Pack URL bug

[o2genum]

The new pack:/// url feature seems to be buggy.

From time to time Chromium is unable to read random packed files (I run the same executable each time, without rebuilding, and sometimes it happens). For big files it happens more frequently. Add something like 30K-line javascript to reproduce it. Unexpected end of input, unexpected token, etc., each time on a random line.

If I click build.js:1 to go to that line, dev tools close. The affected files are missing Sources tab. It I click Application tab, dev tools close.

Maybe the contributor @lhyqy5 knows what's wrong.

[David-Desmaisons]

Could you share the corresponding project/file?

[o2genum]

I'll make up a sample project to reproduce it.

[o2genum]

Here it is. Just a clean solution template + added a heavy main.js.

Open chromium dev tools, about every second run, it logs an error to the console.

https://github.com/o2genum/NeutroniumPackUriExample

[o2genum]

Must be related to the copying of data between the streams: PackUriResourceHandler.cs

Related ChromoumFX docs.

[David-Desmaisons]

@o2genum , I just corrected it. It was one nasty issue due to the way of ChromiumFx manage memory (using weak references) that may lead to hard to reproduce bug.

[o2genum]

Ok. I suspected the copying of data in PackUriResourceHandler.cs and it is coded a bit awkwardly, indeed. It only works because ResponseLength is set.

I corrected it according to the ChromiumFX docs (the code below), maybe you can include this fix, too. This is unrelated to the issue, though.

Will you make a release for this fix?

            ReadResponse += (s2, e2) =>
            {
                if (resInfo != null)
                {
                    var buffer = new byte[e2.BytesToRead];
                    var bytesRead = resInfo.Stream.Read(buffer, 0, e2.BytesToRead);
                    System.Runtime.InteropServices.Marshal.Copy(buffer, 0, e2.DataOut, bytesRead);
                    e2.BytesRead = bytesRead;

                    if (bytesRead != 0)
                    {
                        e2.SetReturnValue(true);
                    }
                    else
                    {
                        e2.SetReturnValue(false);
                    }
                }
            };

            ProcessRequest += (s3, e3) =>
            {
                e3.SetReturnValue(true);
                e3.Callback.Continue();
            };

[David-Desmaisons]

Are you refering to this alteration:

   if (bytesRead != 0)
   {
         e2.SetReturnValue(true);
   }
  else
   {
         e2.SetReturnValue(false);
   }

?

The new version should be available shortly.

Could you comunicate what kind of application you are creating with Neutronium?
Would you be ok to share some screenshots on Neutronium documentation to ilustre aplication made with Neutronium?

[o2genum]

This alteration, also unnecessary Continue() in ReadResponce, also wrong order (Continue(), then SetReturnValue()) in ReadRequest and ProcessRequest.

Ok, I'll share the app description and screenshots when I got it working (we're in process of switching to Neutronium-based UI).

[David-Desmaisons]

Ok Thanks.
I will take a look at the alteration, and commit them after review.

[David-Desmaisons]

I updated handler based on this two references:

https://bitbucket.org/chromiumfx/chromiumfx/src/25c83dc7ef6f204e85645725bf27fdfcd639e176/ChromiumWebBrowser/WebResource.cs?at=default&fileviewer=file-view-default

https://bitbucket.org/chromiumfx/chromiumfx/issues/21/how-do-i-add-a-custom-schema

I did not find any reference on e2.SetReturnValue(false);

[o2genum]

Ok. I only looked at what the docs say.

e.SetReturnValue(false) is for signalling end of input, but when e.ResponseLength is set, this is not necessary (Chromium stops reading before the end of input is met). So I think it's ok as it is.

[David-Desmaisons]

Thanks for the link. I made another alteration to the original base on this doc.