[AC-144] Initial implementation of impression capping in MAC - #7465
Draft
jonesetc wants to merge 59 commits into
Draft
[AC-144] Initial implementation of impression capping in MAC#7465jonesetc wants to merge 59 commits into
jonesetc wants to merge 59 commits into
Conversation
jonesetc
force-pushed
the
ac-144-impression-capping
branch
from
July 7, 2026 18:01
5b803ae to
39a27bb
Compare
jonesetc
force-pushed
the
ac-144-impression-capping
branch
from
July 7, 2026 18:09
39a27bb to
afa38e4
Compare
I'm working on a new UniFFI parser (mozilla/uniffi-rs#2841) and it currently requires that types used in the exported functions are publicly available from other crates. I could maybe rework the parser to handle this another way, but this feels cleaner to me anyways. It feels weird if a type can be used by foreign languages but not other Rust crates.
…ro. (mozilla#7464) Seeing mozilla#7439 fly past reminded me that I was intending to try and remove the BridgedEngine entirely - this doesn't quite achieve that, but does get closer.
… we've merged all keys (mozilla#7458)
Moved the code to RetryingAccount, which should give us this functionality for free. Also, tweaked the fxa-client example a bit.
Current behaviour is to update an existing record with the same id.
Before it was named `httpconfig` for historical reasons, this one seems more natural.
…ozilla#7469) Nowadays we're using the Rust error ping which is handled internally in the Kotlin module. We no longer need to forward these events for recording in sentry. The android side of this was removed in https://bugzilla.mozilla.org/show_bug.cgi?id=1991443.
I believe this should fix our `cargo audit` CI failures.
to avoid leaks from foreign providers of the encryptor decryptor trait.
…#7476)" (mozilla#7479) This reverts commit 7304877.
* doc: Adding details to out of date documentation for building application services * fix: review and glean_sym step added * Update docs/building.md Co-authored-by: bendk <bdk@mozilla.com> --------- Co-authored-by: bendk <bdk@mozilla.com>
The `nimbsus_events.unenrollment` metric is supposed to contain the reason field, but it just hasn't been reported. Additionally, this removes "is_rollout" field from the `nimbus_events.enrollment` metric as (a) that data is not already reported, (b) the data is not already available on the `EnrollmentChangeEvent` (and adding it would be a breaking change), and (c) the field is not all that useful.
…pes (mozilla#7475) * RMST-464: remote-settings: Skip verification of unknown signatures types * Update changelog * Add missing field in tests
…#7474) Split up `RemoteSettingServiceInner` into multiple structs, each with their own Mutex and add a bit of policy around locking them. Right now this is just enforced by putting all the methods that lock them in one impl block. The issue we're trying to avoid is having `RemoteSettingService::sync` block `RemoteSettingService::make_client`. Also, start using `viaduct::Client` which should be possible now since all applications are using the new backend.
…la#7481) When queried for the list of all available Firefox Labs, the Nimbus Client was not filtering the recipes by whether they passed targeting and bucketing. This is now the case. There is now a new function `can_enroll` that returns an enum, `CanEnrollResult`, that distinguishes all the different cases that prevent enrollment. The majority of this function was refactored out of `evaluate_enrollment`, which is now a much simpler function. Additionally, the `targeting(expr, helper)` function has been removed. This function took an expression and a `NimbusTargetingHelper` and returned an `EnrollmentStatus` if and only if the targeting evaluation did not succeed or resulted in an error. This API was too awkward to work into `can_enroll`. The majority of this change is test fallout from removing this function, though the tests make more sense now that they are testing the results of JEXL evaluation and not comparing `EnrollmentStatus`es. In order to remove a bunch of unnecessary clones, the `NimbusTargetingHelper` methods now take a `&str` instead of a `String`. The UDL has been updated to use `[ByRef]` so that the FFI contract is unchanged.
You will probably never do that in a "real" browser window, so make it a little clearer about the expected use.
For the run_maintenance_optimize background task, we add the 0x10000 flag in addition to the default 0x12. This causes optimization of all tables even if they were never queried by the provided connection. This fixes a rare scenario where Firefox for Android can get stuck with incorrect table stats on the moz_origins table, leading to poor address bar autocomplete performance. For the Drop implementation on PlacesDb, we add the now-default flag 0x10, which previously didn't exist, and was being explicitly overridden in the current implementation. This adds a row limit on analyze which can help to prevent poor shutdown performance, matching the change made in the fix for bug 2017227. Co-authored-by: Mark Hammond <mhammond@skippinet.com.au>
…on (mozilla#7490) The connection is otherwise only closed when the Store is dropped, which on Firefox Desktop happens during GC at shutdown -- past the late-write barrier, so the flush crashes debug builds. Wrap the connection in Mutex<Option<AutofillDb>> behind a fallible lock_db() helper (mirroring LoginStore); shutdown() closes it early, and later operations return DatabaseClosed.
Most other components are on proc-macros and most UniFFI development is happening with proc-macros. I checked the generated code and the only differences I could see is were in the comment and I liked the newer versions better. However, it's likely I'm missing something and there will be some unintinded breaking changes. We just started the nightly cycle, so this seems like a good time to switch over.
…ozilla#7492) * RMST-472 - Make remote-settings v2 routes the default going forward * upating changelog
…zilla#7508) * Only build `glean-sym` on Android and iOS * Use android-components publish.gradle for app-services components
… and `libsForTests` publication (mozilla#7509) * Only build `glean-sym` on Android and iOS * Use android-components publish.gradle for app-services components * Package `libmegazord.so` into full-megazord AAR with NSS dependencies and `libsForTests` publication
…s components (mozilla#7510) * Only build `glean-sym` on Android and iOS * Use android-components publish.gradle for app-services components * Package `libmegazord.so` into full-megazord AAR with NSS dependencies and `libsForTests` publication * Remove unused `error_support` macro re-exports in application-services components
…ds (mozilla#7511) * Only build `glean-sym` on Android and iOS * Use android-components publish.gradle for app-services components * Package `libmegazord.so` into full-megazord AAR with NSS dependencies and `libsForTests` publication * Remove unused `error_support` macro re-exports in application-services components * Skip the `kotlin-android` plugin under AGP 9 for mozilla-central builds
…config check for nested Gradle (mozilla#7506)
… append v2 (mozilla#7517) * Bug 2058816 - updating get_base_url and get_url in remote_settings config to append v2 instead of v1 for custom url's * Updating remote_settings unit tests * updating search and nimbus-cli refs
PR mozilla#7508 hardcoded `"org.mozilla.appservices"`, dropping the `.nightly` suffix on nightly builds. To fix, we add `appServicesGroupId` to resolve the proper `groupId` based on the context (`mozilla-central` or standalone `application-services`). `configurePublish`'s parameters were also being shadowed in `publish.gradle`, which this resolves too.
…mozilla#7518) * feat(remote-settings): RMST-484: verify signatures with sync_if_empty * Update changelog
…7487) * test: Adds new python smoke tests for building against iOS; fenix * test: Adds scheme, tests * fix: Modularization and missing xcframework change * fix: linting * fix: Lints, docs * fix: Adds deprecation notes * fix: Adds regex for fenix + docs * fix: Clarifies a todo * fix: missed err_msg change * feat: Adds HNT tests, fixes some review issues * fix: Some missing docs * fix: Small edits * fix: adds hnt test * fix: Build against all * fix: Small log correction * fix: some doc changes * fix: some readme linting * fix clarifies HNT acronym
…ozilla#7523) * fix(remote-settings): Do not quote timestamps with v2 API * Update CHANGELOG
For some reason the JSON stopped working when imported. Adding an extra `datasource` field fixes this.
…lla#7513) Firefox Desktop is moving its address storage to this component and needs two things that are not currently exposed. Storage APIs for importing records already persisted elsewhere - add_address_with_meta, add_many_addresses_with_meta, update_address_with_meta and add_many_address_tombstones - taking a caller-supplied guid, timestamps and change counter. The bulk variants run each record in a savepoint so one bad record neither aborts the migration nor leaves a row behind. Carrying the change counter means update_internal_address now takes a CounterUpdate of Increment, Leave or Set(i64). A bridged sync engine, so Desktop's Sync framework can drive address sync: Store::addresses_bridged_engine() exposes the address engine the crate already has, which only implemented sync15::SyncEngine and so was unreachable from Desktop's mozIBridgedSyncEngine. Nothing on mobile calls any of this; it is additive for the Desktop migration.
…enabled. (mozilla#7527) Bug 2054009 added a dependency from viaduct to rusqlite, which broke the nimbus-fml toolchain build because it doesn't have the x86_64-linux-musl-gcc compiler. viaduct doesn't actually need to link to sqlite outside the ohttp feature, so properly encode that in its Cargo.toml. Because nimbus-fml doesn't enable viaduct's ohttp feature, it won't try to link to sqlite anymore.
* feat: Adds ads-client shutdown for sqlite and telemetry * fix: small note * fix: fixes clippy * fix: Adds changelog.md update * feat: Adds mutexes from review * test: vendoring with telemetry removal disabled * test: Trying vendoring with Drop * test: remove dro * fix: doc updates * fix: Adds a comment * fix: switches to RwLock, switches to Option * fix: reorders calls, fixes changelog
I like the symmetry with `S` for `FxaState`.
A while back we purposely misspelled "Auth" as "Ath" to get around the Sentry redaction rules. Now that we're using the error ping and Grafana we don't need this anymore.
There's no need or use-case for syncs queuing up. Any "immediate" requests are time-sensitive, and the chance of a conflict is small.
Changed the `trace!` into a `breadcrumb!`. This way when we see errors we can know which URL it was from. I noticed a few in the last week and knowing the URL would have been very helpful.
Bug 2048396 removed the sync of moz.build files when vendoring application services into the Firefox tree. As pointed out by phabricator.services.mozilla.com/D298757#10668401, keeping them here is potentially confusing and I ended up confused indeed, so let's remove them as suggested.
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
19 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.
Initial implementation of opt-in impression capping.
Telemetry:
There should be no breaking changes as there is only a new option added to FFI functions, but there has been a new optional param added to native rust interface.
Documentation still needs to be updated.
Pull Request checklist
[ci full]to the PR title.