can1357
a5a983c26b
- Replaced @xterm/headless backing with libghostty-vt WASM for grapheme-aware widths. - Removed legacy/modern width-model split now that the terminal matches the renderer. - Reframed width-disagreement regressions around Ghostty row containment. - Added non-destructive in-band resize handling and regression coverage.
38 lines
1.4 KiB
TypeScript
38 lines
1.4 KiB
TypeScript
import { formatSeed, runStressScenario, type Scenario, type StressScenarioResult } from "./render-stress-harness";
|
|
|
|
// Subprocess entry for the randomized render-stress pool. The parent test spawns
|
|
// one `bun` process per scenario, writes the scenario JSON to stdin, and reads a
|
|
// single JSON {@link StressScenarioResult} line back on stdout. Running each
|
|
// scenario in its own process gives full isolation — fresh Ghostty WASM VT,
|
|
// fresh `process.platform`/env patches, no shared global state to coordinate —
|
|
// and lets the parent enforce a hard timeout by killing the process, which a
|
|
// Web Worker could not deliver reliably.
|
|
|
|
function serializeError(error: unknown): { error: string; stack?: string } {
|
|
if (error instanceof Error) {
|
|
return error.stack === undefined ? { error: error.message } : { error: error.message, stack: error.stack };
|
|
}
|
|
return { error: String(error) };
|
|
}
|
|
|
|
async function main(): Promise<void> {
|
|
const scenario = JSON.parse(await Bun.stdin.text()) as Scenario;
|
|
let result: StressScenarioResult;
|
|
try {
|
|
// patchEnv defaults on: this process owns its env + platform for its one
|
|
// scenario, then exits, so the patch never has to be unwound.
|
|
await runStressScenario(scenario);
|
|
result = { ok: true };
|
|
} catch (error) {
|
|
result = {
|
|
ok: false,
|
|
scenario: scenario.name,
|
|
seed: formatSeed(scenario.seed),
|
|
...serializeError(error),
|
|
};
|
|
}
|
|
await Bun.write(Bun.stdout, JSON.stringify(result));
|
|
}
|
|
|
|
await main();
|