for a long time, gecko conspired with spidermonkey to give a free pass from the request model to js code running on the main thread, as long as it didn't try to pass the context to another thread. i can't remember when spidermonkey revoked that pass, but it was a while ago, and no one updated the tests.
Created attachment 386225 [details] [diff] [review] use requests i just need someone to confirm that i've used the request model correctly. note that this runs w/o fails, it does get plenty of assertions ("Script compiled without principals").
Comment on attachment 386225 [details] [diff] [review] use requests I owe jorendorff a review and I am out on paternity leave, so redirecting to Jason. /be
Comment on attachment 386225 [details] [diff] [review] use requests Thanks for updating these. In the new EvaluateScript function, please do printf("%s", msg);. I think gcc warns about that otherwise. In TestArgFormatter, it looks like you could just declare the AutoRequest at the top and stay in a request for the duration of the function. Then you could avoid the do-while(0) loop. That approach might also work in TestSecurityManager. I didn't look closely. Would you be willing to s/%%/%/ and again use printf("%s", msg) instead of a plain printf(msg)?
Comment on attachment 386225 [details] [diff] [review] use requests I meant to say, r=me with those changes, and once again thanks for the patch. I'm not a superreviewer; maybe mrbkap...?
Created attachment 389590 [details] [diff] [review] updated i've decided not to broaden the autorequests, i'm treating them as units and for tests this is perfectly fine.