- If `SMTP_ADDRESS` is set, configure Action Mailer to use it, along
with additional optional SMTP-related settings.
- Otherwise, don't set config (for compatibility with engines like
fizzy-saas).
- Remove sendmail env setup, since by default there is no sendmail in
the container, and custom deployment can use whatever config the want
directly. If we end up needing this, we can bring it back via its
own env.
This addresses a DoS vulnerability where the response might be massive
leading to OOM errors, as the response is read in full in memory by
default.
To prevent this, we need to read the body in chunks, checking the
size of the chunks we've read and raising if we go over a certain limit.
I've set the limit to 100 KB because the responses to these requests
should be fairly small or even empty, and we only care about the status
code in the end.
(Caught one such uniqueness exception in the wild)
* 200 OK -> 204 No Content, default status
* No need to touch the subscription when found
* Drop superflous test
Rails memoizes `@app_env_config`, so running push subscriptions test
after join codes test will result in running with `show_exceptions:
:none`, causing RecordInvalid to raise instead of returning 422 status.
References #1996
When ActiveStorage::Record uses `connects_to` for read replica support,
it creates a separate connection pool from ApplicationRecord. This causes
`after_commit` callbacks to fire in non-deterministic order - the
Attachment's `create_variants` callback can fire before the User model's
upload callback completes, resulting in FileNotFoundError.
The fix removes replica connection configuration from ActiveStorage::Record
so it shares the same connection pool as application models, ensuring
proper callback ordering.
Also reverts test workarounds that were added to work around this issue,
since the root cause is now fixed.
See: https://github.com/rails/rails/issues/53694
As suggested by @flavorjones, and explained in:
https://blog.yossarian.net/2025/11/21/We-should-all-be-using-dependency-cooldowns
This applies:
- 7 days as default: from the article, a 7-day
cooldown would have prevented 8 out of 10 recent
supply chain attacks. It seems reasonably long
but not crazy long.
- 14 days for major versions, to give some time
to the community to find problems.
Register a middleware with Rails.error that adds identity_id and
account_id from Current attributes. Only evaluated when an error
is actually reported, avoiding the cost on successful requests.
Move thousands of info-level Sentry events to logs. Query Loki rather
than downsampling events: `event.action = "run.ruby-script",
script.name = "autotuner"`
I find it helpful (especially when going through a long list of
results) to know how far back in time I'm looking.
A note on the timestamp being used: it's the search index
timestamp. Using the card or comment `updated_at` is fraught because
we do so much touching of records (e.g., comment is touched when a
reaction is added; card is touched when a comment is added). Let's
display the search record's `created_at`, which is the ordering used
for the overall results.
Kevin McConnell