* Storage: harden reconcile for concurrent writes and fix board transfer
Reconcile now uses two-cursor approach: captures cursor before and after
the storage scan, aborting if they differ (entries added during scan).
Job retries 3x with 1-minute waits and limits concurrency to 1 per owner.
Board transfer now correctly moves storage for card description embeds
and comment embeds, not just direct attachments. Uses batched queries
to handle cards with thousands of comments efficiently.
Also fixes N+1 queries in attachment grouping via lookup maps.
* Storage: fix per-attachment reconcile and enforce no blob reuse
The storage ledger tracks per-attachment (not per-blob) as a business
abstraction for quotas. This fixes reconcile to match that model and
adds enforcement to prevent blob reuse in tracked contexts.
* Reconcile now uses joins(:blob).sum() for per-attachment counting
* New validation prevents reusing blobs across tracked attachments
* Race-safe storage_total creation with create_or_find_by
* Job efficiency: skip find_by when object already available
* Backfill script checks per-attachment, not just per-blob
Remove data-action from the sanitizer allowlist to disallow injection
of potentially malicious Stimulus actions in user-provided
content. The lightbox controller now uses imageTarget callbacks to
handle clicks on image links.
Also add the file name as a caption in the light box, and fix the
caption color for dark mode visibility.
* Configure all sources with CSP_* vars. Space separated source list.
* Fall back to `config.x.content_security_policy.*`
* Move our sources to fizzy-saas
References https://github.com/basecamp/fizzy-saas/pull/24
An append-only storage ledger replaces deadlock-prone synchronous
counter updates and drift-prone async updates.
All content storage (card images, card/comment/board description embeds)
is tracked. Account exports (which expire) and avatars are not tracked.
Storage is accounted for by Account and Board. Both consume the same
event stream independently: no bubble-up storage bumps triggering
deadlocks due to lock sequencing. Each calculates its own total from
the underlying ledger with independent cursors and materialization.
* Storage::Entry: Append-only ledger recording attach/detach/transfer
events with delta bytes. Single event stream indexed for both
Account and Board cursor queries.
* Storage::Total: Polymorphic snapshot cache with cursor (last_entry_id)
tracking which entries have been materialized.
* Storage::Totaled concern: Provides bytes_used (fast snapshot) and
bytes_used_exact (snapshot + pending) query modes, plus
materialize_storage! to roll up pending entries.
* Storage::Tracked concern: For models owning attachments (Card,
Comment, Board). Provides board_for_storage_tracking for models
where board is determined differently (Board returns self).
Handles board transfers by recording transfer_out/transfer_in entries.
* Storage::AttachmentTracking: Hooks ActiveStorage::Attachment lifecycle
to record attach/detach entries. Handles ActionText::RichText embeds
by traversing to the actual model. Snapshots context in before_destroy
to handle cascading deletes where parent record may be gone by
after_destroy_commit.
* MaterializeJob: Rolls up pending entries into snapshot. Concurrency
limited per owner to prevent duplicate work.
* ReconcileJob: On-demand reconciliation against actual attachment
storage for support/debugging. Compares ledger total to real bytes
from card images, card embeds, comment embeds, and board embeds.
Usage:
* account.bytes_used / board.bytes_used: fast, slightly stale bytesize
* account.bytes_used_exact / board.bytes_used_exact: real-time bytesize
* Storage::Entry: audit trail for debugging and point-in-time queries
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
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"`
External sources:
- challenges.cloudflare.com for Turnstile (script-src, frame-src)
- storage.basecamp.com for Active Storage (connect-src)
User tools: loosen style/img/font/media/worker-src to avoid fighting
accessibility extensions, privacy tools, and custom fonts.
Loosen font-src and add media-src to allow browser extensions like
accessibility fonts (OpenDyslexic) and user style managers to work
without CSP violations.
Add default-src and connect-src directives to close security gaps where
unspecified directives were implicitly allowing all sources.
Default to a very narrow policy since there are no CDNs or third-party
resources to contend with.
Configurable via:
- config.x.content_security_policy.* for fizzy-saas gem overrides
- DISABLE_CSP to skip entirely
- CSP_REPORT_ONLY to enable report-only mode
- CSP_REPORT_URI for violation reporting
Process variants synchronously on attachment to close the window between
image upload and variant availability, guaranteeing that we won't have
lazy variant processing attempts in GET requests.
Tradeoff is that we do variant processing in upload requests, which is
actually desirable. We're working with images that should take
milliseconds to resize given that we'll already have the file on hand.
References https://github.com/rails/rails/pull/51951
Add SSRF protection for web push endpoints:
- Resolve endpoint IP once and pin it for connection
- Validate endpoints resolve to public IPs
- Whitelist permitted push service hosts
Add missing IP ranges to SsrfProtection:
- 100.64.0.0/10 (Carrier-grade NAT, RFC6598)
- 198.18.0.0/15 (Benchmark testing, RFC2544)
Note: link-local (169.254.0.0/16) is already covered by ip.link_local?
Patch load_schmema! to set the default value for UUID primary keys. This
removes the need to patch ApplicationRecord + Rails models individually.
It also means we no longer need to patch the default in for the integer
primary key in Search::Record::SQLite.
In the cable schema we use:
```
t.binary "payload", size: :long, null: false
```
This will fail when resetting the db otherwise with an error due to the unexpected size property.
* main: (116 commits)
Ensure avatar thumbnails are square
Update useragent to recognize twitterbot/facebot
Add defensive styles for non-square avatar images
Update test for copy changes
Missed commit
AI: standardize on https://agents.md
Make it clear this is just notifications, not comprehensive activity
AI: configure MCP servers for Chrome, Grafana, and Sentry (#1727)
Allow requests from Google Image Proxy
Update to basecamp's useragent fork
Clean up a little bit the CSRF reporting code
Claude: production observability guidance (#1725)
Prevent autoscroll to the root columns container to prevent jump on page load
Include full name string so you can type your name to filter
Prioritize current user and assigned users in assignment dropdown
Check and report on Sec-Fetch-Site header for forgery protection
bundle update
Bump bootsnap from 1.18.6 to 1.19.0
Bump rails from `077c3ad` to `17f6e00`
Fix cards getting stuck in edit mode
...
Patch the sqlite adapter to add CHECK constraints for string and text
column limits. We'll do them inline, so that any column changes
automatically update the constraints.
Instead of writer pinning, we'll track the last transaction ID of each
write in the session. Then on each read we'll wait for the replica to
report that this transaction is available.
If it doesn't become available within a reasonable timeout, we'll
proceed anyway, and accept the possibility of a stale read.
The hope here is that most of the time, the replica is caught up in the
time between a write request and the following read request. If it's
not, we now have a little tolerance to wait for it, which hopefully
proves enough to stale reads are not encountered in normal use.
We also disable the writer affinity opt-out mechanism that we had
before, since we will no longer be using writer affinity at the load
balancer.
Jeremy Daer