-
Notifications
You must be signed in to change notification settings - Fork 0
Stability
From 1.0.0, loadbearer follows semantic versioning
for the surface below. The canonical version of this is
VERSIONING.md
in the repo.
-
The CLI. The set of subcommands (
run,compare,score,soak,info,mem,list,baseline,net-server) and their flags. Removing or renaming a subcommand or a flag, or changing what a flag accepts, is breaking. Adding a subcommand or an optional flag is not. -
The
schema-tagged JSON formats. Every machine-readable document carries aschemastring:schemaproduced by loadbearer.result/1loadbearer run --output/--jsonloadbearer.soak/1loadbearer soak --output/--jsonloadbearer.compare/1loadbearer compare --jsonloadbearer.mem/1loadbearer mem --jsonRemoving, renaming or retyping a field bumps the number after the slash (
/1→/2) and is a major release. Adding an optional field does not — consumers must ignore unknown fields.loadbearer info --jsonis themachineblock ofloadbearer.result/1. -
Exit codes.
0on success, non-zero on any failure.
-
Absolute scores and letter grades for a given machine. They move whenever
the baseline is recalibrated or a kernel is refined. Use
comparefor a baseline-independent answer. - The reference baseline. It is calibration data, not API. Re-anchoring it to more or newer hardware is a normal minor/patch change.
-
Human-readable output — the
--plainlayout, the TUI, the "why" wording. - The diagnostic log — location, line format, which events are logged.
- Internal Rust API. loadbearer is a binary, not a published library crate.
The minimum supported Rust version is the rust-version in Cargo.toml.
Raising it is a minor release, called out in the changelog.
Scripting against a schema-tagged file, an exit code, or a subcommand: only a
major bump can break you. Comparing two machines: use loadbearer compare — its
verdict does not depend on the baseline and is stable across versions.
Pages
- Benchmark Methodology
- Scoring & Grades
- Comparing Machines
- The Baseline
- Configuration
- Accuracy Notes
- Fleet Deployment
- Memory by Program
- Stability
Commands
-
run— benchmark & grade -
compare— head-to-head verdict -
score— re-grade a result file -
soak— sustained-load / throttle test -
baseline— build a baseline -
net-server— real link test -
mem— per-program memory (ps_mem-style) -
info·list— inventory & catalogue
Not in the grade
-
network · gpu · battery health ·
--net-targetlink ·--soak