[dmorfav]

Fixes firebase/firebase-ios-sdk#16888

Timestamp.init(from:) checks the string format with a regex but not the ranges, then parses it with try! Google_Protobuf_Timestamp(jsonString:). Strings like "2026-13-01T00:00:00Z" pass the regex, SwiftProtobuf rejects them, and the try! crashes the process. Since AnyCodableValue tries Timestamp first for every JSON string, one String field with such a value crashes the app while it caches the server response, and also when AnyValue decodes it.

Change

• CodableTimestamp: try! → try in init(from:), which already throws. Decoding as Timestamp now throws, and AnyCodableValue falls back to .string through its existing try?.

Tests

TimestampCodableTests gets two new tests, with six values that match the regex but are out of range:

• testOutOfRangeValuesShouldThrow
• testOutOfRangeValuesInAnyValueAreKeptAsStrings

Without the fix, both crash with Fatal error: 'try!' expression unexpectedly raised an error. With the fix, swift test --filter FirebaseDataConnectUnit passes (82 tests), and so does scripts/style.sh test-only.

This PR is independent of #122: the two touch different lines. The only overlap is the # Unreleased heading in the CHANGELOG.

[gemini-code-assist]

Code Review

This pull request fixes a crash when decoding out-of-range timestamp strings by replacing try! with try during parsing, and adds corresponding unit tests. The review feedback suggests wrapping the raw SwiftProtobuf error in a domain-specific DataConnectCodecError.invalidTimestampFormat error to avoid leaking implementation details of the underlying library.

[aashishpatil-g]

LGTM

[aashishpatil-g]

Thanks for the fix PR @dmorfav !!

[dmorfav]

Thanks for the fix PR @dmorfav !!

Thanks for the review! The workflow runs (check, spm) are waiting for maintainer approval since this is my first contribution from a fork.

[dmorfav]

@aashishpatil-g the CI failure is unrelated to this change: spm-package-resolved fails resolving firebase-ios-sdk (it requires swift-tools 6.2.1, the runner has 6.1.0). The scheduled spm run on main has been failing the same way since 2026-09-25. Happy to rebase once CI is fixed, or feel free to merge if that check isn't required.

FYI @paulb777

[aashishpatil-g]

Yes, the CI failure is unrelated.

[dmorfav]

I’m keeping an eye on PR #124 so I can update mine once it’s merged.

[dmorfav]

@aashishpatil-g this PR is also green and ready for review!

[aashishpatil-g]

LGTM - resolved the CHANGELOG conflict.