Trying to use URLRequestInfo::SetStreamToFile and URLResponseInfo::GetBodyAsFileRef - c++

Question

I am working on an NaCl plugin for Chrome, and trying to download a URL resource file locally, into the temporary cache of Chrome, but without success.
Here is how I proceed:
// I indicate that I want the resource to be downloaded to a file
m_URLRequestInfo.SetStreamToFile( true );
// I open the request:
m_URLLoader.Open( m_URLRequestInfo, m_CCFactory.NewCallback( &MyClass::OnOpen ) );
...
// My callback (OnOpen) is eventually called.
// I then check to make sure the status code is 200 using this call:
m_URLLoader.GetResponseInfo().GetStatusCode()
// Then I ask to download the whole file:
m_URLLoader.FinishStreamingToFile( m_CCFactory.NewOptionalCallback( &MyClass::OnFileDownloaded ) );
...
// My other callback (OnFileDownloaded) gets eventually called,
// and again the status code is 200.
// Then I query the FileRef using this call:
pp::FileRef l_FileRef = m_URLLoader.GetResponseInfo().GetBodyAsFileRef();
The returned pp::FileRef seems to be fine, but pp::FileRef::GetFileSystemType() returns PP_FILESYSTEMTYPE_EXTERNAL, and then the call to pp::FileRef::GetPath() fails (it returns an UNDEFINED pp::Var).
So from this point, I am lost. I don't know what else I should do to get a valid pp::FileRef that points to a local file in the browser's cache. My final goal is to open this local file (an image file in my case) using a standard system file IO like fopen().
Thanks for any light !

Answer 1

Is there a reason you can't use the nacl_io library instead? With it you can write something like this:
#include <stdio.h>
#include <sys/mount.h>
// Mount an HTTP filesystem that reads files from "http://example.com/path/...".
mount("http://example.com/path/", "/mnt/http", "httpfs", 0, "");
// Performs a URL Request of "http://example.com/path/my_image.png".
FILE* file = fopen("/mnt/http/my_image.png", "r");
...
Take a look a the nacl_io demo in the SDK. It is located at $NACL_SDK_ROOT/examples/demo/nacl_io.

Answer 2

After reading more thoroughly the documentation and running more tests, I finally figured out what I was doing wrong.
When we call pp::URLLoader::FinishStreamingToFile, then the file is downloaded in Browser's cache, but it cannot be opened/read using regular stdio services like fopen, fread etc. We need to use the pp::FileIO class services to open the obtained pp::FileRef and read the content of the file.
So here is what I did to successfully load and read a file that was downloaded for me by the Browser. Basically, I continued to use the C++ PPAPI services.
(1) upon callback from m_URLLoader->FinishStreamingToFile, we then call m_FileIO->Open to open the downloaded file using the obtained FileRef;
(2) upon callback from m_FileIO->Open, we then call m_FileIO->Query to obtain the size of the downloaded file (and some other file attributes);
(3) upon callback from pp::FileIO::Query, we then check the file attribute type (e.g. not a folder), allocate a memory buffer large enough to hold the whole file content, and start to call pp::FileIO::Read to obtain the file's content;
(4) upon callback from pp::FileIO::Read, if the obtained nResult argument is 0, then we reached the EOF and we finished reading the file content into our memory buffer; if the obtained nResult > 0, then it indicates the number of successfully read bytes, and we call m_FileIO->Read again to continue reading bytes, and storing them at a different offset location into our memory buffer; if the obtained nResult < 0 then an error occurred and we must terminate the reading process.
Many steps, and many callbacks to manage, but in the end this works smoothly.