Initial commit

node_modules/execa/lib/verbose/output.js

@@ -0,0 +1,60 @@
+import {BINARY_ENCODINGS} from '../arguments/encoding-option.js';
+import {TRANSFORM_TYPES} from '../stdio/type.js';
+import {verboseLog, serializeVerboseMessage} from './log.js';
+import {isFullVerbose} from './values.js';
+
+// `ignore` opts-out of `verbose` for a specific stream.
+// `ipc` cannot use piping.
+// `inherit` would result in double printing.
+// They can also lead to double printing when passing file descriptor integers or `process.std*`.
+// This only leaves with `pipe` and `overlapped`.
+export const shouldLogOutput = ({stdioItems, encoding, verboseInfo, fdNumber}) => fdNumber !== 'all'
+	&& isFullVerbose(verboseInfo, fdNumber)
+	&& !BINARY_ENCODINGS.has(encoding)
+	&& fdUsesVerbose(fdNumber)
+	&& (stdioItems.some(({type, value}) => type === 'native' && PIPED_STDIO_VALUES.has(value))
+	|| stdioItems.every(({type}) => TRANSFORM_TYPES.has(type)));
+
+// Printing input streams would be confusing.
+// Files and streams can produce big outputs, which we don't want to print.
+// We could print `stdio[3+]` but it often is redirected to files and streams, with the same issue.
+// So we only print stdout and stderr.
+const fdUsesVerbose = fdNumber => fdNumber === 1 || fdNumber === 2;
+
+const PIPED_STDIO_VALUES = new Set(['pipe', 'overlapped']);
+
+// `verbose: 'full'` printing logic with async methods
+export const logLines = async (linesIterable, stream, fdNumber, verboseInfo) => {
+	for await (const line of linesIterable) {
+		if (!isPipingStream(stream)) {
+			logLine(line, fdNumber, verboseInfo);
+		}
+	}
+};
+
+// `verbose: 'full'` printing logic with sync methods
+export const logLinesSync = (linesArray, fdNumber, verboseInfo) => {
+	for (const line of linesArray) {
+		logLine(line, fdNumber, verboseInfo);
+	}
+};
+
+// When `subprocess.stdout|stderr.pipe()` is called, `verbose` becomes a noop.
+// This prevents the following problems:
+//  - `.pipe()` achieves the same result as using `stdout: 'inherit'`, `stdout: stream`, etc. which also make `verbose` a noop.
+//    For example, `subprocess.stdout.pipe(process.stdin)` would print each line twice.
+//  - When chaining subprocesses with `subprocess.pipe(otherSubprocess)`, only the last one should print its output.
+// Detecting whether `.pipe()` is impossible without monkey-patching it, so we use the following undocumented property.
+// This is not a critical behavior since changes of the following property would only make `verbose` more verbose.
+const isPipingStream = stream => stream._readableState.pipes.length > 0;
+
+// When `verbose` is `full`, print stdout|stderr
+const logLine = (line, fdNumber, verboseInfo) => {
+	const verboseMessage = serializeVerboseMessage(line);
+	verboseLog({
+		type: 'output',
+		verboseMessage,
+		fdNumber,
+		verboseInfo,
+	});
+};