Widen Proc_Decl (R1214) for Fortran 2008 initial-proc-target

[perazz]

Problem

Proc_Decl.match() in src/fparser/two/Fortran2003.py (R1214) hardcodes the right-hand side of => as Null_Init. That is correct for Fortran 2003 but too narrow for Fortran 2008: F2008 widened the R1214 rule
so the right-hand side can also be an initial-proc-target (R1216) (a procedure-name). Fortran2008/ has no override yet, so any declaration like

procedure(foo_cb), pointer, nopass :: eval => dummy_output

raises FortranSyntaxError under std='f2008', even though gfortran -std=f2008 compiles and runs it without complaint.

Standard reference

From §12.4.3.6 of the Fortran 2008 standard (ISO/IEC 1539-1:2010), publicly available as J3 working document
10-007r1:

R1214  proc-decl           is procedure-entity-name [ => proc-pointer-init ]
R1216  proc-pointer-init   is null-init
                           or initial-proc-target
R1217  initial-proc-target is procedure-name
C1219 (R1214) If => appears in proc-decl, the procedure entity shall
              have the POINTER attribute.

F2003 R1214 was procedure-entity-name [ => null-init ] - same rule number, narrower body. C1219 is unchanged between standards.

Change

• New class Proc_Decl in src/fparser/two/Fortran2008/proc_decl_r1214.py, subclassing
  Fortran2003.Proc_Decl. match() calls the F2003 match first (which handles null-init), and on failure falls back to BinaryOpBase.match(Procedure_Entity_Name, "=>", Name, string) for the F2008 initial-proc-target branch. The RHS is matched as a bare Name, consistent with how fparser treats other name-bearing rules
  (a stricter implementation would introduce a dedicated Initial_Proc_Target class enforcing C1220; this is a deliberate minimal-viable choice and can be tightened later without changing the tree shape).
• Registration in src/fparser/two/Fortran2008/__init__.py alongside the existing R12xx entries.
• CHANGELOG.md entry under the current unreleased section, summarising the widening and citing J3/10-007r1 §12.4.3.6.

Tests

New module
src/fparser/two/tests/fortran2008/test_proc_decl_r1214.py with 12 cases covering:

• the F2003 null-init branch (regression guard),
• the F2008 initial-proc-target branch (parametrised over several names),
• malformed inputs raising NoMatchError,
• end-to-end module parse for both branches under std='f2008',
• explicit FortranSyntaxError check confirming the widened form is still rejected under std='f2003' (the F2003 grammar is unchanged).

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 92.22%. Comparing base (a197a85) to head (4e3df50).

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #502      +/-   ##
==========================================
+ Coverage   92.21%   92.22%   +0.01%     
==========================================
  Files          87       88       +1     
  Lines       13832    13853      +21     
==========================================
+ Hits        12755    12776      +21     
  Misses       1077     1077              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.