Skip to content

(DO NOT MERGE) Identity integration branch - #1608

Open
aram356 wants to merge 434 commits into
mainfrom
rc/identity
Open

aram356 wants to merge 434 commits into
mainfrom
rc/identity

Conversation

@aram356

@aram356 aram356 commented Nov 14, 2026 •

Copy link
Copy Markdown
Collaborator

Summary

  • rc/identity is the Trusted Server integration branch for server-side identity on every adapter (Server-side identity on every adapter #1558), the counterpart of (DO NOT MERGE) Identity integration branch edgezero#495. It carries the ready PRs the Fastly and Axum MVP builds on, joined by history-preserving merges, so identity work can start on the combined stack and CI can test it against main.
  • Each included PR keeps its own review and still lands on main by itself. This PR shows the combined diff and its CI.
  • The branch only moves forward. When an included PR's head moves, or a waiting PR becomes ready, the new head is merged on top.

Included

Merged in this order on main 1810c763a:

Order PR Head Merge commit Merge What it brings
1 #1557 76e1dd5f6 9084af2d5 Clean Sourcepoint TCF consent bridge
2 #1525 cd8977cad 7b01cab86 Clean Explicit per-provider HTTP auction endpoints
3 #1526 e529e5576 ad63e68a4 Clean Awaitable shared EC persistence (#1522)
4 #1503 5e10c24e4 aa9c03461 Clean ts ec show and ts ec eids (#1432)
5 #1506 cee4acb48 30a74def1 31 hunks Cutover to EdgeZero PR 275 (#1507)
6 none 6b32d2b61 Fix commit Lint allowance where #1503 meets #1506
7 #1506 858f95087 4bc4c4847 Clean #1506 repinned to EdgeZero rc/identity at 10166628 (see below)
8 #1506 7ee7e89a5 4a414a00b 1 hunk #1506's round-3 fixes, repinned to EdgeZero rc/identity at 3d92ee38
9 #1506 47a539528 9bd4b1415 Clean #1506 repinned to EdgeZero rc/identity at f84ffbbb
10 #1614 9552a89e0 ad1f77576 Clean Identify and batch sync answered locally on every adapter (#1582), ported onto #1506
10 #1525 f5d8b7f68 ad1f77576 Clean #1525 restacked on #1614, with its #1506 resolution in its own merge
10 #1526 b53c6e2d6 ad1f77576 Clean #1526 restacked on #1525, with its #1506 resolution in its own merge
11 #1487 64674f94c 7b1dba233 Clean Server-side identity provider collection with one page-load write (#1567), ported onto #1526 with #1557 merged
12 #1488 8183c3cae 7b4a43f00 Clean Vendor identity providers and their cookie, refresh and consent hooks, ported onto #1487
12 #1489 a7262c7db 7b4a43f00 Clean Bounded identity mirror, with pull sync as an after-response provider, ported onto #1488

Moved since 7b4a43f00

None.

The identity integration chain (2026-11-17)

The Trusted Server PRs now stack linearly on the cutover, each merging the one below it and resolving its own conflicts, so this branch takes the top as a plain merge. On GitHub the chain is stack 1621: #1506, #1614, #1525, #1526, #1487, #1488, #1489.

PR Merge in the PR Resolution
#1614 9627ca12a, then 9552a89e0 T1 ported onto #1506: local 501 and 405 answers on #1506's Fastly per-request services and Spin's request API; route tests on #1506's scripted outbound client in place of CapturingOrigin, which implemented the removed platform HTTP client
#1525 45cdb1ee3 The endpoint union and the dropped dormant Cloudflare test from 30a74def1; #1506's test ProviderConfig literals gain allow_insecure_http: false
#1526 f87078412 30a74def1 and 4a414a00b: files only #1506 and #1526 touch match this branch exactly; #1614's identify handler awaits the async handle_identify
#1487 6c0bdae31, 64674f94c Merges #1526 and #1557. #1487's KV and finalize code moves onto #1526's async store, its provider lookups onto EdgeZero's outbound client with start_batch_until, and its Fastly page lookups are collected inside the async finalize under #1506's single block_on. A lookup response first observed after the cutoff is dropped, as the cutover spec records for bids. Details in #1487's comment
#1488 8183c3cae Merges #1487. This PR's cookie reads, GDPR exclusion and opt-out memory run inside #1487's batch dispatch; its after-response provider cookies reach the lookups through post-delivery work
#1489 a7262c7db Merges #1488. Pull sync runs as #1489's after-response provider through #1487's lookup batch on the post-delivery client, under the one post-delivery block_on; the cutover's pull_sync.rs and its 15 s grouped sender stay deleted with #1489's design

GitHub does not let a stacked PR change its base, so stack 1542 (#1525 to #1528) was dissolved before #1525 moved onto #1614, and stack 1563 (#1487 to #1489) before #1487 moved onto #1526; #1527 and #1528 stay open on their old bases. ad1f77576 merges #1526's head, and its tree differs from #1506 plus this branch's earlier resolutions only by #1614.

How #1506's conflicts were resolved

30a74def1 resolves #1506's conflicts with #1525 and #1526 on #1506's structure. git show --remerge-diff 30a74def1 shows exactly what the resolution changed.

Area #1525 / #1526 side #1506 side Resolution
Fastly EC finalize (adapter-fastly/src/main.rs) Awaits EC persistence, bridged with block_on inside edgezero_main Deletes edgezero_main; finalize runs in finalize_response The block_on bridge moves into finalize_response. EdgeZero calls that hook after its dispatch executor has returned, so nothing nests
Post-send pull sync (main.rs) Async pull-sync helpers in main.rs Moves post-send work to post_delivery.rs, which already awaits it #1506's version kept; no production change needed
Pull sync and batch sync (ec/pull_sync.rs, ec/batch_sync.rs) Awaits on snapshot and upsert calls Sends through EdgeZero's send_all_until and a new request builder #1506's send path with #1526's awaits; KV-touching tests become async
Endpoint rule (auction/plan.rs) https, or http with allow_insecure_http https, or http to loopback Union: https, loopback http, or any http with the opt-in
Code #1506 deleted Edits to it Deleted Stays deleted; the edits moved into #1506's replacements

4a414a00b resolves the one conflict of #1506's round-3 head, in ec/pull_sync.rs. #1506 added a sync test, pull_sync_group_failure_names_every_partner_of_the_group, where #1526 had made the pull-sync tests async. The new test moves to #1526's async KV API (a tokio test, with seed_present_snapshot and graph.get awaited), and the existing test keeps this branch's async header.

6b32d2b61 is the one non-merge commit. #1503's ec_admin CLI test reads a body with Body::into_bytes, which #1506's clippy.toml disallows; neither PR fails alone. It allows the lint for that test crate the way #1506's own test crates do.

The EdgeZero repin

On 2026-11-17 #1506 moved to EdgeZero rc/identity at f84ffbbb. It carries PR 275's head ea2549fb, 16 fix and docs commits inside EdgeZero, and #422's stock-celld store service. No Trusted Server code changes; this branch takes it as the plain merge 9bd4b1415.

On 2026-11-16 #1506 moved to EdgeZero rc/identity at 3d92ee38, which carries PR 275's head b65e7144. PR 275's new commits answer three Trusted Server asks:

#1506's Fastly logging_for now uses that macro (#1611). This branch takes the repin as part of merge 4a414a00b. The earlier repin to 10166628 follows.

The repin lives in #1506 itself, not in this branch, so this branch takes it as a plain merge. EdgeZero's rc/identity adds the PRs stacked on PR 275 to the pin. Three of them reach Trusted Server:

EdgeZero change Trusted Server change in #1506 Commit
IABTechLab/edgezero#409 removes Forwarded and X-Forwarded-* fields on Axum before routing The Axum router records its authenticated forwarder decision in EdgeZero's received-head preparer (IABTechLab/edgezero#499), which runs before that removal. Code and tests come from 3da80afd7 in #1605 32526960d
IABTechLab/edgezero#460 adds cache tags to OutboundCachePolicy, so it is no longer Copy Test code clones it, and ExpectedTransport drops Copy 32526960d
EdgeZero's workspace pins getrandom exactly, as one of its audited exact pins The lockfile moves from getrandom 0.4.3 to 0.4.2 32526960d
None The spec's D5 now pins an rc/identity commit rather than PR 275's head 858f95087

Without the preparer, two ts dev proxy tests fail on Axum. One shows the publisher rewriting links to the loopback authority instead of the browser's. The other shows trace enable returning 403 when the proxy does not rewrite Host. With the preparer both pass, and removing it makes the new preparer test fail.

Open points for the owning PRs

Point Owning PR
The endpoint error text and its docs describe only the loopback rule, not the allow_insecure_http opt-in #1525, which now lands after #1506
No test sends to an opted-in non-loopback http endpoint through EdgeZero's client #1525, which now lands after #1506
docs/superpowers/specs/2026-11-10-shared-async-ec-design.md says EcFinalizeState carries services, which #1506 removed #1526
ec_admin.rs needs the disallowed_methods allowance from 6b32d2b61 #1503 or #1506, whichever lands second

Waiting on

None. Every in-review PR under #1558 with a non-draft PR is included.

Not included

PR What it is Why not here
#1566 Server-side identity collection on the Axum dev server Stacked on #1488; its Axum part is rebuilt on the shared identity app
#1531 Native prepared startup Rebases onto #1506 first (plan item A1)
#1532 to #1540 Native Axum baseline stack Stacks on #1531
#1605 Follow-up above an older #1506 checkpoint Closed. Its forwarder change lives in #1506 (32526960d)
#1449, #1543 Reproducible Fastly builds and release publishing Release tooling; the identity MVP does not wait on it

EdgeZero pin

Pinned by EdgeZero revision Next
#1506 rc/identity at f84ffbbb (IABTechLab/edgezero#495), which carries PR 275 at ea2549fb Moves with EdgeZero's rc/identity; an EdgeZero release tag once PR 275 merges

Refs #1558. Planned in #1559.

Test plan

At 7b4a43f00, run locally on macOS:

Check Result
cargo fmt --all -- --check, all eight clippy aliases and the core build-digest clippy Pass
test-fastly (4,063), test-fastly-reuse (243), test-axum (64), test-cloudflare (66), test-spin (89), parity (33), core build-digest test (10) Pass
Host CLI tests Pass: 1,000 on rerun. The first run failed collects_lazy_gpt_slot_only_when_scroll_is_enabled, a real-Chrome audit test that timed out waiting for the browser to exit; it passed 3 of 3 alone, and the merge changes nothing under commands/audit
scripts/lint-wasm-clock.sh, docs and Markdown format, URL-host lint Pass
#1488 at 8183c3cae and #1489 at a7262c7db: the same gates (4,076 and 4,063 Fastly tests, 972 CLI) Pass
Every in-review PR head is an ancestor of 7b4a43f00 Pass
CI on this PR Running at 7b4a43f00

Earlier, at 7b1dba233, run locally on macOS:

Check Result
cargo fmt --all -- --check, all eight clippy aliases and the core build-digest clippy Pass
test-fastly (3,929), test-fastly-reuse (242), test-axum (64), test-cloudflare (66), test-spin (89), parity (33), host CLI tests (993), core build-digest test (10) Pass
scripts/lint-wasm-clock.sh, docs and Markdown format, URL-host lint Pass
#1487 at 64674f94c: the same gates Pass
CI on this PR 31 of 31 pass at 7b1dba233

Earlier, at ad1f77576, run locally on macOS (the gates ran on the same merge before #1614's three-line host fix, which then passed the URL-host lint and its four Fastly tests):

Check Result
cargo fmt --all -- --check, all eight clippy aliases and the core build-digest clippy Pass
test-fastly (3,734), test-fastly-reuse (226), test-axum (63), test-cloudflare (65), test-spin (88), parity (33), host CLI tests (993), core build-digest test (10) Pass
scripts/lint-wasm-clock.sh, docs and Markdown format, URL-host lint Pass
#1614, #1525 and #1526 each at their new head: fmt, all clippy aliases, scripts/lint-wasm-clock.sh, the four adapter suites, test-fastly-reuse and parity Pass
CI on this PR Superseded by 7b1dba233

Earlier, at 9bd4b1415, run locally on macOS:

Check Result
cargo fmt --all -- --check, all eight clippy aliases and the core build-digest clippy Pass
test-fastly (3,726), test-fastly-reuse (224), test-axum (60), test-cloudflare (62), test-spin (85), parity (33), host CLI tests (993) Pass
Core build-digest test, scripts/lint-wasm-clock.sh, JS tests (3,310) and format, docs and Markdown format, docs-proposal scripts, Node pin guard, URL-host lint Pass
CI on this PR Running at 9bd4b1415

Earlier, at 4a414a00b, run locally on macOS:

Check Result
cargo fmt --all -- --check Pass
All eight clippy aliases and the core build-digest clippy Pass
test-fastly (3,726), test-fastly-reuse (224), test-cloudflare (62), test-spin (85), parity (33), host CLI tests (993) Pass
test-axum (60) 59 passed, 1 failed in the gate run: publisher_input_cap_matrix_axum read an incomplete body (IncompleteBody). The test then passed 3 of 3 alone and the alias passed 5 of 5 full runs. #1506's head 7ee7e89a5 also passed 5 of 5. The cause is not reproduced
Core build-digest test, scripts/lint-wasm-clock.sh, JS tests (3,310) and format, docs and Markdown format, docs-proposal scripts, Node pin guard, URL-host lint Pass
CI on this PR Running at 4a414a00b

Earlier, at 4bc4c4847, run locally on macOS:

Check Result
Every head in Included is an ancestor of this branch through history-preserving merges Pass
cargo fmt --all -- --check Pass
All eight clippy aliases and the core build-digest clippy Pass
test-fastly (3,723), test-fastly-reuse (223), test-axum (60), test-cloudflare (62), test-spin (85), parity (33), host CLI tests (981) Pass
Core build-digest test, scripts/lint-wasm-clock.sh, JS tests (3,310) and format, docs and Markdown format, docs-proposal scripts, Node pin guard, URL-host lint Pass
The same 27 gates on #1506's own head 858f95087 Pass
CI on this PR 31 of 31 pass at 4bc4c4847

dhruv8sh and others added 30 commits November 12, 2026 19:41
The coordinator now reports only a 404 as no match before parsing. A 204 reaches the provider, so LiveRamp's 204 opt-out clears its sources and is remembered, while ID5 and UID2 map their own 204 to no match.

Signed-off-by: dhruv8sh <dhruv8sh@proton.me>
LookupInput carries the coordinator's request time and RequestCookies carries it to cookie readers, so the UID2 and ConnectID expiry checks and ID5's ts field use the navigation's one clock reading. Provider tests now read cookies at a fixed time.

Signed-off-by: dhruv8sh <dhruv8sh@proton.me>
Validation dropped refresh credentials whose refresh time had passed, so a live UID2 identity imported after its refresh_from, or kept through a failed refresh, could never be refreshed again. Due state is now stored and the source stays due; rate_limit_per_hour bounds the retries.

Signed-off-by: dhruv8sh <dhruv8sh@proton.me>
A server-side refresh updates the row but not the browser's cookies, so the next navigation's older __uid_2 token replaced the refreshed token and its refresh state. Providers can now ask the coordinator to keep a live stored UID whose absolute expiry outlives the cookie's; UID2 does. A cookie opt-out still wins, and a newer SDK token is still taken.

Signed-off-by: dhruv8sh <dhruv8sh@proton.me>
Cookie and consent-derived opt-outs no longer get the 30-day lockout: lookups now say whether they remember, keep or forget the provider's opt-out, and only an opt-out a vendor server reports is remembered. The vendor's cookies are read before a remembered opt-out is checked, so an ID in them forgets the opt-out and is stored. ID5's "0" stays an opt-out.

Signed-off-by: dhruv8sh <dhruv8sh@proton.me>
Validation already clamped absolute expiries; ProviderExpiry::After is now clamped to the same 30 days, so the bound no longer depends on each provider.

Signed-off-by: dhruv8sh <dhruv8sh@proton.me>
Generation reuses the generated UID the row still holds after it expires, so the ID stays stable for as long as the row lives rather than changing every 30 days.

Signed-off-by: dhruv8sh <dhruv8sh@proton.me>
Signed-off-by: dhruv8sh <dhruv8sh@proton.me>
The unchanged-cookie and failed-refresh regressions now run dispatch, collect and plan_write for a 404, timeout, 502, invalid reply and rate limit, each with a no-cookie control, and assert that no write is planned.

Signed-off-by: dhruv8sh <dhruv8sh@proton.me>
…ient

The fetch body now carries provider (provider_name, defaulting to the publisher domain), domain and origin like ID5's Prebid Server module, instead of o, v, tml and nbPage.

Signed-off-by: dhruv8sh <dhruv8sh@proton.me>
Prebid's Lockr AIM module does not run where GDPR applies, so a Lockr cookie on such a request is a leftover; cookie readers now ignore Lockr cookies there. The example tables use Lockr keys that differ from the vendors' own cookies, and the docs say the default Lockr keys are already read as own cookies.

Signed-off-by: dhruv8sh <dhruv8sh@proton.me>
…pping file

Signed-off-by: dhruv8sh <dhruv8sh@proton.me>
LiveIntent now applies the shared usable-ID check, so a blank or 0 ID reads as no match. A LiveRamp envelope entry without a type no longer fails the whole reply.

Signed-off-by: dhruv8sh <dhruv8sh@proton.me>
Signed-off-by: dhruv8sh <dhruv8sh@proton.me>
Signed-off-by: dhruv8sh <dhruv8sh@proton.me>
Prebid's LiveIntent module sets the resolved Trade Desk ID's ext.provider to the page domain, falling back to liveintent.com, and copies the reply's m next to segments in the nonId EID ext. Carry the publisher domain to parse_response through the request context and build both ext values the same way.

Signed-off-by: dhruv8sh <dhruv8sh@proton.me>
The coordinator drops sources a provider excludes where GDPR applies only when it writes the row, so a UID2 UID stored outside GDPR still reached bidders until the next committed write. Mark those sources in the partner registry for every configured provider and skip them in live_eids for a GDPR visitor.

Signed-off-by: dhruv8sh <dhruv8sh@proton.me>
Prebid's ID5 module sends the cached reply's signature as s so ID5 can return the same ID. Keep a non-empty reply signature as the UIDs' private refresh state, due a tenth of the TTL before they expire, and send it as s in the next fetch.

Signed-off-by: dhruv8sh <dhruv8sh@proton.me>
The GDPR source exclusion applied only to stored rows, so a UID2 EID the browser sent in the /auction body or the ts-eids cookie still reached bidders for a GDPR visitor. drop_vendor_denied_client_eids now also drops client EIDs whose source a provider excludes where GDPR applies, which covers /auction and the server-side auction.

Signed-off-by: dhruv8sh <dhruv8sh@proton.me>
… sensitive information'

Co-authored-by: Copilot Autofix powered by AI <62310815+github-advanced-security[bot]@users.noreply.github.com>
…phic value'

Co-authored-by: Copilot Autofix powered by AI <62310815+github-advanced-security[bot]@users.noreply.github.com>
The Copilot Autofix change generated the key with aes_gcm::aead::OsRng, which this aes-gcm build does not export, so the crate stopped compiling. A fixed key in test-only code is intended.

Signed-off-by: dhruv8sh <dhruv8sh@proton.me>
The encrypted_reply test helper built its AES-256-GCM key from a constant, which CodeQL flags as a hard-coded cryptographic value. Generate the key from the system RNG instead and return it alongside the body so the decrypting tests pass it as the reply context.

The getrandom feature is enabled on aes-gcm for dev builds only.

Signed-off-by: dhruv8sh <dhruv8sh@proton.me>
Every adapter rendered a failed request's report through its own copy of
http_error, which logged at error whatever the status, and
read_request_body logged the same rejection again. A 413, 408, or 400 from
a body read therefore produced one warn line and one error line.

The adapters now share trusted_server_core::error::http_error. It logs the
report once, at warn for a client error and at error otherwise, and the
four adapter copies are gone. read_request_body no longer logs: its report
names the limit or carries the read error. The auction and batch sync
routes, which answer a 413 without the helper, log the report at warn
themselves.

WarningCapture now records errors as well as warnings, with each record's
level, so a test can show that a request logs one line at the expected
level.
Each adapter read the request's outbound client, failed with the same
internal error when it was missing, and turned the proxy marker off in its
own copy of one block. Core's outbound::request_outbound_client now does
that once, and the four copies are gone.

Cloudflare and Spin each had an identical publisher_input_cap, and Axum
inlined its first half. Core's publisher::publisher_input_cap takes the
platform cap as a parameter: Cloudflare passes 10 MiB, Spin 8 MiB, and Axum,
which has no runtime limit of its own, u64::MAX.
CodeQL alert 283 (rust/cleartext-logging) flagged
RecordedRequest::assert_headers: its assert_eq printed every recorded and
expected header value, a Fastly API key among them, and its message printed
the request's method and URI.

A failure now names the missing, unexpected, and differing headers only.
assert_transport_settings also stops printing the URI and the authority it
sent, and names the request by its method. The management API test calls
its upsert request upsert_request instead of secret.
Spec test 11 says canonical 404 and 405 carry x-ts-version on every
runtime, but the Fastly leg had no test: the Viceroy TRACE canary checked
only the 405 status. It now also requires a non-empty x-ts-version and
x-ts-fastly-version, which Viceroy's FASTLY_SERVICE_VERSION supplies. CI
runs the canary in both Fastly integration jobs.

The Axum, Cloudflare, and Spin canonical tests compared the header with
TS_GIT_VERSION, so a build without a git version passed on an absent
header. They now expect the compiled-in version and fail when the build
has none. The EOF harness compared the detached responses' version headers
with the routed page's, which also passed when both were absent; it now
requires both headers on the routed page first.
On Fastly a forbidden or invalid operator response header warned twice per
request: once when the entry point built the per-request synthesized set,
which runs apply_finalize_headers on an empty response, and again when the
routed response was finalized.

Core now splits the warning from the application.
apply_response_headers_with_cache_privacy skips the entries without a log,
and warn_skipped_operator_response_headers logs one warning per skipped
entry. Fastly calls it from per_request_synthesized_headers, which the
entry point runs once for every request. Axum, Cloudflare, and Spin call it
from their finalize helper, which their middleware runs once per request.
EC pull sync and the Fastly Tinybird sender logged a failed request's
status and kind themselves, and pull sync logged a request-build error
with Debug, which includes the error's message. That contradicted the
outbound module's rule, and spec section 2.11, that every client failure
and request-construction error goes through outbound_error.

Both now build the report with outbound_error and
OutboundRequestSource::Internal, then log it at warn: pull sync with an
EdgeCookie context for a build failure, a group failure, a request
failure, or an unreadable response, and Tinybird with a Proxy context for
a build or send failure. outbound_error adds its typed detail and keeps
the provider's text out of the logs.
T1 now builds on the EdgeZero cutover, so the partner API change ports onto
#1506's adapters:

- Core batch sync keeps T1's local 501 and 405 answers next to #1506's async
  handle_batch_sync_with_writer.
- Fastly takes #1506's per-request services. The identify and batch-sync
  handlers answer locally unless the method is the primary one and an EC
  record store is configured, and both paths claim every method.
- The route tests drop the CapturingOrigin fixture, which implemented the
  platform HTTP client #1506 removed. They route through #1506's scripted
  outbound client instead and assert that no partner API request reaches
  it; a control test shows a publisher page does reach it.
#1614 now carries #1506's EdgeZero cutover, so #1525's per-provider HTTP
opt-in meets the cutover here, resolved the way rc/identity resolved it in
30a74de:

- The endpoint rule is the union: https, plain http to loopback (#1506), or
  any http with allow_insecure_http (#1525).
- Cloudflare keeps #1506's startup_loads_multi_provider_plan; the dormant
  multi-provider test is dropped because the cutover loads every plan.
- #1506's test ProviderConfig literals in the Fastly app and the auction
  orchestrator gain allow_insecure_http: false.
#1525 now carries #1506's EdgeZero cutover and #1614, so #1526's awaitable EC
persistence meets them here. The resolution is the one rc/identity made in
30a74de and 4a414a0:

- Fastly runs the async EC finalize from finalize_response, which EdgeZero
  calls after its dispatch executor returns, and drops edgezero_main.
- Pull sync and batch sync keep #1506's EdgeZero send paths with #1526's
  awaits, and their KV-touching tests become async, including #1506's
  pull-sync group failure test.
- The post-delivery test blocks on the now-async snapshot load.

#1614's identify handler awaits #1526's async handle_identify.
The partner API requests target publisher.example.com, and the fallback control test scripts the origin URL from the test settings, as the cutover's other Fastly tests do.
Takes #1614's 9552a89, which moves its ported Fastly partner API tests onto allowlisted example hosts.
Takes #1525's f5d8b7f, which brings #1614's move of its ported Fastly partner API tests onto allowlisted example hosts.
… rc/identity

#1526's head b53c6e2 now stacks on #1525 (f5d8b7f), which stacks on #1614 (9552a89), which stacks on #1506. Each PR merged the one below it and resolved its own conflicts, so this merge needs no resolution. #1525's and #1526's resolutions against #1506 are the ones this branch made in 30a74de and 4a414a0, and #1614 (T1) is new here.
Brings PR #1487 onto PR #1526's base (PR #1506's EdgeZero cutover, PR #1614,
PR #1525 and PR #1526's awaitable EC persistence) and keeps PR #1487's
behavior in the new base's structure.

How each area was resolved:

- Provider lookups: the lookup transport now sends through EdgeZero's
  outbound client in the auction fan-out's shape. `LookupTransport::start`
  takes every built request and starts them as one batch with
  `HttpClient::start_batch_until` and a cutoff of the identity budget,
  taken on the client's own clock;
  `gather` reads the batch until it finishes, maps slots back to providers
  and reports unresolved slots as timeouts. `PlatformLookupTransport` and the
  coordinator's web-time `Deadline` are replaced by `OutboundLookupTransport`;
  requests keep the 64 KiB response cap, cache bypass and strict TLS.
- Deleted modules: `platform/http.rs` and the Fastly `platform/deadline.rs`
  stay deleted. PR #1487's final-snapshot change to the Fastly deadline
  select (a 1 ms surfacing wait per pending body and counting arrived headers
  as progress) has no counterpart in EdgeZero's batch and is not ported, nor
  is its doc-only change to `select_until`.
- Async KV: `ec/kv.rs`, `ec/mod.rs` and `ec/finalize.rs` take PR #1487's
  versions converted to PR #1526's async store; the page-load and
  after-response writes, the write planner, after-response commit, batch
  sync's provider-owned check and their tests await the store. Unchanged
  tests take PR #1526's async versions, and PR #1526's non-Send yielding
  store test is kept. `generate_if_needed` stays async as on the base.
- Fastly entry point: `main.rs` takes the base's lifecycle. `finalize`
  collects the navigation's page-phase lookups inside the async EC finalize
  under its one `block_on`, using the partner registry built once in
  `AppState`. After-response lookups run in post-delivery work, after pull
  sync, on the post-delivery client and clock. Lookups still travel through
  the request-local slot, because an in-flight batch is not `Send`; the slot
  is now cleared before each request and after delivery.
- Fastly app: keeps the base's local 501/405 identify and batch-sync answers,
  publisher origin options and response-body scope owner, with PR #1487's
  partner registry, identity coordinators, navigation permit and dispatch.
- Other adapters, core call sites and tests: kept both sides; PR #1487's
  tests that used the removed platform HTTP client or a custom auction
  provider now use the scripted outbound client, and the Fastly identity
  tests route with the host geo lookup the entry point's `prepare` runs.
- Cargo: the lockfile adds only the identity-providers crate to the base's.
- Docs and CHANGELOG: auction collection text follows the base; PR #1487's
  identity entries and spec now describe the EdgeZero batch transport.
#1487's head 64674f9 stacks on #1526's b53c6e2 and merged #1557's 76e1dd5; each merge resolved its own conflicts, so this merge needs no resolution. It brings server-side identity provider collection with one page-load write (#1567), with lookups on EdgeZero's outbound client and the KV code on #1526's async store.
Brings PR #1488's vendor identity providers onto PR #1487's new head, which
carries PR #1487 onto PR #1526 (EdgeZero cutover, local partner API answers,
HTTP auction opt-in and awaitable EC persistence).

How each area was resolved:

- Coordinator: PR #1488's cookie reads, GDPR source exclusion, opt-out
  memory and merged results run inside PR #1487's batch-shaped dispatch. The
  lookups a navigation still needs are built first and start together in one
  EdgeZero batch; `PendingLookups::is_empty` counts started requests along
  with cookie, generated and early results.
- Fastly entry point: `main.rs` takes PR #1487's EdgeZero version. The
  provider cookies PR #1488 hands to after-response lookups go into the
  after-response `LookupContext` in post-delivery work; the request-side
  cookie capture merged unchanged.
- Async KV: PR #1488's new write-plan and identify tests await the store.
- Providers crate, CLI and settings: merged without conflict; providers stay
  pure and make no HTTP calls of their own.
- Cargo: the textual merge pointed `crypto-common` at a `getrandom` the
  lockfile no longer held; cargo resolved it to the locked 0.4.2. The rest
  of the lockfile delta is PR #1488's own.
Brings PR #1489's pull-sync provider and EID mirror onto PR #1488's new
head, which carries PR #1487 onto PR #1526 (EdgeZero cutover, local partner
API answers, HTTP auction opt-in and awaitable EC persistence).

How each area was resolved:

- Pull sync: core `ec/pull_sync.rs` stays deleted, so the base's EdgeZero
  group sender and its tests go with it. Pull sync runs as PR #1489's
  `pull-sync` after-response provider, sent through the post-delivery
  client in the lookup batch. Post-delivery work is now telemetry, then
  after-response lookups; the separate pull-sync step, its graph factory
  and the unused clock are removed, and PR #1489's "nothing due" check runs
  before the live row read. `require_identity_graph`, now used only by
  tests, is test-only.
- Coordinator: PR #1489's dispatch order and per-navigation pull-sync cap
  run inside PR #1487's batch-shaped dispatch; the cap counts each queued
  pull-sync request.
- Fastly app: `EcFinalizeState` drops both the old `services` and
  `is_real_browser`; the bot gate moves into after-response capture as in
  PR #1489, with the base's request services and async batch sync.
- Async KV and finalize: PR #1489's removals in `kv.rs` (snapshot partner
  upsert) and the pull-sync marker stand; the mirror and the inner finalize
  are awaited, and PR #1489's new tests await the store.
- Publisher: keeps the base's EdgeZero origin request without the marker
  preload inputs PR #1489 removed.
- Providers crate: PR #1489's pull-sync providers and PR #1488's SharedID
  cookie defaulting both apply when the phase coordinators are built.
- Docs and CHANGELOG: the base's auction wording with PR #1489's identity
  bullets; the onboarding step names after-response lookups.
…ntity

#1489's head a7262c7 stacks on #1488 (8183c3c), which stacks on #1487 (64674f9); each merged the one below it and resolved its own conflicts, so this merge needs no resolution. It completes the provider stack (#1567): the vendor identity providers and cookie hooks (#1488), and the bounded identity mirror with pull sync as an after-response provider on EdgeZero's outbound client (#1489).

This branch has not been deployed

No deployments
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants