Linux/OSX: browser creation fails once in a while

[cztomczak]

It happened only once. Couldn't reproduce anymore.

Was it CEF fault? or GTK fault?

From the logs, compared to other successful launches, I see that there was one message missing, about the sandbox missing "Running renderer without sandbox". So it seems that the call to cef_create_browser_sync() failed. Did it return a null pointer? Need to add a check for that.

It would be a good idea to do some stress testing. Do 1000 launches of application and see if we can reproduce the problem.

[cztomczak]

Added commit 1a7d869 to check whether browser failed.

[cztomczak]

Reported problem on CEF Forum:
http://www.magpcss.org/ceforum/viewtopic.php?f=6&t=11470

[cztomczak]

Marshall says that there were numerous reports of problems when using the synchronous version of browser creation function. The solution will be to create browser asynchronously. A minor fix for windows is required to call WindowResized() after some delay.

Fix commited in revision f95b131.

[cztomczak]

Unfortunately the problem still persists. Browser creation fails. But this time I don't see an empty window. When browser creation fails in CEF, then CEF sends a signal to destroy window. Thus cef_quit_message_loop() is called and application quits.

Happens about 1/40 launches of application.

[cztomczak]

Same problem noticed also on OSX, branch 1750.

It seems that branch 1750 is just unstable. Latest Google Chrome stable is 32 / branch 1700. We will have to wait for until 1750 makes it into mainstream.

Still it would be good to create an example that can reproduce the issue, using cefsimple C++ example. See the topic on the CEF Forum from third post.

[cztomczak]

After applying some fixes to CEF structures initialization, and by providing app and client handlers to cef_execute_process/cef_initialize (see Issue #10 ), I can't reproduce this issue anymore on Linux.

The issue still persists on OS X.

[jhvst]

Same bug noticed on Ubuntu 12.04. I'm using the resources you've linked in the latest readme.

[cztomczak]

There is an Issue created in CEF to stress test application launching, to detect and fix this problem:
https://code.google.com/p/chromiumembedded/issues/detail?id=1207

The crash may not necessarily be happening during browser creation. I've observed similar issues in branch 1650 on Linux when using CEF C++ API in CEF Python. The crash occurs during launch of a subprocess after a call to CefInitialize, no browser is being created. See this post for more details:
http://www.magpcss.org/ceforum/viewtopic.php?p=22227#p22227

[cztomczak]

In Issue #13 quirkey states that this issue disappeared after implementing CefLifespanHandler. He also made some other changes to code that could affect CEF execution, see quirkey's fork for details.

EDIT. I meant Issue 13.

[cztomczak]

After creating some tests the issue was still reproducible in quirkey's fork.

I have some good news. By offloading CPU during CEF initialization I was able to get rid of segmentation faults in CEF Python. This should work for CEF2go as well. See:
https://code.google.com/p/cefpython/issues/detail?id=131#c2

EDIT: quirkey confirmed that this fixed the issue in cef2go. See #13 (comment)

[cztomczak]

A corresponding issue in CEF tracker (star it so that it gets more attention):
https://code.google.com/p/chromiumembedded/issues/detail?id=1362