[HarshPopat23]

Summary

Parameter-header processing currently rebuilds descriptor-name validation structures and repeatedly scans the supplied headers. This PR reduces that work while retaining the existing public API and validation behavior.

Depends on #3031. This is a draft stacked on the benchmark PR. After #3031 merges, this branch should be rebased onto main so the final diff contains only the optimization, regression coverage and comparative measurements.

Changes

Descriptor validation

• Compare borrowed descriptor names instead of constructing lowercase string copies.
• Use an allocation-free duplicate scan for lists containing up to eight descriptors.
• Retain ordered-set validation for larger lists, using borrowed views and case-insensitive comparison.
• Continue validating descriptors on every call; no public descriptor cache or new lifetime requirements are introduced.

Header lookup

• Avoid constructing prefixed header-name strings on successful validation paths.
• For larger requests-more than eight descriptors and sixteen supplied headers—build a sorted index of borrowed mirrored-header views and use binary lookup.
• Retain the linear scan for smaller requests.
• Construct full header names when needed for error responses.

Behavior preserved

• Case-insensitive descriptor and header matching.
• Per-descriptor validation and error ordering.
• Rejection of duplicate headers matching a declared parameter.
• Acceptance of duplicate unused headers as opaque input.
• Existing UTF-8 and sentinel-decoding behavior.
• Exact numeric comparison and safe-integer boundaries.

Benchmark results

Measured using the unchanged fixtures from #3031, with three alternating baseline/candidate process runs on the same local Linux runner, GCC 13.3, Release configuration and system allocator.

Values below are medians across process runs of the harness’s per-operation realTime statistic. They are not request-latency percentiles.

Workload	Baseline	Candidate	Change
Validate headers - 2 descriptors	418.99 ns	222.58 ns	−46.88%
Validate headers - 64 descriptors	69.47 µs	24.61 µs	−64.58%
Reject headers - 2 descriptors	1.43 µs	1.24 µs	−13.83%
Reject headers - 64 descriptors	69.55 µs	26.16 µs	−62.39%
Generate headers - 2 descriptors	496.96 ns	396.24 ns	−20.27%
Generate headers - 64 descriptors	23.46 µs	23.94 µs	+2.06%

All measured small-input cases improved. Large header generation was approximately 2% slower, so this PR does not claim improvement across every workload.

The complete measurement procedure and limitations are documented in benchmark/MCP-header-results.markdown.

Regression coverage

Adds an indexed-path regression case covering:

• Mixed-case header names and reversed header order.
• Unicode values requiring sentinel encoding.
• Duplicate matching headers versus duplicate unused headers.
• Incorrect argument types.
• Exact decimal values at the safe-integer boundary.
• Integers outside the safe range.
• Case-insensitive duplicate descriptor names.

Verification

• Release build passed.
• MCP unit suite: 1,421 passed, zero failed.
• Three alternating baseline/candidate benchmark comparisons completed.
• Repository clang-format checks passed.
• git diff --check passed.

Limitations

The indexing thresholds are internal heuristics, not descriptor-count limits or a proven optimal crossover.

Measurements cover the existing 2/64-descriptor fixtures with short names. Other cardinalities, longer names, platforms and allocators may behave differently. The full repository test matrix and allocation profiling were not run.

These results describe local helper operations, not end-to-end MCP transport or LLM latency.

[augmentcode]

🤖 Augment PR Summary

Optimizes MCP parameter-header processing to reduce repeated allocations and scans.

• Replaces lowercase descriptor-name copies with borrowed string views.
• Checks case-insensitive uniqueness with a direct scan for up to eight descriptors.
• Uses a case-insensitive ordered set for larger descriptor lists.
• Continues validating descriptors on every call without introducing caching.
• Builds a sorted mirrored-header index above eight descriptors and sixteen headers.
• Uses binary lookup for indexed requests and linear scans otherwise.
• Defers full prefixed header-name construction to validation error paths.
• Preserves duplicate handling, validation ordering, decoding and numeric comparisons.
• Adds indexed-path regression coverage for casing, duplicates, encoding and numeric boundaries.
• Adds MCP benchmarks covering metadata, headers, result writing and subscriptions.
• Documents local validation improvements, the generation slowdown and measurement limitations.

_{🤖 Was this summary useful? React with 👍 or 👎}

[augmentcode]

Review completed. No suggestions at this time.

Comment augment review to trigger a new review at any time.

[cubic-dev-ai]

4 issues found across 6 files

_{You’re at about 90% of the monthly reviewed-line limit. You may want to disable incremental reviews to conserve quota. Reviews will continue until that limit is exceeded. If you need help avoiding interruptions, please contact contact@cubic.dev.}

Prompt for AI agents (unresolved issues)

Check if these issues are valid — if so, understand the root cause of each and fix them. When an issue isn't valid or won't be fixed in this PR, reply in its thread with the reason and then resolve the thread. If appropriate, use sub-agents to investigate and fix each issue separately.


<file name="benchmark/MCP-header-results.markdown">

<violation number="1" location="benchmark/MCP-header-results.markdown:5">
P3: 'PR 1' is undefined. Refer to the baseline PR by its number (#3031) or to the fixture definitions in `benchmark/MCP.markdown`.</violation>

<violation number="2" location="benchmark/MCP-header-results.markdown:5">
P2: The environment description conflicts with the methodology in `benchmark/MCP.markdown`, which requires an otherwise idle machine. A shared runner is not idle. Add the commit SHA, CPU/OS, system load, SIMD flags and `cmake -LA -N build` output that MCP.markdown requires, or note that these runs do not meet that requirement.</violation>

<violation number="3" location="benchmark/MCP-header-results.markdown:7">
P3: The table gives only medians, so the per-run values behind them are not recorded. Include the three baseline and three candidate timings for each case, or link the retained `before-N.json`/`after-N.json` outputs, as the reproduction instructions require.</violation>
</file>

<file name="src/core/mcp/parameter_headers.cc">

<violation number="1" location="src/core/mcp/parameter_headers.cc:177">
P3: The `substr(10)` offset is a magic number tied to the separate "Mcp-Param-" literal. Define one shared prefix constant and derive the offset from its size so the two cannot diverge.</violation>
</file>

<sub>Reply with feedback, questions, or to request a fix.

View guided diff | Re-trigger cubic</sub>

[cubic-dev-ai]

P2: The environment description conflicts with the methodology in benchmark/MCP.markdown, which requires an otherwise idle machine. A shared runner is not idle. Add the commit SHA, CPU/OS, system load, SIMD flags and cmake -LA -N build output that MCP.markdown requires, or note that these runs do not meet that requirement.

Prompt for AI agents

Check if this issue is valid — if so, understand the root cause and fix it. When an issue isn't valid or won't be fixed in this PR, reply in its thread with the reason and then resolve the thread. At benchmark/MCP-header-results.markdown, line 5:

<comment>The environment description conflicts with the methodology in `benchmark/MCP.markdown`, which requires an otherwise idle machine. A shared runner is not idle. Add the commit SHA, CPU/OS, system load, SIMD flags and `cmake -LA -N build` output that MCP.markdown requires, or note that these runs do not meet that requirement.</comment>

<file context>
@@ -0,0 +1,22 @@
+
+Dependent on the MCP baselines in #3031. Public signatures and validation policy are unchanged.
+
+Local Linux shared runner, GCC 13.3, CMake Release (including configured LTO/SIMD), system allocator. Identical PR 1 fixtures; three alternating baseline/candidate process runs. Values are the median across runs of the harness realTime statistic (fastest of three measurement windows), not per-request p95. No compilation ran during the retained comparisons. An earlier executable-relink collision failed to run the candidate and was discarded.
+
+| Case | Baseline median ns/op | Candidate median ns/op | Change |
</file context>

[cubic-dev-ai]

P3: 'PR 1' is undefined. Refer to the baseline PR by its number (#3031) or to the fixture definitions in benchmark/MCP.markdown.

Prompt for AI agents

Check if this issue is valid — if so, understand the root cause and fix it. When an issue isn't valid or won't be fixed in this PR, reply in its thread with the reason and then resolve the thread. At benchmark/MCP-header-results.markdown, line 5:

<comment>'PR 1' is undefined. Refer to the baseline PR by its number (#3031) or to the fixture definitions in `benchmark/MCP.markdown`.</comment>

<file context>
@@ -0,0 +1,22 @@
+
+Dependent on the MCP baselines in #3031. Public signatures and validation policy are unchanged.
+
+Local Linux shared runner, GCC 13.3, CMake Release (including configured LTO/SIMD), system allocator. Identical PR 1 fixtures; three alternating baseline/candidate process runs. Values are the median across runs of the harness realTime statistic (fastest of three measurement windows), not per-request p95. No compilation ran during the retained comparisons. An earlier executable-relink collision failed to run the candidate and was discarded.
+
+| Case | Baseline median ns/op | Candidate median ns/op | Change |
</file context>

[cubic-dev-ai]

P3: The table gives only medians, so the per-run values behind them are not recorded. Include the three baseline and three candidate timings for each case, or link the retained before-N.json/after-N.json outputs, as the reproduction instructions require.

Prompt for AI agents

Check if this issue is valid — if so, understand the root cause and fix it. When an issue isn't valid or won't be fixed in this PR, reply in its thread with the reason and then resolve the thread. At benchmark/MCP-header-results.markdown, line 7:

<comment>The table gives only medians, so the per-run values behind them are not recorded. Include the three baseline and three candidate timings for each case, or link the retained `before-N.json`/`after-N.json` outputs, as the reproduction instructions require.</comment>

<file context>
@@ -0,0 +1,22 @@
+
+Local Linux shared runner, GCC 13.3, CMake Release (including configured LTO/SIMD), system allocator. Identical PR 1 fixtures; three alternating baseline/candidate process runs. Values are the median across runs of the harness realTime statistic (fastest of three measurement windows), not per-request p95. No compilation ran during the retained comparisons. An earlier executable-relink collision failed to run the candidate and was discarded.
+
+| Case | Baseline median ns/op | Candidate median ns/op | Change |
+|---|---:|---:|---:|
+| MCP_Validate_Headers_2 | 418.99 | 222.58 | -46.88% |
</file context>

[cubic-dev-ai]

P3: The substr(10) offset is a magic number tied to the separate "Mcp-Param-" literal. Define one shared prefix constant and derive the offset from its size so the two cannot diverge.

Prompt for AI agents

Check if this issue is valid — if so, understand the root cause and fix it. When an issue isn't valid or won't be fixed in this PR, reply in its thread with the reason and then resolve the thread. At src/core/mcp/parameter_headers.cc, line 177:

<comment>The `substr(10)` offset is a magic number tied to the separate "Mcp-Param-" literal. Define one shared prefix constant and derive the offset from its size so the two cannot diverge.</comment>

<file context>
@@ -141,24 +166,57 @@ auto mcp_validate_parameter_headers(
+    indexed.reserve(headers.size());
+    for (const auto &entry : headers) {
+      if (starts_with_ignore_case(entry.first, "Mcp-Param-")) {
+        indexed.emplace_back(entry.first.substr(10), entry.second);
+      }
+    }
</file context>

[cubic-dev-ai]

1 issue found across 1 file (changes from recent commits).

_{You’re at about 90% of the monthly reviewed-line limit. You may want to disable incremental reviews to conserve quota. Reviews will continue until that limit is exceeded. If you need help avoiding interruptions, please contact contact@cubic.dev.}

Prompt for AI agents (unresolved issues)

Check if these issues are valid — if so, understand the root cause of each and fix them. When an issue isn't valid or won't be fixed in this PR, reply in its thread with the reason and then resolve the thread. If appropriate, use sub-agents to investigate and fix each issue separately.


<file name="cmake/common/targets/executable.cmake">

<violation number="1" location="cmake/common/targets/executable.cmake:42">
P2: The comment says this override covers every reference to the allocator, including allocator test targets, but it is set only on targets created through `sourcemeta_executable()`. The stacktrace test programs in `test/stacktrace/CMakeLists.txt` link `${SOURCEMETA_CORE_ALLOCATOR_TARGET}` directly and get no `WHOLE_ARCHIVE` override. Either apply the same property to those targets or narrow the comment to what this function covers.</violation>
</file>

<sub>Tip: Review your code locally with the cubic CLI to iterate faster.

View guided diff | Re-trigger cubic</sub>

[cubic-dev-ai]

P2: The comment says this override covers every reference to the allocator, including allocator test targets, but it is set only on targets created through sourcemeta_executable(). The stacktrace test programs in test/stacktrace/CMakeLists.txt link ${SOURCEMETA_CORE_ALLOCATOR_TARGET} directly and get no WHOLE_ARCHIVE override. Either apply the same property to those targets or narrow the comment to what this function covers.

Prompt for AI agents

Check if this issue is valid — if so, understand the root cause and fix it. When an issue isn't valid or won't be fixed in this PR, reply in its thread with the reason and then resolve the thread. At cmake/common/targets/executable.cmake, line 42:

<comment>The comment says this override covers every reference to the allocator, including allocator test targets, but it is set only on targets created through `sourcemeta_executable()`. The stacktrace test programs in `test/stacktrace/CMakeLists.txt` link `${SOURCEMETA_CORE_ALLOCATOR_TARGET}` directly and get no `WHOLE_ARCHIVE` override. Either apply the same property to those targets or narrow the comment to what this function covers.</comment>

<file context>
@@ -36,6 +36,13 @@ function(sourcemeta_executable)
+    if(SOURCEMETA_OS_LINUX AND NOT BUILD_SHARED_LIBS AND Mimalloc_FOUND)
+      # Load the replacement entry points before the static C++ runtime can
+      # supply its own new/delete definitions. Override all references to this
+      # target, including explicit allocator links on allocator test targets.
+      set_property(TARGET "${TARGET_NAME}" PROPERTY
+        "LINK_LIBRARY_OVERRIDE_${SOURCEMETA_CORE_ALLOCATOR_TARGET}" WHOLE_ARCHIVE)
</file context>