`BrokenPipeError` traceback when piping `cylc config` into `less`

[MetRonnie]

Upon exiting less you see the traceback, so does not block any functionality but just looks bad.

$ cylc config ~/cylc-src/wflow | less
Traceback (most recent call last):
  File ".../cylc-8.6.3-2/bin/cylc", line 8, in <module>
    sys.exit(main())
             ^^^^^^
  File ".../cylc-8.6.3-2/lib/python3.12/site-packages/cylc/flow/scripts/cylc.py", line 646, in main
    ret = _main(opts, cmd_args)
          ^^^^^^^^^^^^^^^^^^^^^
  File ".../cylc-8.6.3-2/lib/python3.12/site-packages/cylc/flow/scripts/cylc.py", line 706, in _main
    return execute_cmd(command, *cmd_args)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File ".../cylc-8.6.3-2/lib/python3.12/site-packages/cylc/flow/scripts/cylc.py", line 329, in execute_cmd
    entry_point.load()(*args)
  File ".../cylc-8.6.3-2/lib/python3.12/site-packages/cylc/flow/terminal.py", line 298, in wrapper
    wrapped_function(*wrapped_args, **wrapped_kwargs)
  File ".../cylc-8.6.3-2/lib/python3.12/site-packages/cylc/flow/scripts/config.py", line 159, in main
    asyncio.run(_main(parser, options, *ids))
  File ".../cylc-8.6.3-2/lib/python3.12/asyncio/runners.py", line 195, in run
    return runner.run(main)
           ^^^^^^^^^^^^^^^^
  File ".../cylc-8.6.3-2/lib/python3.12/asyncio/runners.py", line 118, in run
    return self._loop.run_until_complete(task)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File ".../cylc-8.6.3-2/lib/python3.12/asyncio/base_events.py", line 691, in run_until_complete
    return future.result()
           ^^^^^^^^^^^^^^^
  File ".../cylc-8.6.3-2/lib/python3.12/site-packages/cylc/flow/scripts/config.py", line 218, in _main
    config.pcfg.idump(
  File ".../cylc-8.6.3-2/lib/python3.12/site-packages/cylc/flow/parsec/config.py", line 185, in idump
    self.mdump(mkeys, sparse, prefix, oneline, none_str, handle=handle)
  File ".../cylc-8.6.3-2/lib/python3.12/site-packages/cylc/flow/parsec/config.py", line 203, in mdump
    self.dump(keys, sparse, prefix, none_str, handle=handle)
  File ".../cylc-8.6.3-2/lib/python3.12/site-packages/cylc/flow/parsec/config.py", line 211, in dump
    printcfg(
  File ".../cylc-8.6.3-2/lib/python3.12/site-packages/cylc/flow/parsec/util.py", line 194, in printcfg
    handle.write(msg)
  File ".../cylc-8.6.3-2/lib/python3.12/site-packages/colorama/ansitowin32.py", line 47, in write
    self.__convertor.write(text)
  File ".../cylc-8.6.3-2/lib/python3.12/site-packages/colorama/ansitowin32.py", line 177, in write
    self.write_and_convert(text)
  File ".../cylc-8.6.3-2/lib/python3.12/site-packages/colorama/ansitowin32.py", line 205, in write_and_convert
    self.write_plain_text(text, cursor, len(text))
  File ".../cylc-8.6.3-2/lib/python3.12/site-packages/colorama/ansitowin32.py", line 211, in write_plain_text
    self.wrapped.flush()
BrokenPipeError: [Errno 32] Broken pipe
Exception ignored in: <colorama.ansitowin32.StreamWrapper object at 0x7fe13bee5b50>
BrokenPipeError: [Errno 32] Broken pipe

[oliver-sanders]

Doesn't reproduce with a simple workflow but does with a more complex one.

Looks like a race condition between less exiting and the kill signal (SIGHUP?) being actioned by cylc config.

[ColemanTom]

Oh, I just made a similar issue #7250

To replicate, don't use less, use grep -q '\w', that will cause the error every single time.

It is nothing to do with less. If you use something that breaks the pipe prior to all output being written (e.g. grep -q), then Cylc is not handling BrokenPipeError.

[oliver-sanders]

Curious coincidence, this isn't a newly added issue is it?

[ColemanTom]

Curious coincidence, this isn't a newly added issue is it?

Coincidence as far as I can tell. It's been around since at least 8.3.5 (I don't have earlier versions installed).

$ for v in 8.3.5 8.5.2 8.6.3; do export CYLC_VERSION=$v; ( cylc config | grep -q '\w'; ) |& tail -n1; done
BrokenPipeError: [Errno 32] Broken pipe
BrokenPipeError: [Errno 32] Broken pipe
BrokenPipeError: [Errno 32] Broken pipe