feat (parser): parse empty CLASS-ID paragraph with "unimplemented" warning

[kmr-srbh]

• Add support for parsing an "empty" CLASS-ID paragraph with no contained paragraphs and divisions.
• Emit a "unimplemented" warning and exit.

[kmr-srbh]

@GitMensch @nberth could you please review?

[nberth]

Very good! One side question though (please ignore if you have discussed this with @GitMensch already): I would have though this kind of changes for OOP would target the gitside-master branch (so GnuCOBOL 4.x instead of GnuCOBOL 3.x).

[nberth]

Minor: we typically put two empty lines below an AT_CLEANUP.

[kmr-srbh]

Done in 2907b0e.

[nberth]

I think it may be worth adding additional tests here:

• One with two definitions of a class with the same name (to trigger the redefinition error);
• One where the class name that follows END CLASS does not correspond to the current CLASS-ID.

(Besides, the empty lines are not necessary in the COBOL code.)

[kmr-srbh]

We are not creating the AST currently so I think we cannot check those errors now. I think throwing the above errors requires calls to setup_program() and clean_up_program(). I will make sure to add checks for these in the PR where we start with the AST. :)

Besides, the empty lines are not necessary in the COBOL code.

Done in 2907b0e.

[GitMensch]

@nberth the plan is to bring the syntax and possibly first AST into GC3 so its parser is not trashed when OO is fed into it - returning UNSUPPORTED, then bringing that with a merge over to master/trunk where the AST is finished (by that also the syntax checks which may be not complete in 3.x) and the codegen is implemented. If I don't misremember, then that was also outlined in the GSoC proposal. Thanks for the reviews, I'll get in when you feel the use of that. Nicolas Berthier ***@***.***> schrieb am Freitag, 10. April 2026 um 16:20:

…

@nberth commented on this pull request. Very good! One side question though (please ignore if you have discussed this with @GitMensch already): I would have though this kind of support would target the `gitside-master` (so GnuCOBOL 4.x instead of GnuCOBOL 3.x). In tests/testsuite.src/syn_definition.at: > @@ -2947,3 +2947,19 @@ AT_DATA([prog.cob], [ AT_CHECK([$COMPILE_ONLY -Wno-unfinished prog.cob], [0], [], []) AT_CLEANUP + Minor: we typically put two empty lines below an `AT_CLEANUP`. In tests/testsuite.src/syn_definition.at: > @@ -2947,3 +2947,19 @@ AT_DATA([prog.cob], [ AT_CHECK([$COMPILE_ONLY -Wno-unfinished prog.cob], [0], [], []) AT_CLEANUP + +AT_SETUP([CLASS-ID Syntax Check]) +AT_KEYWORDS([CLASS-ID]) + +AT_DATA([prog.cob], [ + IDENTIFICATION DIVISION. + + CLASS-ID. MyClass. + END CLASS MyClass. + +]) I think it may be worth adding additional tests here: - One with two definitions of a class with the same name (to trigger the redefinition error); - One where the class name that follows `END CLASS` does not correspond to the current `CLASS-ID`. (Besides, the empty lines are not necessary in the COBOL code.) — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you were mentioned.