Files
PaperClipAI/doc
DottaandPaperclip 54b83b15a3 Reuse authenticated nested workspace verification (#19854)
## Thinking Path

> - Paperclip preserves the work that its agents produce.
> - Portable workspace migration can save nested workspaces before their
parent.
> - The parent must prove that a committed child snapshot covers any
files it omits.
> - The existing owner downloads and restores the same immutable child
archive before and after parent capture.
> - The fresh local-file checks are required, but the repeated immutable
archive check is avoidable within one live source fence.
> - This pull request reuses only an authenticated archive proof with an
unchanged reference and provider identity.
> - It also bounds independent child branches at two workers while
retaining every fresh local and Git check.

## Linked Issues or Issue Description

Refs #19805 and #19688. A duplicate search found no separate fix for
repeated child archive verification.

**What existing behavior does this improve?**

Online Git parent migration after nested workspaces have committed
portable snapshots.

**Subsystem affected**

server/ — workspace migration and nested snapshot coverage.

**Current behavior**

Each parent verification pass downloads, authenticates and restores
every child archive, then captures the child's current local tree. The
second pass repeats the immutable download and restore. Independent
child branches also run in sequence. A large parent can exhaust its
existing budget before parent capture begins.

**Proposed behavior**

One private verifier belongs to the existing live source-owner
transaction. It can reuse a completed archive proof only when the full
snapshot reference, operation, child closure, ETag and byte length still
match. Missing or weak ETags and non-cancellation HEAD errors keep full
verification. Caller cancellation still stops work. A changed identity
forces fresh authenticated readback. The owner still captures current
local bytes and Git state in both passes.

**Reason and benefit**

Remove duplicate immutable work without treating an old local-file check
as current authority. Two independent topmost branches can overlap.
Recursive descendants stay serial within those branches, so the bound
does not grow with depth.

## What Changed

- Add a private per-fence verifier to the existing tree owner. It clears
its proof map at the end of the import and refuses a different owner
transaction.
- Reuse only completed ciphertext, GCM, plaintext, and canonical restore
verification with a matching fresh provider identity.
- Retain all current database, operation, recursive coverage,
local-file, Git and manifest checks on each pass.
- Bound independent branches at two. Abort queued work on failure and
await active peers before scratch cleanup or fence release.
- Use the verifier only in the existing online Git importer. Keep the
offline full-verification path and all capture deadlines unchanged.
- Document resource admission, fallback behavior and the hosted
acceptance limit.

## Verification

- Genuine baseline RED through the real parent importer: child GET
counts 2 + 2, canonical restores 4, maximum overlapping readbacks 1. The
four fresh local child captures already pass.
- Exact positive GREEN: child GET counts 1 + 1, canonical restores 2,
maximum overlapping readbacks 2. All four fresh local child captures
remain.
- The fixture uses real encrypted archives and the canonical S3
provider's local HTTP HEAD parsing with actual payload-derived ETags. It
uses fake fixture credentials only.
- Final focused suite: **13/13 PASS**, 148.36 seconds, at `eaf17d3340`.
This includes recursive max-two completion, failure/cancellation
joining, identity/ref/content changes, same-fence alias reuse and
stale-owner refusal.
- Both affected existing migration suites: **256/256 PASS**, 201.84
seconds, at initial head `eaf17d3340`.
- The HEAD-failure review repair has genuine two-case RED evidence and
three passing failure/cancellation controls. Full layered
source/test-tree typecheck and module boundaries pass at repaired head
`763648adad`. Final exact-head controls: **21/21 PASS**, 86.63 seconds
(all 16 new controls plus five existing parent cases).
- The first extended recursive fixture correctly failed because it
created a Git branch after taking an earlier snapshot. The fixture now
creates all branches before any import; the source-change guard remains
intact.
- Exact-head CI: 56 success, four expected skips, and one neutral
Paperclip Review. Required gates pass; Greptile is fresh 5/5. Its
optional aggregate-handler cleanup recommendation is resolved with the
intentional fail-closed rationale.
- Local verification is macOS arm64 / Node26.4.0. The observed hosted
parent still needs to complete within its unchanged budget after normal
delivery. No hosted timing improvement is claimed here.

## Risks

- Two simultaneous child verifications can use more I/O, memory and
scratch space. Operators must count logical verifications, even in one
process.
- HEAD availability and provider identity are required for proof reuse.
Providers without a usable ETag perform the full check.
- Reuse lasts only for one live transaction. A retry, changed reference,
changed provider identity, process exit or owner change requires fresh
verification.
- Faster verification does not waive changed child bytes or guarantee a
specific hosted completion time. No provider turn, authority marker,
source deletion or larger timeout is introduced.

## Model Used

OpenAI GPT-6 through Codex, with repository and code-execution tools.
The exact serving model revision and context-window size are not exposed
to this agent.

## Checklist

- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` / `Closes
#` / `Refs #` OR (b) described the issue in-PR following the relevant
issue template
- [x] I have not referenced internal/instance-local Paperclip issues or
links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip`
URLs)
- [x] My branch name describes the change (e.g. `docs/...`, `fix/...`)
and contains no internal Paperclip ticket id or instance-derived details
- [x] I have run tests locally and they pass
- [x] I have added or updated tests where applicable
- [x] I have updated relevant documentation to reflect my changes
- [x] I have considered and documented any risks above
- [x] All Paperclip CI gates are green
- [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [x] I will address all Greptile and reviewer comments before
requesting merge

---------

Co-authored-by: Paperclip <noreply@paperclip.ing>
2026-11-07 10:45:17 -06:00
..
…
…
…
…
…