With dom.ipc.processCount > 1 some devtools tests fail. The reason seem to be that hud = yield openConsole() is returning something else than it should. devtools/client/webconsole/test/browser_bug1045902_console_csp_ignore_reflected_xss_message.js | Timed out while waiting for: Confirming that CSP logs messages to the console when ‘reflected-xss’ directive is used! - devtools/client/webconsole/test/browser_webconsole_bug_1010953_cspro.js | Timed out while waiting for: Confirmed that CSP and CSP-Report-Only log different messages to the console. - devtools/client/webconsole/test/browser_webconsole_bug_1010953_cspro.js | A promise chain failed to handle a rejection: - at chrome://mochikit/content/browser-test.js:799 - TypeError: invalid 'in' operand ex devtools/client/webconsole/test/browser_webconsole_bug_1247459_violation.js | Test timed out devtools/client/webconsole/test/browser_webconsole_bug_1247459_violation.js | failed to match rule: CSP policy URI warning displayed successfully - devtools/client/webconsole/test/browser_webconsole_bug_770099_violation.js | Timed out while waiting for: CSP policy URI warning displayed successfully - devtools/client/webconsole/test/browser_webconsole_bug_770099_violation.js | A promise chain failed to handle a rejection: - at chrome://mochikit/content/browser-test.js:799 - TypeError: invalid 'in' operand ex devtools/client/webconsole/test/browser_webconsole_bug_770099_violation.js | Test timed out - devtools/client/webconsole/test/browser_webconsole_show_subresource_security_errors.js | Timed out while waiting for: Subresource STS warning displayed successfully - devtools/client/webconsole/test/browser_webconsole_show_subresource_security_errors.js | A promise chain failed to handle a rejection: - at chrome://mochikit/content/browser-test.js:799 - TypeError: invalid 'in' operand ex devtools/client/webconsole/test/browser_webconsole_show_subresource_security_errors.js | Test timed out -

After understanding the complex and painful call path for logging, that propagates messages both in JS and C++ from child to the parent (just to confuse the poor reader) it turned out that the reason for these messages not showing up on the console is a simple uint32_t vs uint64_t. I wonder how could this land without warning but don't have time to figure it out. If the window id is big enough and gets truncated we decide that the window is unknown and don't report the error. This fixes a lot of test, and possible some intermittents without e10s-multi too.

Blocking the bug that introduced this bug.

Comment on attachment 8787539 [details] [diff] [review] using uint32_t instead of uint64_t in CSP log functions. v1 Review of attachment 8787539 [details] [diff] [review]: ----------------------------------------------------------------- I think all of this stuff is written in JS, so there wouldn't be any warnings.

Bug 1298680 landed earlier.