зеркало из https://github.com/mozilla/gecko-dev.git
Bug 997440 - Fix some error reporting nastiness in the CPS tests. r=bz
This commit is contained in:
Родитель
51dbaf59da
Коммит
e2886812ec
|
@ -42,11 +42,20 @@ function runAsyncTests(tests) {
|
|||
}
|
||||
},
|
||||
consoleError: function (scriptErr) {
|
||||
// As much as possible make sure the error is related to the test. On the
|
||||
// other hand if this fails to catch a test-related error, we'll hang.
|
||||
let filename = scriptErr.sourceName || scriptErr.toString() || "";
|
||||
if (/contentpref/i.test(filename))
|
||||
do_throw(scriptErr);
|
||||
// Previously, this code checked for console errors related to the test,
|
||||
// and treated them as failures. This was problematic, because our current
|
||||
// very-broken exception reporting machinery in XPCWrappedJSClass reports
|
||||
// errors to the console even if there's actually JS on the stack above
|
||||
// that will catch them. And a lot of the tests here intentionally trigger
|
||||
// error conditions on the JS-implemented XPCOM component (see erroneous()
|
||||
// in test_getSubdomains.js, for example). In the old world, we got lucky,
|
||||
// and the errors were never reported to the console due to happenstantial
|
||||
// JSContext reasons that aren't really worth going into.
|
||||
//
|
||||
// So. We make sure to dump this stuff so that it shows up in the logs, but
|
||||
// don't turn them into duplicate failures of the exception that was already
|
||||
// propagated to the caller.
|
||||
dump("AsyncRunner.jsm observed console error: " + scriptErr + "\n");
|
||||
}
|
||||
});
|
||||
|
||||
|
|
Загрузка…
Ссылка в новой задаче