mengzhx/deepseek-harness
1
0
Fork 0
mirror of https://github.com/deepseek-ai/deepseek-harness.git synced 2026-09-11 04:00:38 +00:00
Code Issues Packages Projects Releases Wiki Activity

docs(code-runtime-python): drop the fixed empty-open limitation; declare the test helpers

Browse Source 
The empty-open continuation skip (5b61a8fe6) made the Known Limitations entry
stale — the held fragment array no longer grows per empty frame — so the entry
is removed on both sides. The public-surface list now declares
resolvePythonBin and readProcessStart, which the '.' entry re-exports for the
test suite.
This commit is contained in:
Chinesezjc
2026-08-31 15:06:05 +08:00
committed by Tianyi Cui
parent 716060a04a
commit ca0e3e573e
3 changed files with 4 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files
packages/code-runtime/code-runtime-python/README.md
+1 -2
View File
@@ -29,7 +29,7 @@ Choose this package to run Python model code through the code-runtime seam: regi
### What you get
The package's default export is the `PythonCodeRuntime` plugin. Its public surface also re-exports the host-side protocol vocabulary: `validateChildFrame` (rebuilds every inbound frame), the lossless-JSON codec and meters (`encodeJsonPlain`, `checkDoneValue`, `hasUnsafeIntegerToken`, `hasNonLosslessNumber`), and `logTruncationMarker` (the shared truncation-marker text). Every cap is a validated `Config` field with a default: `cpuSeconds` (60), `maxWallMs` (600000), `addressSpaceMb` (512, not applied on Darwin), `maxLogBytes` (65536), `maxValueBytes` (32768), `graceMs` (3000), and `pythonBin` (`python3`, resolved against `PATH` before the child spawns with an empty environment; a basename with no `PATH` match is rejected at load rather than silently falling to the platform default `PATH`).
The package's default export is the `PythonCodeRuntime` plugin. Its public surface also re-exports the host-side protocol vocabulary: `validateChildFrame` (rebuilds every inbound frame), the lossless-JSON codec and meters (`encodeJsonPlain`, `checkDoneValue`, `hasUnsafeIntegerToken`, `hasNonLosslessNumber`), `logTruncationMarker` (the shared truncation-marker text), plus `resolvePythonBin` (interpreter lookup against the current `PATH`) and `readProcessStart` (process-start statistics for tests). Every cap is a validated `Config` field with a default: `cpuSeconds` (60), `maxWallMs` (600000), `addressSpaceMb` (512, not applied on Darwin), `maxLogBytes` (65536), `maxValueBytes` (32768), `graceMs` (3000), and `pythonBin` (`python3`, resolved against `PATH` before the child spawns with an empty environment; a basename with no `PATH` match is rejected at load rather than silently falling to the platform default `PATH`).
### The wire
@@ -118,7 +118,6 @@ These limits define what the package does and does not cover; they are current p
- **No state persists across runs** — every request executes in a fresh subprocess; a persistent REPL-style kernel stays deferred until a backend brings its own logging scheme.
- **An fd-3 frame whose raw length exceeds 64 MiB settles the run as a worker-exit** — `maxLogBytes`/`maxValueBytes` are load-bounded to the same parser cap so an honest child's frames always fit; a model-constructed binding ARGUMENT above 64 MiB (a value with no seam-level budget) trips the same cap — an accepted residual of the OOM guard.
- **A combined log-and-value peak is not modelled by the load gate** — a model daemon thread that keeps writing while the completion value is metered and framed can add the two peaks in a way no gate admits or rejects; the run dies as `worker-exit`, containment holds, and only the failure classification is degraded.
- **An empty `open` continuation frame bills zero and holds one host slot** — a forged `{"type":"log","text":"","open":true}` flood grows the held fragment ARRAY without touching `logBudget` (each empty fragment's billed cost is `max(cost - 2, 0) = 0`). Accepted residual: per frame the host cost is far below its ~30-byte fd-3 wire cost, the flood is bounded by pipe throughput, and the trust level of model code equals bash.
- **A 1-second dual-limit `ulimit -t 1` CPU overrun is reported as `worker-exit`, not a timeout** — when the host starts under a hard CPU limit equal to the soft and that limit is 1, `_clamped` cannot lower the soft, so the kernel SIGKILLs the busy loop and SIGXCPU is never delivered; containment holds, only the classification is degraded.
- **No byte cap on intermediate binding values** — the implementation remains bounded by the lossless-JSON serialization cost and process memory, and a provider or executor may apply its own fetch cap.