anthony sottile

sounds good -- will follow #2216! (fwiw the --canary also failed in the same way)

this should probably happen anyway -- right now that file descriptor is torn down at an undefined time

> @asottile-sentry (CC @bukzor) > > Would the overhead of making this a new CI job be acceptable if it runs only on PRs that change code in the directory...

> > running only on HC changes would be "fine" but I think that's impossible to detect due to how tangled our codebase is > > I believe > >...

please stop commenting +1, many of us are subscribed to receive updates on this issue - if you want to see it happen, click the +1 react on the original...

> need to address the ci fails in [getsentry/pypi#663](https://github.com/getsentry/pypi/pull/663) here's your root cause: https://github.com/getsentry/snuba-sdk/commit/b7a64da26d17987fd233351a4d37a883fd7fdd9d#r138544221

> > here's your root cause: [getsentry/snuba-sdk@b7a64da#r138544221](https://github.com/getsentry/snuba-sdk/commit/b7a64da26d17987fd233351a4d37a883fd7fdd9d#r138544221) > > @asottile-sentry @joshuarli Why did this start breaking just for these last few builds? Parsimonious has been in the Snuba SDK for...

what are we trying to solve here. this feels like extra gatekeeping. we should trust developers to do the right thing and if they don't that's a people management problem

my 2c is we should just ban custom sql

partially reverting #1672 fixes this so I suspect the removal of `objects` there isn't correct CC @flaeppe ```diff --- a/mypy_django_plugin/transformers/models.py +++ b/mypy_django_plugin/transformers/models.py @@ -833,9 +833,9 @@ class MetaclassAdjustments(ModelClassInitializer): ): del...