Thank you for updating Change log entry section 👏

^{Visited at: 2026-05-05 21:15:41 UTC}

Benchmarks

Benchmark execution time: 2026-05-06 21:44:52

Comparing candidate commit afff006 in PR branch fix-tracestate with baseline commit 3619470 in branch master.

Found 0 performance improvements and 0 performance regressions! Performance is the same for 45 metrics, 1 unstable metrics.

---------------------------------(------|---^--------)-------------------------------->
                              -0.6%    0%  0.3%     +1.2%
                                 |          |        |
         lower bound of the CI --'          |        |
sample mean (center of the CI) -------------'        |
         upper bound of the CI ----------------------'

----------------------------------------|---------|---(---------^---------)---------->
                                       0%        1%  1.3%      2.2%      3.1%
                                                  |   |         |         |
       significant impact threshold --------------'   |         |         |
                      lower bound of CI --------------'         |         |
       sample mean (center of the CI) --------------------------'         |
                      upper bound of CI ----------------------------------'

suggestion (non-blocking): short-circuit on invalid encoding before split

At lib/datadog/tracing/distributed/trace_context.rb, split_tracestate (the return unless tracestate line):

The byteslice+chop pattern fixes the codex-flagged truncation case, but invalid UTF-8 input under the byte limit still hits split(',', 33) and raises ArgumentError: invalid byte sequence in UTF-8. Propagation#extract's rescue catches it, so the request continues — but each invalid-tracestate request emits a logger.error. A short-circuit at the top keeps that path quiet and aligns with the W3C spec, which requires tracestate to be ASCII-only.

return unless tracestate
return unless tracestate.valid_encoding?

valid_encoding? returns true for ASCII-8BIT regardless of byte values, so Rack's typical binary headers are unaffected. It only catches UTF-8-tagged-but-actually-invalid input.

Beyond the size limits, the rewrite changes trace_state to be whitespace-normalized: "v=1, v2=2 " extracts as "v=1,v2=2" because vendors.each(&:strip!) then vendors.join(',') reconstructs without OWS. More spec-compliant (W3C says OWS in tracestate is allowed but not significant), but it's customer-observable for any service inspecting digest.trace_state after extraction. The current changelog entry covers parsing limits — extending it to "Add parsing limits and normalize OWS in tracestate propagation headers" would catch this side effect.

@p-datadog Added the whitespace changes to the changelog. Thank you!

but invalid UTF-8 input under the byte limit still hits split(',', 33)

@p-datadog This came up on every step of these changes, and I chose not to address these immediately because they don't seem to be reported errors today, and a proper fix likely requires pervasive inspection of all the HTTP header code flow to know the best place to implement a UTF-8 validation/normalization.

Typing analysis

Note: Ignored files are excluded from the next sections.

Untyped methods

This PR introduces 3 partially typed methods, and clears 3 partially typed methods.

sig/datadog/tracing/distributed/baggage.rbs:18
└── def initialize: (fetcher: untyped, ?baggage_key: ::String, ?baggage_tag_keys: untyped) -> void
sig/datadog/tracing/distributed/baggage.rbs:20
└── def inject!: (untyped digest, untyped data) -> (nil | untyped)
sig/datadog/tracing/distributed/baggage.rbs:22
└── def extract: (untyped data) -> TraceDigest?

sig/datadog/tracing/distributed/baggage.rbs:16
└── def initialize: (fetcher: untyped, ?baggage_key: ::String, ?baggage_tag_keys: untyped) -> void
sig/datadog/tracing/distributed/baggage.rbs:18
└── def inject!: (untyped digest, untyped data) -> (nil | untyped)
sig/datadog/tracing/distributed/baggage.rbs:20
└── def extract: (untyped data) -> (nil | untyped)

If you believe a method or an attribute is rightfully untyped or partially typed, you can add # untyped:accept on the line before the definition to remove it from the stats.

🎯 Code Coverage (details)
• Patch Coverage: 100.00%
• Overall Coverage: 97.22% (+0.00%)