Wrong Request URL in IRenderProcessMessageHandler.OnContextCreated on Redirect?

[dbuechel]

We're using IRenderProcessMessageHandler.OnContextCreated(...) to inject a JavaScript API in our software: https://github.com/SafeExamBrowser/seb-win-refactoring/blob/master/SafeExamBrowser.Browser/Handlers/RenderProcessMessageHandler.cs#L33. We have now found out that on redirects, webBrowser.Address does not provide the correct URL of the redirected page, but rather the URL of the previous page from which one has been redirected (see SafeExamBrowser/seb-win-refactoring#84 (comment) for more details).

Is there some other API that should be used to inject JavaScript on redirects, or could it be that this is a bug in the API?

[amaitland]

What version are you using?

We have now found out that on redirects, webBrowser.Address does not provide the correct URL of the redirected page, but rather the URL of the previous page from which one has been redirected

Each frame will have a V8Context, so using Address is always likely to be problematic. You can get the URL for the frame. http://cefsharp.github.io/api/95.7.x/html/P_CefSharp_IFrame_Url.htm

Address is also updated asynchronously, when the property is updated may differ between chromium versions.

[dbuechel]

What sort of redirect is it? Header? Meta? Javascript?

It appears to be a JavaScript redirect, see https://scottdellinger.com/SEBTarget.php and https://scottdellinger.com/SEBRedirect.php

Is the redirect cross origin?

No, it isn't cross origin.

What exactly is the behaviour you are seeing? Can you walk me through what's going on here so I get a better understanding.

We calculate a hash value (see first link above) which incorporates the URL of the actual page and set that hash as a JavaScript constant for the actual page. This works fine, unless the page is being redirected to. In that case, the URL appears to be wrong, i.e. the URL of the previous, redirecting page and thus the calculation of the hash is incorrect. Performing a page reload on such pages however fixes the issue.

Does FrameLoadStart get called with the correct frame information?

Yes, FrameLoadStart appears to provide the correct URL (see log excerpt below). Disabling Chromium site isolation however unfortunately does not resolve the issue.

2021-12-16 16:41:31.093 [26] - ERROR: [Browser Instance #1] Frame Load Start: https://scottdellinger.com/SEBTarget.php, Explicit, DirectLoad
2021-12-16 16:41:31.096 [26] - DEBUG: [Browser Instance #1] Navigated to 'https://scottdellinger.com/SEBTarget.php'.
2021-12-16 16:41:31.167 [26] - WARNING: [RenderProcessMessageHandler #1] BEK: 03aff983d03efcb9adbae6076dc73650e73108d2a48df00718e02d71742eb961
2021-12-16 16:41:31.168 [26] - WARNING: [RenderProcessMessageHandler #1] CK: 8af361a19f07e93783181c45f14c82d40822c7cbce9549266e415e5b4270d91f
2021-12-16 16:41:31.168 [26] - WARNING: [RenderProcessMessageHandler #1] URL (frame): 'https://scottdellinger.com/SEBTarget.php'
2021-12-16 16:41:31.169 [26] - WARNING: [RenderProcessMessageHandler #1] URL (webBrowser): 'https://scottdellinger.com/SEBTarget.php'
2021-12-16 16:41:31.541 [26] - WARNING: [RenderProcessMessageHandler #1] BEK: 03aff983d03efcb9adbae6076dc73650e73108d2a48df00718e02d71742eb961
2021-12-16 16:41:31.542 [26] - WARNING: [RenderProcessMessageHandler #1] CK: 8af361a19f07e93783181c45f14c82d40822c7cbce9549266e415e5b4270d91f
2021-12-16 16:41:31.543 [26] - WARNING: [RenderProcessMessageHandler #1] URL (frame): 'https://scottdellinger.com/SEBTarget.php'
2021-12-16 16:41:31.544 [26] - WARNING: [RenderProcessMessageHandler #1] URL (webBrowser): 'https://scottdellinger.com/SEBTarget.php'
2021-12-16 16:41:31.545 [26] - ERROR: [Browser Instance #1] Frame Load Start: https://scottdellinger.com/SEBRedirect.php, LinkClicked
2021-12-16 16:41:31.546 [26] - DEBUG: [Browser Instance #1] Navigated to 'https://scottdellinger.com/SEBRedirect.php'.

Clicking the redirect button on the page mentioned above...

2021-12-16 16:41:46.334 [26] - WARNING: [RenderProcessMessageHandler #1] BEK: 5649623ac329299124c9a36539b6c273e8422cac23b4d3498e2a3c107daf5b00
2021-12-16 16:41:46.335 [26] - WARNING: [RenderProcessMessageHandler #1] CK: f70a386b103a4ed688ad3315edab4cbb77a37118500903ceb6bf384dc46291f7
2021-12-16 16:41:46.336 [26] - WARNING: [RenderProcessMessageHandler #1] URL (frame): 'https://scottdellinger.com/SEBRedirect.php'
2021-12-16 16:41:46.336 [26] - WARNING: [RenderProcessMessageHandler #1] URL (webBrowser): 'https://scottdellinger.com/SEBRedirect.php'
2021-12-16 16:41:46.338 [26] - ERROR: [Browser Instance #1] Frame Load Start: https://scottdellinger.com/SEBTarget.php, LinkClicked
2021-12-16 16:41:46.339 [26] - DEBUG: [Browser Instance #1] Navigated to 'https://scottdellinger.com/SEBTarget.php'.

2021-12-16 16:41:58.297 [01] - DEBUG: [Browser Instance #1] The user confirmed reloading the current page...

2021-12-16 16:41:58.530 [26] - WARNING: [RenderProcessMessageHandler #1] BEK: 03aff983d03efcb9adbae6076dc73650e73108d2a48df00718e02d71742eb961
2021-12-16 16:41:58.530 [26] - WARNING: [RenderProcessMessageHandler #1] CK: 8af361a19f07e93783181c45f14c82d40822c7cbce9549266e415e5b4270d91f
2021-12-16 16:41:58.531 [26] - WARNING: [RenderProcessMessageHandler #1] URL (frame): 'https://scottdellinger.com/SEBTarget.php'
2021-12-16 16:41:58.532 [26] - WARNING: [RenderProcessMessageHandler #1] URL (webBrowser): 'https://scottdellinger.com/SEBTarget.php'
2021-12-16 16:41:58.533 [26] - ERROR: [Browser Instance #1] Frame Load Start: https://scottdellinger.com/SEBTarget.php, Reload
2021-12-16 16:41:58.534 [26] - DEBUG: [Browser Instance #1] Navigated to 'https://scottdellinger.com/SEBTarget.php'.

[amaitland]

Yes, FrameLoadStart appears to provide the correct URL

Are the frame identifiers the same in FrameLoadStart and OnContextCreated?

OnContextCreated is really very simple from a CefSharp point of view, simply sends a message in the render process to the browser process when the V8Context is created. CEF deals with all the frame related handling.

• https://github.com/cefsharp/CefSharp/blob/cefsharp/96/CefSharp.BrowserSubprocess.Core/CefAppUnmanagedWrapper.cpp#L192
• https://github.com/cefsharp/CefSharp/blob/cefsharp/96/CefSharp.Core.Runtime/Internals/ClientAdapter.cpp#L1268

This is probably a bug in CEF. Using FrameLoadStart is likely your best alternative.

Disabling Chromium site isolation however unfortunately does not resolve the issue.

Might be worth trying to disable the back forward cache. Relevant command line argument is listed in #3902 (comment)

[dbuechel]

Thanks a lot for your help. Yes, the frame identifiers are the same. I have tried disabling the back forward cache, unfortunately to no avail as well.

Using FrameLoadStart is likely your best alternative.

How would I go about using FrameLoadStart, specifically regarding injection of our JavaScript API? I thought in that event it isn't yet possible to inject JavaScript and thus used OnContextCreated for our implementation.

[amaitland]

How would I go about using FrameLoadStart, specifically regarding injection of our JavaScript API?

Exactly the same. Execute the script against the given frame.

I thought in that event it isn't yet possible to inject JavaScript

Injecting JavaScript is fine.

Commands to be executed in the renderer process that may fail during early frame initialization (ExecuteJavaScript, LoadRequest, etc.) will now be queued until after the JavaScript context for the frame has been created.

As per https://bitbucket.org/chromiumembedded/cef/issues/2421/add-onframecreated-onframedeleted#comment-60616371

ExecuteJavaScriptAsync should now be queued. I haven't yet updated the CefSharp documentation as i haven't done any testing to understand the exact behaviour change in CEF.

For reference there are also new frame related notifications via http://cefsharp.github.io/api/96.0.x/html/P_CefSharp_WinForms_ChromiumWebBrowser_FrameHandler.htm

[dbuechel]

Excellent, thanks a lot for your help. Using FrameLoadStart now and everything seems to work as expected, the issue appears to be fixed.