Fix crash on function-like macros with no arguments

[kunalKumar-13]

Problem

A file containing a function-like macro with no arguments makes fortls abort while preprocessing:

re.PatternError: invalid group reference 10

The file, and every module it defines, then silently fails to parse — which is why #486 reports that go-to-definition, hover and completion all stop working, including for code outside the module.

Minimal reproduction from the issue:

#define ok() if(ie/=0) then; return; end if;

module A
    implicit none
contains
    subroutine B
    integer :: ie
    ie = 1
    ok()
    end subroutine
end module

The reporter noticed that #define ok ... works while #define ok() ... does not, which is the giveaway: the object-like and function-like paths differ.

Cause

expand_func_macro splits the argument list with def_args.split(","). For a zero-argument macro the argument list is the empty string, and "".split(",") returns [""], not [] — so the macro is treated as having one unnamed argument.

The substitution loop then runs:

re.sub(r"\b()\b", r"\1", sub)

That pattern matches the empty string at every word boundary, so group references get injected all through the replacement text. In the reported macro body one lands immediately before a digit in ie/=0, producing the literal \10 — which re reads as a reference to group 10 when the text is later used as a substitution template. Hence the error, and hence why the value 10 looks arbitrary.

Solution

Drop empty names before the argument count is taken. A zero-argument macro then compiles to \bok\s*\(\) and no argument substitution runs over its body.

Argument names are also passed through re.escape, which they were not before.

Testing

Two tests added to test/test_preproc_parser.py:

• test_pp_zero_argument_function_macro — the issue's macro expands to its body instead of raising
• test_pp_function_macro_arities — zero, one and two argument macros all expand correctly, so the fix does not regress the argument path

Both fail without the change. Full suite: 183 passed. pre-commit (flake8, black, isort, pyupgrade) clean.

One note: the two-argument expectation asserts ((1) + ( 2)) with the leading space intact. Argument capture does not strip whitespace, which is pre-existing behaviour and left alone so this change stays about the crash.

Fixes #486

[Copilot]

Pull request overview

Fixes a fortls preprocessor crash triggered by zero-argument function-like macros (e.g., #define ok() ...) by correctly treating empty argument lists as arity 0 and safely escaping argument names during substitution.

Changes:

• Filter out empty argument names when expanding function-like macros to prevent erroneous substitution on () and subsequent re group-reference failures.
• Escape macro argument names with re.escape before building substitution patterns.
• Add regression tests covering zero-, one-, and two-argument macro expansion; document the fix in the changelog.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 1 comment.

File	Description
fortls/parsers/internal/parser.py	Fixes macro argument parsing/substitution for zero-argument function-like macros and escapes argument names.
test/test_preproc_parser.py	Adds regression tests for zero-argument macros and arity handling.
CHANGELOG.md	Notes the crash fix and links to the reported issue.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[kunalKumar-13]

Good catch — the docstring was not raw, so it held a literal backspace instead of the text. Fixed in 17950af.

[kunalKumar-13]

The red X here is not this change. All 8 failing jobs fail at the same single step, Upload coverage to Codecov; Lint and Unittests pass in every one of them. It has been happening to every PR in the repo since about mid-July, including the project's own pre-commit-ci branch — I wrote it up with the evidence in #539.

I re-ran the CI sequence locally against this branch on Python 3.13 to be sure the change itself is clean:

black --check .                      ->  61 files unchanged
python -m fortls.schema; git diff    ->  no drift
pytest --doctest-modules -n auto     ->  196 passed

Nothing needed from me here as far as I can tell, but happy to rebase if that would help.

[kunalKumar-13]

The red builds here are not from this change. On every failing job the only failing step is Upload coverage to Codecov — Setup, Lint and Unittests all pass (196 passed), and the step fails on the action's own signature check:

gpg: Signature made Thu Jul  9 01:37:57 2026 UTC
gpg:                using RSA key 27034E7FDB850E0BBC2C62FF806BB28AED779869
gpg: Can't check signature: No public key
==> Could not verify signature. Please contact Codecov if problem continues
    Exiting...

It is repo-wide rather than specific to this PR: the most recent Tests run on master fails on the same step in the same job, and every push-triggered run since 2026-07-13 has failed the same way. #533 bumps codecov/codecov-action and may be the remedy, though its own checks predate the breakage so I can't say that from the runs alone.

Happy to rebase once CI is healthy if that would help.