From MozillaWiki
Jump to: navigation, search


  • Ola: Probably a CSP bug related to manifest.
  • Mike: We'll check.

Needs Decision (miketaylr)

Issue 1669 - Use TravisCI only for cross-browser func testing (rather than BrowserStack, etc) (brizental) '

  • Bea: Short intro in issue. Browser stack takes about 20 minutes to test and it is not reliable. Saucelabs is even slower. With Travis we have the oportunity to test Chrome and Firefox. Installing a VM is not possible.
  • Mike: Problems we want to solve: 1. Test run in more than FF, 2. Test run for contributors who are not member of the webcompat org and don't develop on a fork. (No access to secrets in Travis / Browser stack).
  • Bea: Biggest problem that just core devs can run the tests.
  • Mike: Bea had a great idea, so since Travis has the headless chrome browser plugin, we could test there. But Edge might be also a good thing to test. What do you think?
  • Tom: More browsers are always a good thing to test, but when it is an improvement to everyone... Can we use both?
  • carol: agreed with Tom it would be useful to use BrowserStack for IE testing
  • Mike: Interesting idea!
  • Bea: We could set up a cron job that would test everything like every 2 weeks in e.g. Browserstack?
  • Mike: Asks the internet if this is possible and it agrees.
  • Ola: I think it's a great idea. The only request I would have, would be that we run browserstack tests before we deploy.
  • Mike: Action item would be to enable Chrome in Travis, weekly/monthly cron job and fire up a browserstack test before deploy. ## Roundtable, current work
  • Bea: I've been working on BrowserStack work, both local and on Travis. Also been testing with SauceLabs. I'm also interested in working on the issue related to tests not working on forks. What I'd like to do is remove all secure vars from Travis so things work everywhere.
  • Mike: One issue I'm working on #1564 / about CSP. CSP is an HTTP security mechanism, so the site does not hacked. There are 2 things in there, which we use right now, which I do not like. Initial commit: Flash message - just for one request, that comes from the flask application. There is no great way to get text from python to JS. Best practice is to dump inline.script in the template. In this current commit I'm dumping the text in data-attr on the tag and read that.
  • Ola: I've fixed a few bugs in the last week. I also did some ascii art in the webcompat-cli tool. Architecture is done, art is done, just need to work on the content. And created a new repo for that today.
  • Carol:
  • Mike: Guillame, Karl, Adam are on vacation. Ola is going tomorrow. Tom is on vacation next week. Mike sometime in August.

Check PR's (miketaylr)

Going through Pull Requests

Automate starting up server in test mode (bea)

  • mike: Here's what I have... probably not useful. ~/.functions testw () { wd="/Users/miket/dev/compat/" osascript -e "tell application \"Terminal\" to do script \"cd $wd; . env/bin/activate && python -t\"" > /dev/null osascript -e "tell application \"Terminal\" to do script \"cd $wd; sleep 2; intern\"" > /dev/null }
  • bea: can we paste a warning from running tests if they're not in test mode?
  • carol: if you want to test creating a new issue manually, you need to run the server in non test mode; also for me it is no longer difficult to remember running the server in test mode but it should be clearly documented when you need test mode/not

Broken Voices of the Web

Web Compatibility Progress