Fortran parser must report error in dimension string correctly

[climbfuji]

This bug occurs with the metadata/fortran parser version used by @JulieSchramm when parsing variables in CCPP schemes. I mistakenly typed

      real(kind=kind_phys), dimension(:,;), intent(out) :: plvl, tlvl

instead of

      real(kind=kind_phys), dimension(:,:), intent(out) :: plvl, tlvl

(note the semicolon instead of the colon). The error that the parser reported was

Missing local_variables, ['plvl', 'tlvl'] in GFS_rrtmg_pre_run

while it should be something like invalid dimension specification or something like that.

[climbfuji]

I tested this with the current head of develop (Sep 18, 2025). Depending on the exact syntax used, the error message is slightly different. The first case is helpful albeit not exactly the correct message:

      real(kind_phys),    intent(in)    :: effrr_in(:,;)

-- Running ccpp_capgen.py from /home/dom/work/ccpp-framework/ccpp-f-fparse-semicolon/test/var_compatibility_test
'/home/dom/work/ccpp-framework/ccpp-f-fparse-semicolon/scripts/ccpp_capgen.py' '--debug' '--host-files' 'module_rad_ddt.meta,test_host_data.meta,test_host_mod.meta,test_host.meta' '--scheme-files' 'effr_calc.meta,effrs_calc.meta,effr_diag.meta,effr_pre.meta,effr_post.meta,rad_lw.meta,rad_sw.meta' '--suites' 'var_compatibility_suite.xml' '--host-name' 'test_host' '--output-root' '/home/dom/work/ccpp-framework/ccpp-f-fparse-semicolon/build/test/var_compatibility_test/ccpp'
-- ccpp-capgen stdout: Invalid variable_list, 'effrr_in(:,', at /home/dom/work/ccpp-framework/ccpp-f-fparse-semicolon/test/var_compatibility_test/effr_calc.F90:45

The second case, which is the error reported originally in this issue, is still the same:

      real(kind_phys), dimension(:,;), intent(in) :: effrr_in


-- Running ccpp_capgen.py from /home/dom/work/ccpp-framework/ccpp-f-fparse-semicolon/test/var_compatibility_test
'/home/dom/work/ccpp-framework/ccpp-f-fparse-semicolon/scripts/ccpp_capgen.py' '--debug' '--host-files' 'module_rad_ddt.meta,test_host_data.meta,test_host_mod.meta,test_host.meta' '--scheme-files' 'effr_calc.meta,effrs_calc.meta,effr_diag.meta,effr_pre.meta,effr_post.meta,rad_lw.meta,rad_sw.meta' '--suites' 'var_compatibility_suite.xml' '--host-name' 'test_host' '--output-root' '/home/dom/work/ccpp-framework/ccpp-f-fparse-semicolon/build/test/var_compatibility_test/ccpp'
-- ccpp-capgen stdout: Missing local_variables, ['effrr_in'] in effr_calc_run

CMake Error at cmake/ccpp_capgen.cmake:80 (message):
  CCPP cap generation FAILED: result = 1
Call Stack (most recent call first):
  test/var_compatibility_test/CMakeLists.txt:27 (ccpp_capgen)

In Fortran, a semicolon ; is used to combine code from two lines into one, maybe that is what the parser thinks and then it gets confused.

[gold2718]

In Fortran, a semicolon ; is used to combine code from two lines into one, maybe that is what the parser thinks and then it gets confused.

A more precise statement is the semicolon is used to combine multiple statements on a single line.

I tried the second example in gfortran and it complains about an invalid character (the closing parenthesis). Is that a better error message? Is it worth doing a full parse (to get a better error message) instead of looking for information for the Fortran version of the metadata table (what happens today)?

[climbfuji]

In Fortran, a semicolon ; is used to combine code from two lines into one, maybe that is what the parser thinks and then it gets confused.

A more precise statement is the semicolon is used to combine multiple statements on a single line.

I tried the second example in gfortran and it complains about an invalid character (the closing parenthesis). Is that a better error message? Is it worth doing a full parse (to get a better error message) instead of looking for information for the Fortran version of the metadata table (what happens today)?

I don't know immediately how complicated it would be to catch that we are inside an open parenthesis statement that must be closed before a semicolon is found (or can be continued on the next line, but I think the parser takes care of the line wrappings before parsing the contents)?

[climbfuji]

Closed via #687