Added the frame instance to the OnCertificateError client handler API

[GoogleCodeExporter]

What steps will reproduce the problem?
1. The frame instance is needed to identify cases where the error occurs on the 
main frame.
2.
3.

What is the expected output? What do you see instead?
N/A

What version of the product are you using? On what operating system?
CEF3 Branch 2285, r. 2040, linux64

Please provide any additional information below.

patch attached

Original issue reported on code.google.com by amitkan...@gmail.com on 3 Mar 2015 at 10:10

Attachments:

• 0001-Added-frame-parameter-to-the-OnCertificateError-API.patch

[GoogleCodeExporter]

Thanks for the patch. Some comments/questions:

1. Certificate errors for resources other than the main frame are suppressed 
currently by the following code in 
CefContentBrowserClient::AllowCertificateError:

  if (resource_type != content::ResourceType::RESOURCE_TYPE_MAIN_FRAME) {
    // A sub-resource has a certificate error. The user doesn't really
    // have a context for making the right decision, so block the request
    // hard.
    *result = content::CERTIFICATE_REQUEST_RESULT_TYPE_CANCEL;
    return;
  }

This matches the behavior in Chrome (see 
ChromeContentBrowserClient::AllowCertificateError). What is your use case for 
allowing certificate errors from sub-resources?  Which specific types of 
sub-resources do you want to allow certificate errors for?

If certificate errors are allowed for sub-resources then the resource type 
should also be passed to OnCertificateError so the application has all 
necessary information to make an informed decision.

2. Line 22:

+  CefRefPtr<CefFrame> frame =
+    browser->GetFrame(render_frame_id);

No need to wrap lines that are less than 80 characters.

+  if (!frame.get())
+    return;

If the frame ID is unknown (GetFrame returns NULL) we should pass the main 
frame (browser->GetMainFrame()) instead of not executing the callback.

3. Pass the frame argument to LoadErrorPage in 
ClientHandler::OnCertificateError.

4. Create patch files using the `git diff --no-prefix` command (this removes 
the a/ b/ prefixes from the file paths).

Original comment by magreenb...@gmail.com on 6 Mar 2015 at 6:54

[GoogleCodeExporter]

@#1: Note that if certificate errors are not allowed for sub-resources then 
there is no need to pass the frame as an argument since the callback will only 
ever be executed for the main frame.

Original comment by magreenb...@gmail.com on 6 Mar 2015 at 7:00

[GoogleCodeExporter]

Marshall, we reviewed again our use-case, and now that we know that this 
callback occurs only on the main frame - it really seems a bit silly to pass 
the frame.
The idea was to provide an indication to the user that a frame had an error 
(even it it was an iframe), but since chrome doesn't do this - it's probably OK 
for us to ignore such errors.

Thanks for you time.

Original comment by amitkan...@gmail.com on 8 Mar 2015 at 5:09

[GoogleCodeExporter]

Original comment by magreenb...@gmail.com on 9 Mar 2015 at 5:01

• Changed state: WontFix