtSink, pOIBindInfo, grfPI, dwReserved, pTargetProtocol);
-      handled = true;
       // IE shows a cross that img is not loaded
       return INET_E_REDIRECT_FAILED;
     }
-    if (contentType == CFilter::contentTypeSubdocument && isBlocked)
+    if (CFilter::EContentType::contentTypeSubdocument == m_contentType)
     {
       PassthroughAPP::CustomSinkStartPolicy<WBPassthru, WBPassthruSink>::GetProtocol(this)->m_shouldSupplyCustomContent = true;
       BaseClass::OnStart(szUrl, pOIProtSink, pOIBindInfo, grfPI, dwReserved, pTargetProtocol);
-
       m_spInternetProtocolSink->ReportProgress(BINDSTATUS_MIMETYPEAVAILABLE, L"text/html");
       m_spInternetProtocolSink->ReportData(BSCF_FIRSTDATANOTIFICATION, 0, static_cast<ULONG>(g_myPageBlocked.size()));
-      handled = true;
       return S_OK;
     } 
-    if (contentType == CFilter::contentTypeScript && isBlocked)
+    if (CFilter::EContentType::contentTypeScript == m_contentType)
     {
       BaseClass::OnStart(szUrl, pOIProtSink, pOIBindInfo, grfPI, dwReserved, pTargetProtocol);
       m_spInternetProtocolSink->ReportProgress(BINDSTATUS_MIMETYPEAVAILABLE, L"text/javascript");
       m_spInternetProtocolSink->ReportResult(INET_E_REDIRECTING, 301, L"data:");
-      handled = true;
       return INET_E_REDIRECT_FAILED;
     }
-    if (contentType == CFilter::contentTypeXmlHttpRequest && isBlocked)
+    if (CFilter::EContentType::contentTypeXmlHttpRequest == m_contentType)
     {
       BaseClass::OnStart(szUrl, pOIProtSink, pOIBindInfo, grfPI, dwReserved, pTargetProtocol);
       m_spInternetProtocolSink->ReportResult(INET_E_REDIRECTING, 301, L"data:");
-      handled = true;
-      return INET_E_REDIRECT_FAILED;
-    }
-    if (isBlocked)
-    {
-      BaseClass::OnStart(szUrl, pOIProtSink, pOIBindInfo, grfPI, dwReserved, pTargetProtocol);
-      m_spInternetProtocolSink->ReportResult(S_FALSE, 0, L"");
-      handled = true;
       return INET_E_REDIRECT_FAILED;
     }
   }
@@ -282,7 +235,6 @@
   return isBlocked ? S_FALSE : BaseClass::OnStart(szUrl, pOIProtSink, pOIBindInfo, grfPI, dwReserved, pTargetProtocol);
 }
 
-
 HRESULT WBPassthruSink::OnRead(void* pv, ULONG cb, ULONG* pcbRead)
 {
   if (nullptr == pv)
@@ -348,13 +300,62 @@
   return m_spInternetProtocolSink ? m_spInternetProtocolSink->Switch(pProtocolData) : E_UNEXPECTED;
 }
 
-STDMETHODIMP WBPassthruSink::BeginningTransaction(LPCWSTR szURL, LPCWSTR szHeaders, DWORD dwReserved, LPWSTR *pszAdditionalHeaders)
+STDMETHODIMP WBPassthruSink::BeginningTransaction(LPCWSTR szURL, LPCWSTR szHeaders, DWORD dwReserved, LPWSTR* pszAdditionalHeaders)
 {
   if (pszAdditionalHeaders)
   {
-    *pszAdditionalHeaders = 0;
+    *pszAdditionalHeaders = nullptr;
   }
 
+  CPluginClient* client = nullptr;
+  if (CFilter::EContentType::contentTypeAny == m_contentType && (client = CPluginClient::GetInstance()))
+  {
+    auto acceptHeader = [&]() -> std::string

[Oleksandr, 2014/09/11 08:58:54]

I'd just like to say this looks too cool :D

[sergei, 2014/09/11 09:53:25]

Thanks!

+    {
+      ATL::CComPtr<IWinInetHttpInfo> winInetHttpInfo;
+      HRESULT hr = m_spTargetProtocol->QueryInterface(&winInetHttpInfo);
+      if(FAILED(hr))
+      {
+        return "";
+      }
+      DWORD size = 0;
+      DWORD flags = 0;
+      hr = winInetHttpInfo->QueryInfo(HTTP_QUERY_RAW_HEADERS_CRLF | HTTP_QUERY_FLAG_REQUEST_HEADERS,
+          /*buffer*/nullptr, /* get size */&size, &flags, /*reserved*/ 0);
+      if(FAILED(hr))
+      {
+        return "";
+      }
+      std::string buf(size, '\0');
+      hr = winInetHttpInfo->QueryInfo(HTTP_QUERY_RAW_HEADERS_CRLF | HTTP_QUERY_FLAG_REQUEST_HEADERS,
+        &buf[0], &size, &flags, 0);
+      if(FAILED(hr))
+      {
+        return "";
+      }
+      char acceptHeader[] = "Accept:";
+      auto acceptHeaderBeginsAt = buf.find(acceptHeader);
+      if (std::string::npos == acceptHeaderBeginsAt)
+      {
+        return "";
+      }
+      acceptHeaderBeginsAt += sizeof(acceptHeader);
+      auto acceptHeaderEndsAt = buf.find("\n", acceptHeaderBeginsAt);
+      if (std::string::npos == acceptHeaderEndsAt)
+      {
+        return "";
+      }
+      return buf.substr(acceptHeaderBeginsAt, acceptHeaderEndsAt - acceptHeaderBeginsAt);
+    }();
+    m_contentType = GetContentTypeFromMimeType(ATL::CString(acceptHeader.c_str()));
+    bool isBlocked = client->ShouldBlock(szURL, m_contentType,
+      //client->GetHostFromUrl(m_boundDomain.operator LPCWSTR()).c_str(), /*debug flag but must be set*/true);

[Oleksandr, 2014/09/11 08:58:54]

Leftover code here?

[sergei, 2014/09/11 09:53:25]

Thanks, removed.

+      m_boundDomain, /*debug flag but must be set*/true);
+    if (isBlocked)
+    {

[Oleksandr, 2014/09/11 08:58:54]

Log blocking here as well

[sergei, 2014/09/11 09:53:25]

`client->ShouldBlock` puts into log

+      return E_ABORT;
+    }
+  }
   CComPtr<IHttpNegotiate> spHttpNegotiate;
   QueryServiceFromClient(&spHttpNegotiate);
   return spHttpNegotiate ? spHttpNegotiate->BeginningTransaction(szURL, szHeaders,dwReserved, pszAdditionalHeaders) : S_OK;