Bug 3: Socket Buffer Contamination

Category: JTAG/PRACC Related docs: script-reference

Symptom

load_image reported completion in approximately 0.5 seconds for a 6.9 MB file. At the PRACC transfer rate of ~97 KB/s, this operation should take roughly 70 seconds. The command appeared to succeed, but the loaded data was incomplete or garbage.

Root Cause

The _drain() method in the OpenOCD telnet wrapper had a hardcoded 0.5-second socket timeout:

def _drain(self):
    self.sock.settimeout(0.5)
    try:
        while True:
            data = self.sock.recv(4096)
            ...
    except socket.timeout:
        break  # Assumed command was done

When cmd("load_image ...") was called, _drain() waited for 0.5 seconds of silence on the socket. Since load_image produces no output during the actual transfer (only a completion message at the end), the 0.5-second silence threshold was hit almost immediately. _drain() returned, and cmd() declared success—even though load_image was still running in the background.

Subsequent commands sent to the same socket interleaved with load_image's eventual completion message, corrupting the command/response stream.

Fix

Thread the timeout from cmd() through to _drain(). For long-running commands like load_image, pass a timeout that exceeds the expected duration:

def cmd(self, command, timeout=120):
    ...
    return self._drain(timeout=timeout)

def _drain(self, timeout=0.5):
    self.sock.settimeout(timeout)
    ...

Also remove the except socket.timeout: break pattern that silently swallowed the timeout as a success signal. Instead, raise an explicit error if the expected completion prompt is not received within the timeout window.

Lesson

Hardcoded timeouts in generic I/O functions create invisible failures. When a drain/read function uses a fixed timeout to decide "the command is done," any command that takes longer than that timeout will appear to succeed instantly. The timeout must be a parameter, scaled to the expected duration of the specific operation.