Prevent segmentation fault when Render does not have 3 arguments.

[chobie]

Hi, i'll try to use webkit_server with socket.

Render command cause segmentation fault when it does not have correct argument count.

Thanks,
Shuhei

[jferris]

Thanks for the patch. However, we don't pull in anything without tests. Can you please add tests for this?

[chobie]

Okay, I'll write test case for this patch. By the way, It seems there are no test cases for webkit_server.
Should I use QTestLib?

[jferris]

All the tests are in Ruby, and test the Ruby interface. If there isn't a way to expose an issue without using the Ruby interface, we don't consider that a bug, since the Javascript and network interfaces are internal.

[chobie]

Thank you for your advice.

I've added test case for it.

Capybara::Driver::Webkit.render checks their options. So basically, webkit_server doesn't cause segmentation fault.
for now, I've used command method and invoke Render command directly at this time.
As I'd like to check webkit_server cause segmentation fault.

please close issue if you think this modification are too much :)

[jferris]

@chobie Out of curiosity, why are you calling command directly?

[chobie]

As the render function fixed width and height when given arguments are weird.
I don't have a good idea for reproducing the issue. so I choose using command directly.

Why I create this issue: One day, I looked into the webkit_server implementation.
As I'd like to know how capybara-webkit works fine. It's very straight forward for me.
But some commands has possibility for segmentation fault. IMHO, The webkit_server should not die when
whatever commands given. so I sent the pull request.

Thank you for reading.

[mike-burns]

@chobie I'm slightly confused: were you able to reproduce this in a real app, or is this just a precaution?

[chobie]

@mike-burns I'm sorry for making you confused. It's just a precaution.

[jferris]

Thanks for looking into this, but since there's no realistic way to run into this while using capybara-webkit, I don't think we want to pull this in. The added test will just lengthen the test suite and make it a little harder to refactor.