Community
Participate
Working Groups
Build Identifier: Since fix of issue 334758 (RAP 1.4 M6) java script errors outside of the request are also caught and an error page is displayed. On the one hand we see that several errors are found now which in parts exist a long time. This surely improves the quality. On the other hand in a productive environment this is very annoying because the user has to reload the page and in our case he must login again. In our application the reason for the errors are cell tooltips, disposed tables, ... I think it would be great to have a possiblity to deactivate the error pages which are shown caused by errors outside the requests. For example, the error pages are shown in debug (developer) mode, but not in standard (productive) mode. Hence, it would be possible to recognize the java script errors while testing but the end user does not see an error page as frequently. Newsgroup entry about reverting the behavior: http://www.eclipse.org/forums/index.php/t/220370/ Reproducible: Always
Created attachment 206595 [details] patch for 1.4 branch We decidid to introduce this behavior for RAP 1.4.2. In standard mode, the error will be logged to the javascript console if present, and otherwise ignored. In the debug-variant it still crashes the entire client application. Errors within a server-response are also still crashing, as the server has to rely on the client being in sync. As argued by the reporter, total crashes are very enduser-unfriendly. Often these out-of-request errors are non-critical issues, like a failed DOM access while rendering a hover-effect. In addition, if such an error did not occour during developlent, its likely that they may only appear on a very specific configuration, like a certain browser-version with a plugin/addon of a certain version.
Created attachment 206656 [details] revised patch for 1.4 branch Revised patch using typeof to check for console.
I agree with Tim. Since this kind of DOM errors is mostly beyond our control and mostly less critical, letting the entire application crash is too aggressive for production mode. As requested, the change only affects errors that occur outside of requests. Committed the second patch to 1.4 maintenance branch, so the fix will be part of 1.4.2. I think we should do the same for HEAD.
(In reply to comment #3) > Committed the second patch to 1.4 maintenance branch, so the fix will be part > of 1.4.2. > I think we should do the same for HEAD. First, thanks for solving this issue for RAP 1.4.2. Now I don't have to generate the client.js by myself anymore. I agree with Ralf and think this should be introduced on HEAD, too.
Commited changs to CVS HEAD. Also made some small improvements to the printed message itself. (Prints stack if browser supports it).
