Replace tools.message_exit() with sys.exit()

sys.exit() can print an error and send the appropriate error status to the shell.

changed milestone to %0.10.0

added Easy pick Interface labels

mentioned in merge request !170 (merged)

changed title from Replace tools.message_exit() by sys.exit() to Replace tools.message_exit() with sys.exit()

marked this issue as related to #316

As I understand:

• sys.exit() and sys.exit(0) are equivalent
• sys.exit(1) indicates a program error
• sys.exit("some string") indicates a program error much like sys.exit(1) but displays "some string".

This means we don't plan to handle other exit codes than 0 or 1. Am I right ?

NB: I don't know why we would want to handle other exit codes, just making sure.

We can returns all values:

>>> import sys
>>> sys.exit(18)
echo $? # echo $status with Fish
18

For more information:

>>> help(sys.exit)

But, we would not be able to return any other values than 1 while printing an error message. Which means error.

assigned to @moul

mentioned in commit fc97fb3a

mentioned in merge request !195 (merged)

mentioned in commit 19088742

removed Easy pick label

Actually, there is a part I didn’t manage to migrate since the tests are failing. I didn’t manage to catch with capsys printed message with sys.exit(), where with print() this is no problem. So, let’s keep message_exit() usage. This is fine this way.

closed

changed milestone to %Backlog

unassigned @moul

There is a way to test the string passed into the exception, by passing match argument to pytest.raises(). So, we could get rid of message_exit().

reopened

added Test enhancement labels

removed Interface label

With this solution, PT012 Ruff exception can also be removed:

grep -rn "PT012"
tests/unit/money/test_transfer_cli.py:121:    with pytest.raises(SystemExit) as pytest_exit:  # noqa: PT012
tests/unit/test_public_key.py:113:    with pytest.raises(SystemExit) as pytest_exit:  # noqa: PT012

A bit hard to fix, since values are variable and doesn’t work for some reasons.

marked this issue as related to #501 (closed)