[ronag]

Alternative to #67134. This is a
standalone change against main; it does not depend on that PR.

Keep a fast UTF-8 byte-length path for sequential one-byte strings.
The fast binding's slow fallback returns the Smi sentinel -1 without
reading or flattening the string. JavaScript then calls the regular
binding separately. Both the fast callback and its fallback avoid JS-heap
allocation; allocating string handling uses an ordinary API continuation.

Validation on Linux x64, GCC 14.2.0, release builds of v28.0.0-pre:

• Candidate 44bd1bfa65f. The local regular-callback baseline
  35cf29a820a is main 250037174ee with the removal fix from buffer: avoid fast API calls when computing UTF-8 byte length #67134,
  used only for the benchmark comparison.
• A negative control using a one-byte signature with an allocating slow
  fallback aborts both the direct and HTTP reproducers in 3/3 runs each.
• The candidate passes the direct reproducer 10/10 and HTTP 5/5. Traces
  show actual lazy deoptimization after weak objects are cleared, followed
  by normal completion, including end() and maybePrepareFinalChunk().
• All 531 buffer, HTTP, string-decoder and snapshot tests pass, including
  explicit optimized fast-acceptance and rejection checks.
• Main thread, four workers, and a user-land startup snapshot pass the
  string-representation corpus before and after optimization.
• Additional GC stress, ESLint, focused binding type checking, cpplint,
  import, C++ formatting and whitespace checks pass.

Benchmark tradeoff versus the regular-callback baseline, ten alternating
runs per binary on one fixed logical CPU, two million calls per case:

UTF-8 input	repeat=1	repeat=16	repeat=256
one-byte ASCII	+65.03% +/-12.95	-9.70% +/-12.50	-12.64% +/-6.43
Latin-1	+33.18% +/-10.41	-7.12% +/-6.46	-2.15% +/-5.28
two-byte	-8.04% +/-4.37	-7.07% +/-4.67	-1.60% +/-5.64

Values are throughput changes with 95% confidence intervals in percentage
points. The repeated one-byte strings take the fallback. Eligible flat
inputs improve, while rejected inputs pay for the additional call.
Latin-1 and base64 encoding controls were also measured; raw results cover
27 configurations and 540 samples. Four comparisons remain significant
at 5% after the benchmark tool's Holm-Bonferroni adjustment.

./node benchmark/compare.js --old ./node-baseline --new ./node-candidate \
  --runs 10 --filter buffer-bytelength-string \
  --set encoding=utf8 --set encoding=latin1 --set encoding=base64 \
  --set type=one_byte --set type=latin1 --set type=two_bytes \
  --set repeat=1 --set repeat=16 --set repeat=256 --set n=2000000 \
  --set CPUSET=1 --no-progress --analyze --csv benchmark.csv buffers

Refs: #66870

AI assistance: Codex helped prepare the implementation and tests.

[Copilot]

🔵 Needs a closer look

V8-aware maintainer review and supported-platform GC-stress CI are needed to validate the allocation-free callback/fallback contract and fast-call lazy-deoptimization behavior.

No open issues 🎉

What changed in this PR

This PR addresses the Buffer.byteLength() abort reported in #66870 while preserving a fast UTF-8 byte-length path for sequential one-byte strings. Rejected strings use a separate regular binding call so allocating string handling avoids the unsupported fast-call deoptimization path.

Changes:

• Restricts the native fast callback to one-byte strings and makes its fallback return -1 without inspecting the string.
• Adds a JavaScript wrapper and binding type declaration for the separate fast path.
• Adds coverage for optimized acceptance and rejection, string representations, and garbage-collection-triggered deoptimization.

File	Description
typings/​internalBinding/​buffer.d.ts	Declares the new fast byte-length binding.
test/​parallel/​test-buffer-bytelength-strings.js	Checks byte lengths across string representations before and after optimization.
test/​parallel/​test-buffer-bytelength-safe-fast.js	Tests fast-path acceptance, rejection, and public API results.
test/​parallel/​test-buffer-bytelength-gc.js	Adds regression coverage for deoptimization during string flattening.
src/​node_buffer.cc	Implements the restricted fast callback and non-allocating sentinel fallback.
lib/​buffer.js	Routes rejected fast calls to the regular UTF-8 binding.

_{Review effort: Balanced}

[Copilot]

🔵 Needs a closer look

V8-aware maintainer review and GC/deoptimization stress validation are needed to confirm the fast-call continuation and no-GC invariants across supported configurations.

No open issues 🎉

_{Review effort: Balanced}

[mcollina]

lgtm

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/78734/

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 90.69%. Comparing base (2500371) to head (44bd1bf).
⚠️ Report is 4 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #67136      +/-   ##
==========================================
+ Coverage   90.67%   90.69%   +0.01%     
==========================================
  Files         797      797              
  Lines      278806   278803       -3     
  Branches    53621    53619       -2     
==========================================
+ Hits       252813   252861      +48     
+ Misses      16341    16305      -36     
+ Partials     9652     9637      -15     

Files with missing lines	Coverage Δ
lib/buffer.js	99.74% <100.00%> (+<0.01%)	⬆️
src/node_buffer.cc	74.86% <100.00%> (+3.19%)	⬆️

... and 23 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.