Ruff autofix for subprocess.run(..., check=...) may be misleading; PLW1510 silently fixed even when not reported (with tool.ruff.fix = true)

[stdedos]

I was hinted by pylint I have a `subprocess.run` without explicit `check` argument. As I am also using ruff (for the autofixing), I wanted to see if ruff would handle it (as it didn't yet).

Naturally, PLW1510 [*] `subprocess.run` without explicit `check` argument exists*. I am reading from https://docs.astral.sh/ruff/rules/subprocess-run-without-check/:

Fix is always available

and

Use instead:


import subprocess

subprocess.run(["ls", "nonexistent"], check=True)  # Raises exception.
Or:


import subprocess

subprocess.run(["ls", "nonexistent"], check=False)  # Explicitly no check.

I am also reading from the docs, that subprocess.run uses a default of check=False, which means that a nonzero exit code will be ignored by default, instead of raising an exception.

I am assuming, that "Fix is always available" means that:
it "fixes" the ambiguity that goes into "blindly using subprocess.run" and assuming that it will "just work fine every time" - without the "Trust me, I know what I am doing" check=False explicitness.

It seems that ruff makes the decision for you that you meant check=False.
However, I am worried: How is the autofix of check=False supporting the user, instead of "blindsiding" them / "just silence a lint warning asap"?

(related pylint issue: pylint-dev/pylint#2848)

The alternative that I was really expecting was check=True. I understand it might not be possible to have "dual autofixes".

---

Additionally, I have set tool.ruff.fix = true, as it can be "boring" to run ruff check ..., only to then run ruff check --fix.

However, I am also blindsided by the fact that PLW1510 is not reported, but still fixed. i.e. I would expect

a.py:25:14: PLW1510 [*] `subprocess.run` without explicit `check` argument ✅
a.py:25:101: E501 Line too long (105 > 100)
...

Instead of "blindly" fixing PLW1510, and only reporting unfixables. Maybe in a "clean git repo" it is okay (and visible) what a manual invocation of ruff check does - but not if you are "troubleshooting" how is ruff working / PLW1510, etc

---

*: ... which isn't on the default ruleset, but also select = ["ALL"] is a bit scary, but that's besides the point here.

[stdedos]

Luke-warm-ingly interesting, PLW1510 is used 9 times in tool.ruff.ignore #9772 (comment)

[MichaReiser]

Thanks for raising this. I think we should remove the fix entirely because the purpose of this rule is that the user makes a deliberate choice and Ruff can't make this choice for you

[stdedos]

#17091