Our team used to utilize the try syntax: `try: -b do -p android-lint,android-x86,android-test,android-api-15 -u robocop -t none` (Ref: https://treeherder.mozilla.org/#/jobs?repo=try&revision=a5695c49f80b544d2b4597dae776521f583f14ad ) to cover most of our tests. (Note: The android-lint,android-test part is actually not necessary in old implementation) With Bug 1345863 this is changed. Some workarounds are removed so based on this context the best bet we could go for is: `try: -b do -p android-x86,android-api-15 -j android-lint,android-test -u robocop -t none` (Ref: https://treeherder.mozilla.org/#/jobs?repo=try&revision=949d6bf392cf89f101406e591e3ae4e7e962df8e ) This does not work however. After a short discussion with :dustin, seems like the -p and -j syntax don't go along well with current implementation See: `taskcluster/taskgraph/try_option_syntax.py`: 554 def task_matches(self, attributes): 555 attr = attributes.get ... 587 job_try_name = attr('job_try_name') 588 if job_try_name: 589 if self.jobs is None or job_try_name in self.jobs: 590 if self.platforms is None or attr('build_platform') in self.platforms: 591 return True Since the android-lint job is not platform specific, we will not be able to run it if we specified specific platforms for this try.
My thought was to ignore `-p` for tasks that have a job_try_name attribute but no build_platform attribute. Greg, does that seem reasonable?
Bug 1368733 does exactly what I'm suggesting, so I'll dupe this one over there.