make patchcheck: Allow Tabs in Some Cases

[ericsnowcurrently]

Currently Tools/patchcheck/reindent.py¹ replaces all tabs with spaces. However, in Tools/c-analyzer/cpython._parser.py there are embedded tab-based tables. Up to now I've basically ignored CI failures (from make patchcheck) in _parser.py, but I'd rather reindent.py be a little smarter (even if just a whitelist). ² Tools/scripts/untabify.py sounds like it may have a similar story.

Linked PRs

• gh-92266: Replace tabs with four spaces in Python files #135983
• gh-92266: reindent.py: Add allowlist for files with tabs #135984
• [3.14] gh-92266: Replace tabs with four spaces in Python files (GH-135983) #136035
• [3.13] gh-92266: Replace tabs with four spaces in Python files (GH-135983) #136036
• GH-92266: Remove embedded tabs from c-analyzer/cpython/_parser.py #137622
• gh-92266: Update pre-commit #139411
• [3.14] gh-92266: Update pre-commit (GH-139411) #139466
• [3.14] GH-92266: Remove embedded tabs from c-analyzer/cpython/_parser.py (GH-137622) #139718
• [3.13] gh-92266: Update pre-commit (GH-139411) #139760

Footnotes

1. reindent.check() is used by Tools/patchcheck/patchcheck.py (AKA make patchcheck). ↩

2. Addressing trailing-tabs-in-str-literals is even more relevant since I enabled the check-c-globals CI check. For example, see https://github.com/python/cpython/pull/102735#issuecomment-1470516458. ↩

[serhiy-storchaka]

Why not to use \t instead of literal tab?

[ezio-melotti]

I think this is the mentioned table:

cpython/Tools/c-analyzer/cpython/_parser.py

Lines 130 to 141 in d20bb33

	MACROS = clean_lines('''
	# @begin=tsv@
	glob name value
	Include/internal/*.h Py_BUILD_CORE 1
	Python/**/*.c Py_BUILD_CORE 1
	Parser/**/*.c Py_BUILD_CORE 1
	Objects/**/*.c Py_BUILD_CORE 1
	Modules/_asynciomodule.c Py_BUILD_CORE 1

Here \ts would make it less readable. Would it make sense to move that to a stand-alone .tsv file?

[ericsnowcurrently]

Here \ts would make it less readable.

Correct.

Would it make sense to move that to a stand-alone .tsv file?

I considered that but didn't have time to try it out. I'd rather have that particular data stay in the current file if possible though. One alternative is to switch to csv. I don't recall that these specific tables need tsv.

[ericsnowcurrently]

CC @AlexWaygood

[hugovk]

Up to now I've basically ignored CI failures (from make patchcheck) in _parser.py

Tools/patchcheck/reindent.py is no longer run as part of make patchcheck, and patchcheck is no longer run on the CI.

reindent.py includes a few lints:

• strip trailing whitespace (now done by trailing-whitespace in CI via pre-commit)
• replace tabs with spaces
• reindent to multiples of four-space indents

PR #135983 re-adds the second of these to the CI via pre-commit.

but I'd rather reindent.py be a little smarter (even if just a whitelist).

PR #135984 adds Tools/c-analyzer/cpython/_parser.py to an allowlist in reindent.py so its tabs are not changed. But as reindent.py is no longer part of the CI, and therefore less often run, this is less pressing.

Tools/scripts/untabify.py sounds like it may have a similar story.

It does. We could easily add an allowlist here too, possibly shared with reindent.py. However, this is not run by the CI, by any make commands, nor is recommended by the devguide, so I've not changed it.