fix(session): merge a save into the stored session instead of overwriting it - #94
Merged
tina4stack merged 2 commits intoOct 1, 2026
Conversation
…ting it A save wrote back the whole session as this request loaded it. When requests on one session overlapped, the one that finished later put its stale copy back over the other's changes, and re-created the session if the other request had destroyed it. regenerate() moved the same stale copy to the new id. A save now re-reads the stored session and applies only the keys this request set, changed or removed since it loaded; after clear() it replaces the session. If the stored session is gone, it stays gone: nothing is written and no cookie is sent. If the store cannot be read, nothing is written and the change is kept for a retry. regenerate() re-reads the same way before moving the session, and mints no id for a session that has ended. Signed-off-by: Michael <[email protected]>
This was referenced Oct 1, 2026
Contributor
Author
|
I have read the Tina4 Contributor Licence Agreement and I agree to it. |
A read-only request on a stored session now re-writes the re-read, merged record on #save, re-stamping the backend deadline to now + TINA4_SESSION_TTL. Expiry is therefore measured from the last request that TOUCHED the session, not from its last change, so a user actively reading pages stays logged in. PHP is the reference; Ruby now matches it (Python and Node follow). The slide is scoped to @stored (an id the store actually holds / adopted), honouring ADR-0087's guard that an unadopted or absent id writes nothing. All existing concurrent-save guards keep precedence: no id -> no write; a record another request ended -> forget + end, never re-created; a failed read -> deadline untouched, dirty kept for retry; cleared -> replace. The re-written record is the merge the concurrent-save contract computes, so sliding never clobbers a concurrent change. TINA4_SESSION_TTL=0 still never expires. The read-only contract test is rewritten from "writes nothing" to "a read-only request moves the expiry forward": real FileHandler on a real tmp dir, the stored deadline read off disk before/after, mirroring PHP's SessionConcurrentRequestsTest. Signed-off-by: Andre van Zuydam <[email protected]> Co-Authored-By: Claude Opus 4.8 <[email protected]> Co-Authored-By: Tina4 <[email protected]>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
A session save wrote back the whole session as the request loaded it. When requests on one session overlap, the one that finishes later puts its stale copy back over the other's changes, and re-creates the session if the other request destroyed it.
regenerate()moved the same stale copy to the new id.Change. A save re-reads the stored session and applies only the keys this request set, changed or removed since it loaded; after
clear()it replaces the session. If the stored session is gone, it stays gone: nothing is written and no cookie is sent. If the store cannot be read, nothing is written and the change is kept for a retry.regenerate()re-reads the same way before moving the session, and returnsnilinstead of minting an id for a session that has ended.Test.
spec/session_concurrent_requests_spec.rb: two sessions on one store stand in for overlapping requests (27 examples, including the documented login flow and an SSO callback). 15 fail onv3without the change. Each part of the change, reverted on its own, turns a case red. The full suite shows no new failures againstv3on the same machine. Also run end to end over HTTP on the built-in server, with the file session backend (the database backend keeps no session in Ruby on SQLite at all, a separate defect).Parity. The same change in all four frameworks: tina4stack/tina4-python#189, tina4stack/tina4-php#260, #94, tina4stack/tina4-nodejs#108.