device/monitor: fix pio monitor crash when stdin is a pipe

[Socialpranker]

AI disclosure

Written by Claude (Anthropic, Sonnet 5). The entire patch, root-cause analysis, and verification below were produced by an AI agent. I (the human account owner) reviewed the diff and the verification steps before submitting.

Problem

pio device monitor (and pio run -t monitor) crashes with termios.error: (25, 'Inappropriate ioctl for device') whenever stdin isn't a real TTY, e.g.:

$ echo "" | pio run -t monitor

Root cause

Terminal.__init__ in platformio/device/monitor/terminal.py calls super().__init__(), which is miniterm.Miniterm.__init__ (from pyserial). That constructor unconditionally does self.console = Console(), and pyserial's POSIX Console.__init__ immediately calls termios.tcgetattr(sys.stdin.fileno()) to put the terminal into raw mode. There is no TTY to reconfigure when stdin is a pipe, so this raises before Terminal.__init__'s own body (or even new_terminal(), its caller) gets a chance to react.

Fix

Added NonInteractiveConsole, a small subclass of pyserial's own ConsoleBase (the same base class miniterm itself falls back to on non-POSIX platforms that don't have termios). It reads bytes from stdin directly with sys.stdin.read(1) — no raw-mode reconfiguration needed, since there's no interactive terminal involved — and reports EOF as the monitor's exit character so the writer thread stops cleanly instead of spinning on empty reads forever.

Because the crash happens inside the parent constructor call, Terminal.__init__ temporarily swaps miniterm.Console for a factory that returns NonInteractiveConsole before calling super().__init__(), only when sys.stdin.isatty() is false, restoring it immediately after (via try/finally). This is the narrowest fix I found: it changes nothing when stdin is a real terminal (the reported, working case), and the substitution window is a single synchronous constructor call, not global state kept around afterward.

Testing

I confirmed the crash first, then verified the fix, both against a real serial.serial_for_url("loop://") instance (pyserial's built-in loopback transport — no physical hardware needed) with stdin replaced by a pipe:

Before the fix (git stash on this change): Terminal(ser, ...) raises termios.error: (25, 'Inappropriate ioctl for device') — the exact error from the issue.

After the fix: Terminal(ser, ...) succeeds, uses NonInteractiveConsole; a full start()/join() cycle correctly reads piped bytes and stops cleanly on EOF with no exception recorded in pio_unexpected_exception.

Non-regression: when stdin is a real TTY, Terminal still uses pyserial's real Console — verified with a dedicated test.

Added tests/commands/test_device_monitor_terminal.py (3 tests, all passing) covering: construction doesn't crash when stdin isn't a TTY, the writer loop reads piped data and exits cleanly on EOF, and the TTY path is unchanged. Ran the existing tests/commands/test_device_monitor.py alongside it — all pass, no regressions. Also ran black, isort, and pylint --rcfile=./.pylintrc on both changed files — clean (10.00/10 on pylint).

I don't have physical hardware to test against pio run -t monitor end-to-end on a real board; the fix and tests operate at the Terminal/miniterm layer where the actual bug lives, using pyserial's loopback URL as a stand-in serial device.

Fixes #5113

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}