Currently, we don't have a way to test advertised updates. This differs from unadvertised updates in the following way... Unadvertised ------------ Help > Check for Updates or Firefox > About Advertised ---------- app.update.idletime = 1 app.update.interval = 60 app.update.timer = 60000 Wait for and click notification

ccing Rob Strong on this bug so he can provide his expertise where needed.

Do you mean that there is currently no way to test advertised updates because of bug 646054? If so, wouldn't waiting until bug 646054 lands be the solution?

(In reply to comment #2) > Do you mean that there is currently no way to test advertised updates because > of bug 646054? If so, wouldn't waiting until bug 646054 lands be the solution? Not at all Rob. This is specific to our Mozmill automation. We don't have an automated Mozmill test for advertised updates (only the manual Check For Updates method). This bug is to develop a test to cover the advertised scenario.

That shouldn't be too hard to implement. We should add another command line option for our update test-run, i.e. --advertised. We can pass the type via the persisted object to the tests, which then can differentiate between using the UI or setting the preferences accordingly.

We already have mochitests for testing advertised updates though we don't have tests for when there are extension on Gecko 1.9.2 / Firefox 3.6 like we do on Gecko 2.0 / Firefox 4.0 which is possible there since extensions can be installed / uninstalled without restarting. We don't have tests for the clicking of the nsIAlertsService notification because it is native on everything except Windows. The last time I looked the nsIAlertsService itself only has very minimal tests because of this. It seems to me that we periodically come back to whether or not mozmill should have redundant tests. I personally don't think we should but I won't make a fuss if those that decide these things believe we should so cc'ing a couple of people. As for testing, the existing tests should be able to provide enough info to provide guidance for doing this. http://mxr.mozilla.org/mozilla-central/source/toolkit/mozapps/update/test/chrome/

(In reply to comment #5) > It seems to me that we periodically come back to whether or not mozmill should > have redundant tests. I personally don't think we should but I won't make a > fuss if those that decide these things believe we should so cc'ing a couple of > people. Keep in mind that we need those tests for testing the updates from our own FTP and the mirror servers. The same applies to all the localized builds which are not getting tested by the existing Mochitests. Check spreadsheets like the following one which shows the immense manual work we have to do to verify that those updates are working as expected. I really want that we can limit the manual work for QA as much as possible. https://spreadsheets.google.com/ccc?key=0AmkRt0ylPb8zdFZqV3NxeXBMYV94RXo4bVNiSmJJb2c&hl=en&authkey=CL7nj6MH#gid=2 (In reply to comment #5) > We don't have tests for the clicking of the nsIAlertsService notification > because it is native on everything except Windows. The last time I looked the > nsIAlertsService itself only has very minimal tests because of this. So that would also be a problem for Mozmill, at least until bug 525060 has been fixed.

Speaking to the question of redundancy, in this case we are trying to automate a task which QA performs manually every release cycle. Automation of Advertised Update testing will allow us to sign off much sooner than normal.

(In reply to comment #6) > (In reply to comment #5) > > It seems to me that we periodically come back to whether or not mozmill should > > have redundant tests. I personally don't think we should but I won't make a > > fuss if those that decide these things believe we should so cc'ing a couple of > > people. > > Keep in mind that we need those tests for testing the updates from our own FTP > and the mirror servers. The same applies to all the localized builds which are > not getting tested by the existing Mochitests. The existing update tests don't provide this? If so, in what way do they not? (In reply to comment #7) > Speaking to the question of redundancy, in this case we are trying to automate > a task which QA performs manually every release cycle. Automation of > Advertised Update testing will allow us to sign off much sooner than normal. My concern is over redundancy and I am 100% in agreement with automating manual processes.

Mozmill is dead, WONTFIX the remaining bugs.