try_chrome_dialog_view_browsertest.cc revision 5821806d5e7f356e8fa4b058a389a808ea183019
1// Copyright (c) 2012 The Chromium Authors. All rights reserved. 2// Use of this source code is governed by a BSD-style license that can be 3// found in the LICENSE file. 4 5#include "base/command_line.h" 6#include "chrome/common/chrome_switches.h" 7#include "chrome/test/base/in_process_browser_test.h" 8 9// Unfortunately, this needs to be Windows only for now. Even though this test 10// is meant to exercise code that is for Windows only, it is a good general 11// canary in the coal mine for problems related to early shutdown (aborted 12// startup). Sadly, it times out on platforms other than Windows, so I can't 13// enable it for those platforms at the moment. I hope one day our test harness 14// will be improved to support this so we can get coverage on other platforms. 15// See http://crbug.com/45115 for details. 16#if defined(OS_WIN) 17 18// By passing kTryChromeAgain with a magic value > 10000 we cause Chrome 19// to exit fairly early. 20// Quickly exiting Chrome (regardless of this particular flag -- it 21// doesn't do anything other than cause Chrome to quit on startup on 22// non-Windows) was a cause of crashes (see bug 34799 for example) so 23// this is a useful test of the startup/quick-shutdown cycle. 24class TryChromeDialogBrowserTest : public InProcessBrowserTest { 25public: 26 TryChromeDialogBrowserTest() {} 27 28protected: 29 virtual void SetUpCommandLine(CommandLine* command_line) { 30 command_line->AppendSwitchASCII(switches::kTryChromeAgain, "10001"); 31 } 32}; 33 34// Note to Sheriffs: This test (as you can read about above) simply causes 35// Chrome to shutdown early, and, as such, has proven to be pretty good at 36// finding problems related to shutdown. Sheriff, before marking this test as 37// disabled, please consider that this test is meant to catch when people 38// introduce changes that crash Chrome during shutdown and disabling this test 39// and moving on removes a safeguard meant to avoid an even bigger thorny mess 40// to untangle much later down the line. Disabling the test also means that the 41// people who get blamed are not the ones that introduced the crash (in other 42// words, don't shoot the messenger). So, please help us avoid additional 43// shutdown crashes from creeping in, by doing the following: 44// Run chrome.exe --try-chrome-again=10001. This is all that the test does and 45// should be enough to trigger the failure. If it is a crash (most likely) then 46// look at the callstack and see if any of the components have been touched 47// recently. Look at recent changes to startup, such as any change to 48// ChromeBrowserMainParts, specifically PreCreateThreadsImpl and see if someone 49// has been reordering code blocks for startup. Try reverting any suspicious 50// changes to see if it affects the test. History shows that waiting until later 51// only makes the problem worse. 52IN_PROC_BROWSER_TEST_F(TryChromeDialogBrowserTest, ToastCrasher) {} 53 54#endif // defined(OS_WIN) 55