Output and Errors

What is captured

The current implementation captures:

• userland output that passes through PHP output buffering, such as echo , print , and printf
• warnings and notices routed through the worker error handler
• exceptions caught by the worker execution wrapper

Reading partial output

Worker::getOutput() and Worker::getErrorOutput() may return partial buffers before completion.

<?php

foreach ($group->all() as $worker) {
    echo $worker->getOutput();
    echo $worker->getErrorOutput();
}

Reading final output

After wait() returns, the buffers exposed by the worker should reflect the final captured output seen by the parent process.

<?php

$group->wait();

foreach ($group as $worker) {
    echo $worker->getOutput();
    echo $worker->getErrorOutput();
}

What is not intercepted automatically

The current output transport does not redirect operating-system file descriptors. As a result, the following are not captured automatically:

• direct writes to STDOUT through fwrite(STDOUT, ...)
• direct writes to STDERR that bypass the worker wrapper
• any external program output unless you explicitly capture it inside the worker callback

Error handling behavior

Inside the worker callback:

• PHP errors that reach the configured worker error handler are written to the error output buffer
• uncaught throwables are converted into error output plus a non-zero exit code
• the callback return value is normalized into a process exit code

Exit code normalization

The worker normalizes callback results as follows:

Best practices

• Prefer buffered userland output inside workers.
• Treat errorOutput as a structured diagnostics channel.
• If you need descriptor-level capture, add explicit redirection in your own worker code.
• If you stream a large volume of output, keep using manager- or group-level wait() so the parent continues draining worker sockets.