stdout width inconsistent

[riesentoaster]

When running the tests of Fandango, specifically test_cli.test_help, I'm getting a weird error. The test looks as follows:

import os
import shlex
import shutil
import subprocess
import unittest

from fandango.cli import get_parser


class test_cli(unittest.TestCase):

    def run_command(self, command):
        proc = subprocess.Popen(
            command,
            stdout=subprocess.PIPE,
            stderr=subprocess.PIPE,
        )
        out, err = proc.communicate()
        return out.decode(), err.decode(), proc.returncode

    def test_help(self):
        command = shlex.split("fandango --help")
        out, err, code = self.run_command(command)
        parser = get_parser(True)
        self.maxDiff = 100000
        self.assertEqual(0, code)
        self.assertEqual(out, parser.format_help())
        self.assertEqual(err, "")

Now, if I run pytest -k test_help, the test succeeds. If I however specify -n {1,auto}, I get the following error:

E       AssertionError: 'usag[45 chars]uiet] {fuzz,parse,shell,help,copyright,version[927 chars]n.\n' != 'usag[45 chars]uiet]\n                {fuzz,parse,shell,help,[944 chars]n.\n'
E       - usage: fandango [-h] [--version] [--verbose | --quiet] {fuzz,parse,shell,help,copyright,version} ...
E       + usage: fandango [-h] [--version] [--verbose | --quiet]
E       +                 {fuzz,parse,shell,help,copyright,version} ...
E         
E         The access point to the Fandango framework
E         
E         options:
E           -h, --help            show this help message and exit
E           --version             show version number
E           --verbose, -v         increase verbosity. Can be given multiple times (-vv)
E           --quiet, -q           decrease verbosity. Can be given multiple times (-qq)
E         
E         commands:
E           {fuzz,parse,shell,help,copyright,version}
E                                 the command to execute
E             fuzz                produce outputs from .fan files and test programs
E             parse               parse input file(s) according to .fan spec
E             shell               run an interactive shell (default)
E             help                show this help and exit
E             copyright           show copyright
E             version             show version
E         
E         Use `fandango help` to get a list of commands.
E         Use `fandango help COMMAND` to learn more about COMMAND.
E         See https://fandango-fuzzer.github.io/ for more information.

Seems like the stdout width is different between the two implementations? I assume this is a bug?

[RonnyPfannschmidt]

This is a necessary details

Xdist replaces stdio with io pipes for its protocol

[riesentoaster]

Is there a way to reconfigure those have the same terminal width as is used in subcommands?

parser.format_help() formats itself to this width - 2 according to shutil.get_terminal_size().columns: implementation details

[RonnyPfannschmidt]

Id recommend to impose expected widths with monkeypatching

[riesentoaster]

Would it be possible/a good idea to add a note about this in the docs somewhere? Or did I miss something? That's not behavior I'd instinctively expect from changing the test driver, even if there may be good reasons for the limitation.

[RonnyPfannschmidt]

I vaguely recall that there are some notes on the dtdio behavior

Perhaps additional keywords are needed