Local perf container cannot resolve required executables
Symptom
Local assessment's matching-image perf attempt exits 127 before measurement.
Container shell cannot resolve rustc, cargo, valgrind, or gungraun-runner.
Evidence
target/assess/cargo-x-perf-container.log reports each executable as not found, glibc 2.41, then exit=127.
No instruction counts or budget verdict were produced.
Exact failed launch command is not captured in that log; repeatability requires recovering its environment and invocation.
Executable lookup failed; incorrect PATH versus missing mounts/binaries remains to be established.
Log is local evidence, not a committed fixture.
Failure sequence
Command lookup prevents the benchmark process from starting.
sequenceDiagram
participant A as Assessor
participant C as Container shell
participant P as Perf runner
A->>C: Launch pinned-image verification
C->>C: Resolve toolchain and measurement tools
C-->>A: Executables not found; exit 127
Note over C,P: Perf runner never starts
Impact
Current dirty codec changes lack fresh matching-platform instruction-budget evidence from this assessment.
This is a local verification-invocation failure, not evidence of a Luci job failure or performance regression.
Native Windows/macOS CI remains deferred and unrelated.
Reproduction and resolution
Recover container launch arguments, mounts, working directory, and effective PATH.
Resolve all four executables before invoking the existing Cargo perf command.
Verify rustc 1.98.1, glibc 2.41, Valgrind 3.25.1, and gungraun-runner 0.19.4 against the benchmark setup.
Capture a complete matching-platform run and its budget verdict; do not repin from failed or mismatched runs.
---
id: SMH-ISSUE-5Z81VL0P
type: issue
title: "Local perf container cannot resolve required executables"
specs: [SMH-SPEC-SPEC0001]
---
# Local perf container cannot resolve required executables
## Symptom
Local assessment's matching-image perf attempt exits 127 before measurement.
Container shell cannot resolve rustc, cargo, valgrind, or gungraun-runner.
## Evidence
`target/assess/cargo-x-perf-container.log` reports each executable as not found, glibc 2.41, then exit=127.
No instruction counts or budget verdict were produced.
Exact failed launch command is not captured in that log; repeatability requires recovering its environment and invocation.
Executable lookup failed; incorrect PATH versus missing mounts/binaries remains to be established.
Log is local evidence, not a committed fixture.
## Failure sequence
Command lookup prevents the benchmark process from starting.
```mermaid
sequenceDiagram
participant A as Assessor
participant C as Container shell
participant P as Perf runner
A->>C: Launch pinned-image verification
C->>C: Resolve toolchain and measurement tools
C-->>A: Executables not found; exit 127
Note over C,P: Perf runner never starts
```
## Impact
Current dirty codec changes lack fresh matching-platform instruction-budget evidence from this assessment.
This is a local verification-invocation failure, not evidence of a Luci job failure or performance regression.
Native Windows/macOS CI remains deferred and unrelated.
## Reproduction and resolution
Recover container launch arguments, mounts, working directory, and effective PATH.
Resolve all four executables before invoking the existing Cargo perf command.
Verify rustc 1.98.1, glibc 2.41, Valgrind 3.25.1, and gungraun-runner 0.19.4 against the benchmark setup.
Capture a complete matching-platform run and its budget verdict; do not repin from failed or mismatched runs.
## Contract
[Smith Specification](../../specs/SMH-SPEC-SPEC0001-smith-specification/index.md): required quality gates must pass.