Initial @file by RamyGeorge · Pull Request #289 · OCamlPro/gnucobol

RamyGeorge · 2026-05-03T23:26:46Z

Summary

This pull request is intended to introduce the "@file" command line option to GnuCOBOL.

Motivation

Such option would no longer limit the amount of options to that of a given OS's Terminal size.

Changes

added initial implementation for "@" detection and reading File contents
Called new process_at_file in process_command_line

Misc

I Have left a few notes on current changes and hope for feedback

RamyGeorge · 2026-05-03T23:28:14Z

+		if (atfile_options) break;
+	}
+	if (cobc_compile_mode) {
+


I initially thought i could "naively" realloc argv with new input but a simple Stack Overflow entry proved this can't be done

GitMensch

we'll need a force-push from a new local branch to not includes screen changes, it may or may not cherry-pick the current commit on that)

We then need test cases which also shows what the implementation wants to provide and then can have a look if the proposed changes work that way (= work on getting it there).

GitMensch · 2026-05-05T07:10:20Z

+		if (!path) {
+			free(path_str);
+			fprintf(stderr,"File is non existent or invalid path");
+			return "";
+		}


as seen in the GCC "spec" (documentation for @file) -> there is no warning/error to raise, the "option" is just passed as a "normal" file, so cobc will then do its normal error handling later

GitMensch · 2026-05-05T07:11:57Z

+#define NONOPTION_P ((argv[optind][0] != '-' && argv[optind][0] != '@') || argv[optind][1] == '\0')
+
+  if (argv[optind][0] == '@'){ 
+	return '@';		/*Most definitly is wrong but im messing around*/
+  }



as discussed - we need to do it all up-front and build a completely new argv/argc before even going to the getopt call

GitMensch

As seen in CI that currently leads to a SEGFAULT:

1197. run_extensions.at:1761: testing CALL unusual PROGRAM-ID. ...
../../tests/run_extensions.at:1815: $COMPILE_MODULE A@B.cob
/tmp/_actions-runner-working-dir/gnucobol/gnucobol/_build/tests/testsuite.dir/at-groups/1197/test-source: line 154: 75721 Segmentation fault      (core dumped) ( $at_check_trace; $COMPILE_MODULE A@B.cob ) >> "$at_stdout" 2>> "$at_stderr" 5>&-
--- /dev/null	2026-04-27 01:59:42.626967408 +0000
+++ /tmp/_actions-runner-working-dir/gnucobol/gnucobol/_build/tests/testsuite.dir/at-groups/1197/stderr	2026-05-18 21:19:44.588745273 +0000
@@ -0,0 +1,5 @@
+
+attempt to reference invalid memory address (signal SIGSEGV)
+
+cobc: aborting
+cobc: Please report this!
../../tests/run_extensions.at:1815: exit code was 139, expected 0
1197. run_extensions.at:1761:  FAILED (run_extensions.at:1815)

so there is still some stuff to do :-)

Most important: please always add Changelog entries as those document the changes you did - also helps for review - and getting the habit of always documenting at least on git push is a good thing; similar applies for the testsuite (would go into used_binaries.at)

GitMensch · 2026-05-19T06:58:03Z

 extern int			cobc_wants_debug;
 extern int			cb_listing_xref;
 extern int			cobc_seen_stdin;
+extern int			cobc_compile_mode;	/*0 for in-terminal command options, 1 for @file */


unused so can be dropped

GitMensch · 2026-05-19T06:58:52Z


+char * process_at_file(const char* argument_values,int* compile_mode,int* atfile_size);
+char** expand_processed_at_file(char *options, int *output_argc);
+void rebuild_argv_at_file(int * old_argc, char*** old_argv);


we only need to externalize one -> that should have a "cob_" prefix, the others can be static in cobgetopt.c

GitMensch · 2026-05-19T07:04:18Z

+		free(path_str);
+
+
+		int ret;


as discussed, that won't build when using configure CC="gcc -std=c89" CPPFLAGS="-Werror=declaration-after-statement" (or an old compiler, which is the reason why we check that)

GitMensch · 2026-05-19T07:10:40Z

+		}
+		size_t path_size = endptr - ptr + 1;
+					/* Ptr->endptr carries the directory */
+		char* path_str = malloc(path_size);


instead of a short-lived malloc on the heap use one of the following two approaches:

1 clean, wasteful: local buffer with local #definedd max
2 hacker-way: const char orig = *endptr; *endptr = 0; /*later* *endptr=orig;

still open - you don't need the malloc here and then also don't need to free the string anywhere later

GitMensch · 2026-05-19T07:11:39Z

+process_at_file(const char* argument_values,int* compile_mode,int* atfile_size){
+	char * ptr = strstr(argument_values,"@");
+	if (ptr){
+		*compile_mode = 1;


this parameter var seems obsolete altogether

This parameter still seems obsolete to me, why not dropping it?

Also: just checking for !ptr and return here can make the code easier to follow

GitMensch · 2026-05-19T07:13:57Z

+			if(length+1 >= capacity){
+				 capacity *= 4;
+         		char *newoptions = realloc(options, capacity);
+				options = newoptions;
+			}
+
+			options[length++] = (char)ret;


Is there a reason to not tokenize here?

toekenize = embedding code of expand_processed_at_file here (but directly wit the token), maybe even recursively call FUNC if the arg seen starts with an @?

GitMensch

nice progress

GitMensch · 2026-05-29T21:18:01Z

+		}
+		size_t path_size = endptr - ptr + 1;
+					/* Ptr->endptr carries the directory */
+		char* path_str = malloc(path_size);


still open - you don't need the malloc here and then also don't need to free the string anywhere later

GitMensch · 2026-05-29T21:18:30Z

 		FILE* path = fopen(path_str, "r");
-		size_t capacity = 256;
- 	 	size_t length = 0;
 		char *options = malloc(capacity);


declare options here, but malloc after the path check

GitMensch · 2026-05-29T21:25:06Z

+process_at_file(const char* argument_values,int* compile_mode,int* atfile_size){
+	char * ptr = strstr(argument_values,"@");
+	if (ptr){
+		*compile_mode = 1;


This parameter still seems obsolete to me, why not dropping it?

Also: just checking for !ptr and return here can make the code easier to follow

GitMensch · 2026-05-29T21:28:12Z

+			if(length+1 >= capacity){
+				 capacity *= 4;
+         		char *newoptions = realloc(options, capacity);
+				options = newoptions;
+			}
+
+			options[length++] = (char)ret;


toekenize = embedding code of expand_processed_at_file here (but directly wit the token), maybe even recursively call FUNC if the arg seen starts with an @?

GitMensch · 2026-05-29T21:31:38Z

  /* Test whether ARGV[optind] points to a non-option argument.  */
 #define NONOPTION_P (argv[optind][0] != '-' || argv[optind][1] == '\0')

+


revert this non-change for cleaner diff

GitMensch · 2026-05-29T21:34:07Z

+$COMPILE_MODULE @tempfile.txt prog.cob], [1], [],
+[cobc: error: only one of options 'm', 'x', 'b' may be specified
+])
+AT_CLEANUP()


end with an empty line, but add more tests here: @something (= not an existing file) should be passed as is

if not yet implemented move to another test: multiple @file (including the error case A -> B -> C -> A)

GitMensch · 2026-05-29T21:34:33Z

+        STOP RUN.   
+])
+
+AT_CHECK([echo "-x" > tempfile.txt


just use AT_DATA for that

GitMensch · 2026-05-29T21:35:18Z

+
+
+AT_SETUP(@file command line options)
+AT_KEYWORDS([syntax])


that looks like the wrong keyword, "cobc" is likely useful here, as well as "getopt", that will then also allow you to VGSUFFIX=mytests make check TESTSUITEFLAGS="-k getopt" to verify your changes via valgrind (also see https://sourceforge.net/p/gnucobol/wiki/Hacking%20GnuCOBOL/)

GitMensch · 2026-06-02T09:22:21Z

@@ -500,11 +498,11 @@ process_at_file(const char* argument_values,int* compile_mode,int* atfile_size){
    	return NULL;
 	}


the old code here was better, you just need to add an additional scope {} around that; note that I'd suggest to do the define here and use all upper-case with a name that is unlikely conflict with any system header

codecov-commenter · 2026-06-02T16:37:57Z

⚠️ Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

❌ Patch coverage is 65.51724% with 30 lines in your changes missing coverage. Please review.
⚠️ Please upload report for BASE (gitside-gnucobol-3.x@a3accbe). Learn more about missing BASE report.

Files with missing lines	Patch %	Lines
libcob/cobgetopt.c	64.70%	22 Missing and 8 partials ⚠️
❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@                   Coverage Diff                   @@
##             gitside-gnucobol-3.x     #289   +/-   ##
=======================================================
  Coverage                        ?   67.44%           
=======================================================
  Files                           ?       34           
  Lines                           ?    61615           
  Branches                        ?    16051           
=======================================================
  Hits                            ?    41555           
  Misses                          ?    14107           
  Partials                        ?     5953

☔ 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.

RamyGeorge commented May 3, 2026

View reviewed changes

GitMensch requested changes May 5, 2026

View reviewed changes

Initial implementation for @file

0598be5

RamyGeorge force-pushed the initial-atfile branch from 032019f to 0598be5 Compare May 18, 2026 21:13

new implementation

2ffbb00

GitMensch requested changes May 19, 2026

View reviewed changes

RamyGeorge and others added 3 commits May 29, 2026 20:34

Fixing potenital mem leaks and some tweaks

d056176

Merge branch 'gitside-gnucobol-3.x' into initial-atfile

130c3f7

COB_EXPIMP for windowns linker

17f3707

GitMensch requested changes May 29, 2026

View reviewed changes

GitMensch reviewed Jun 2, 2026

View reviewed changes

Comment thread libcob/cobgetopt.c Outdated

RamyGeorge force-pushed the initial-atfile branch from 53ea68e to 8036fd2 Compare June 2, 2026 10:25

ISO C90 and updates

9dbd453

RamyGeorge force-pushed the initial-atfile branch 2 times, most recently from dd0e41a to 9dbd453 Compare June 2, 2026 16:03

recursive @file + conflicts handled

5337f3c

		/* Test whether ARGV[optind] points to a non-option argument. */
		#define NONOPTION_P (argv[optind][0] != '-' \|\| argv[optind][1] == '\0')

		@@ -500,11 +498,11 @@ process_at_file(const char* argument_values,int* compile_mode,int* atfile_size){
		return NULL;
		}

Conversation

RamyGeorge commented May 3, 2026

Summary

Motivation

Changes

Misc

Uh oh!

Choose a reason for hiding this comment

Uh oh!

GitMensch left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

GitMensch left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

GitMensch left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

codecov-commenter commented Jun 2, 2026

Codecov Report

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants