https://pantsbuild.org/ logo
#development
Title
# development
h

hundreds-father-404

01/23/2020, 9:53 PM
Should we have
--test-fail-fast
in V2? We have
--test-pytest-fail-fast
for example. This would be a top-level goal, not per-language
w

witty-crayon-22786

01/23/2020, 10:17 PM
probably.
to the "top-level" part: definitely top-level if that's straightforward. i'm not sure that it is =/
h

hundreds-father-404

01/23/2020, 10:20 PM
I think it is! Pretty sure the logic would live in the
test_coordinator
. Notably, the language backends only know how to work on one target at a time, so the logic couldn’t live in the language backends I don’t think we have a way to “cancel” `await Get`s though..
w

witty-crayon-22786

01/23/2020, 10:31 PM
more critically, you don't have a way to return early.
👍 1
the coordinator will block until they have all completed.
so the test runners would need to know to raise an exception to fail fast.
👍 1
or... *mumble mumble "changes to MultiGet to add something like https://docs.rs/futures-preview/0.3.0-alpha.19/futures/macro.select.html". but much larger change.
h

hundreds-father-404

01/23/2020, 10:36 PM
Hm, okay not an easy win, then. CC @hundreds-breakfast-49010. I think Toolchain may want this feature in the medium term
w

witty-crayon-22786

01/23/2020, 10:36 PM
but this feels related to the coverage thing
ie: we agree it should be top-level: we just need to decide how that information is pushed down to runners
👍 1