add --output-tree=FILE option to dump the AST to a file

May be useful for debugging code generation.

Use the extension of FILE to choose the format of output:
* .ml : OCaml code format
* any other: JSON format (output validated by jsonlint-php)

cobc/ChangeLog

@@ -1,4 +1,14 @@
 
+2023-07-15  Fabrice Le Fessant <[email protected]>
+
+	* cobc.c, output_tree.c: new options to dump the AST in text format:
+	  --output-tree=<file>, and --output-tree-flags=<flags>. Format is
+	  either OCaml (for files with .ml extension) or JSON. Flags are
+	  '+/-' for enable/disable, 'c' for cb_common, 'l' for locations,
+	  't' for types, 'p' for pointers, 'i' for indentation, 'n' for
+	  newlines, 'A' for all infos. COB_OUTPUT_TREE_FLAGS env variable
+	  can also be used to set the flags.
+
 2023-07-11  Fabrice Le Fessant <[email protected]>
 
 	* parser.y: fix code generation for OPEN/CLOSE with multiple

cobc/Makefile.am

@@ -23,7 +23,7 @@ bin_PROGRAMS = cobc
 cobc_SOURCES = cobc.c cobc.h ppparse.y pplex.c parser.y scanner.c config.c \
 	reserved.c error.c tree.c tree.h field.c typeck.c codegen.c help.c \
 	config.def flag.def warning.def codeoptim.def ppparse.def \
-	codeoptim.c replace.c 
+	codeoptim.c replace.c output_tree.c
 
 #cobc_SOURCES = cobc.c cobc.h ppparse.y pplex.l parser.y scanner.l config.c
 

cobc/cobc.c

@@ -106,7 +106,8 @@ enum compile_level {
 #define	CB_FLAG_GETOPT_EBCDIC_TABLE    14
 #define	CB_FLAG_GETOPT_DEFAULT_COLSEQ  15
 #define	CB_FLAG_MEMORY_CHECK           16
-
+#define CB_FLAG_OUTPUT_TREE            17
+#define CB_FLAG_OUTPUT_TREE_FLAGS      18
 
 /* Info display limits */
 #define	CB_IMSG_SIZE		24
@@ -444,6 +445,8 @@ static int		save_all_src = 0;
 static signed int	save_c_src = 0;
 static signed int	verbose_output = 0;
 static int		cb_coverage_enabled = 0;
+static char*		output_tree_to_file = NULL;
+static char*		output_tree_flags = NULL;
 static int		cob_optimize = 0;
 
 
@@ -608,6 +611,8 @@ static const struct option long_options[] = {
 	{"MT",			CB_RQ_ARG, NULL, '!'},
 	{"MF",			CB_RQ_ARG, NULL, '@'},
 	{"coverage",	CB_NO_ARG, &cb_coverage_enabled, 1},
+	{"output-tree",		CB_RQ_ARG, NULL, CB_FLAG_OUTPUT_TREE},
+	{"output-tree-flags",	CB_RQ_ARG, NULL, CB_FLAG_OUTPUT_TREE_FLAGS},
 	{"P",			CB_OP_ARG, NULL, 'P'},
 	{"Xref",		CB_NO_ARG, NULL, 'X'},
 	{"use-extfh",		CB_RQ_ARG, NULL, 9},	/* this is used by COBOL-IT; Same is -fcallfh= */
@@ -3638,6 +3643,22 @@ process_command_line (const int argc, char **argv)
 			}
 			break;
 
+		case CB_FLAG_OUTPUT_TREE:
+			if (output_tree_to_file)
+				cobc_main_free (output_tree_to_file);
+			output_tree_to_file = cobc_main_strdup (cob_optarg);
+			break;
+
+		case CB_FLAG_OUTPUT_TREE_FLAGS:
+			if (output_tree_flags){
+				char* old = output_tree_flags;
+				output_tree_flags = cobc_main_stradd_dup(output_tree_flags, cob_optarg);
+				cobc_main_free (old);
+			} else {
+				output_tree_flags = cobc_main_strdup (cob_optarg);
+			}
+			break;
+
 		case '@':
 			/* -MF <file> */
 			cb_depend_file = fopen (cob_optarg, "w");
@@ -8079,6 +8100,9 @@ process_translate (struct filename *fn)
 		}
 	}
 
+	if (output_tree_to_file)
+		cb_output_tree_to_file (current_program, output_tree_to_file, output_tree_flags);
+
 	/* Translate to C */
 	codegen (current_program, fn->translate);
 

cobc/codegen.c

@@ -13722,6 +13722,7 @@ codegen_internal (struct cb_program *prog, const int subsequent_call)
 		}
 	}
 
+	/* output the procedure division code */
 	output_internal_function (prog, prog->parameter_list);
 
 	if (!prog->next_program) {

cobc/help.c

@@ -123,6 +123,7 @@ cobc_print_usage_common_options (void)
 	puts (_("  -A <options>          add <options> to the C compile phase"));
 	puts (_("  -Q <options>          add <options> to the C link phase"));
 	puts (_("  --coverage            instrument generated binaries for coverage"));
+	puts (_("  --output-tree=<file>  output the AST to <file> for debugging"));
 	puts (_("  --conf=<file>         user-defined dialect configuration; see -std"));
 	puts (_("  --list-reserved       display reserved words"));
 	puts (_("  --list-intrinsics     display intrinsic functions"));