as: add support for .ident directives

- token.def: enable T_DOT_IDENT
- tunit.h: add S_DOT_IDENT and make it share attributes with S_DOT_FILE,
  describe available section types, add tunit_strtab_section_enter(),
  rearrange declarations
- tunit.c: implement tunit_strtab_section_enter()
- parse.c: generalize .file parser, use it also for .ident
- input.c: handle S_DOT_IDENT, lookup or create .comment strtab section,
  append given string literal to that section

TODO

@@ -43,9 +43,8 @@ Tools:
 - readelf: add support for relocs
 - readelf: add support for program headers
 - as:
-  * support .ident:
-    + have .ident append string to .comment strtab
-    + finalize .comment strtab before output()
+  * tunit_{strtab_,}section_enter() should be generalized and merged
+  * what happens if user defines .comment as say text, and later enters .ident?
   * add support for named sections
   * add support for sub-sections
   * add support for data directives (.word, .asciz, etc)

as/input.c

@@ -87,6 +87,30 @@ static int do_dot_globl(struct scan_state *scan_state, struct tunit *tunit, stru
     return 0;
 }
 
+static int do_dot_ident(struct scan_state *scan_state, struct tunit *tunit, struct stmt *stmt)
+{
+    struct strtab *strtab;
+
+    strtab = tunit_strtab_section_enter(tunit, ".comment");
+    if (!strtab)
+	return -1;
+
+    if (strtab->section.sh_type == SHT_NULL) {
+	/* The ELF specification doesn't specify attribute values for .comment
+	 * sections, apart from sh_type, but GNU binutils formats them as string
+	 * tables with the following attribute values.
+	 */
+	strtab->section.sh_type = SHT_PROGBITS;
+	strtab->section.sh_flags = SHF_MERGE | SHF_STRINGS;
+	strtab->section.sh_entsize = 1;
+	strtab->section.sh_addralign = 1;
+    }
+
+    (void)strtab_enter(tunit, strtab, stmt->u.string.text);
+
+    return 0;
+}
+
 static int do_dot_size(struct scan_state *scan_state, struct tunit *tunit, struct stmt *stmt)
 {
     struct symbol *symbol;
@@ -198,6 +222,8 @@ static int interpret(struct scan_state *scan_state, struct tunit *tunit, struct
 	return do_dot_file(scan_state, tunit, stmt);
     case S_DOT_GLOBL:
 	return do_dot_globl(scan_state, tunit, stmt);
+    case S_DOT_IDENT:
+	return do_dot_ident(scan_state, tunit, stmt);
     case S_DOT_SIZE:
 	return do_dot_size(scan_state, tunit, stmt);
     case S_DOT_TEXT:

as/parse.c

@@ -35,7 +35,7 @@ static int error(struct scan_state *scan_state, const char *msg, enum token toke
     return -1;
 }
 
-static int parse_dot_file(struct scan_state *scan_state, struct stmt *stmt)
+static int parse_dot_file_or_ident(struct scan_state *scan_state, struct stmt *stmt, enum stmt_tag tag, const char *errmsg)
 {
     enum token token;
     union token_attribute token_attr;
@@ -45,11 +45,21 @@ static int parse_dot_file(struct scan_state *scan_state, struct stmt *stmt)
 	stmt->u.string.text = token_attr.text;
 	token = scan_token(scan_state, &token_attr);
 	if (token == T_NEWLINE) {
-	    stmt->tag = S_DOT_FILE;
+	    stmt->tag = tag;
 	    return 1;
 	}
     }
-    return error(scan_state, "junk after .file directive", token, &token_attr);
+    return error(scan_state, errmsg, token, &token_attr);
+}
+
+static int parse_dot_file(struct scan_state *scan_state, struct stmt *stmt)
+{
+    return parse_dot_file_or_ident(scan_state, stmt, S_DOT_FILE, "junk after .file directive");
+}
+
+static int parse_dot_ident(struct scan_state *scan_state, struct stmt *stmt)
+{
+    return parse_dot_file_or_ident(scan_state, stmt, S_DOT_IDENT, "junk after .ident directive");
 }
 
 static int parse_dot_globl(struct scan_state *scan_state, struct stmt *stmt)
@@ -428,6 +438,8 @@ int parse_stmt(struct scan_state *scan_state, struct stmt *stmt)
 	    return parse_dot_file(scan_state, stmt);
 	case T_DOT_GLOBL:
 	    return parse_dot_globl(scan_state, stmt);
+	case T_DOT_IDENT:
+	    return parse_dot_ident(scan_state, stmt);
 	case T_DOT_SIZE:
 	    return parse_dot_size(scan_state, stmt);
 	case T_DOT_TEXT:

as/token.def

@@ -25,6 +25,7 @@
 /* reserved symbols including directives; MUST come first and MUST be listed in increasing alphanumeric order */
 TOKEN(T_DOT_FILE, ".file", TAFMT_NONE)
 TOKEN(T_DOT_GLOBL, ".globl", TAFMT_NONE)
+TOKEN(T_DOT_IDENT, ".ident", TAFMT_NONE)
 TOKEN(T_DOT_SIZE, ".size", TAFMT_NONE)
 TOKEN(T_DOT_TEXT, ".text", TAFMT_NONE)
 TOKEN(T_DOT_TYPE, ".type", TAFMT_NONE)
@@ -57,7 +58,6 @@ TOKEN(T_DOT_BYTE, ".byte", TAFMT_NONE)
 TOKEN(T_DOT_COMM, ".comm", TAFMT_NONE)
 TOKEN(T_DOT_DATA, ".data", TAFMT_NONE)
 TOKEN(T_DOT_HIDDEN, ".hidden", TAFMT_NONE)
-TOKEN(T_DOT_IDENT, ".ident", TAFMT_NONE)
 TOKEN(T_DOT_INTERNAL, ".internal", TAFMT_NONE)
 TOKEN(T_DOT_LOCAL, ".local", TAFMT_NONE)
 TOKEN(T_DOT_LONG, ".long", TAFMT_NONE)

as/tunit.c

@@ -190,6 +190,39 @@ void strtab_init(struct strtab *strtab, const char *name)
     strtab->head = NULL;
 }
 
+static struct strtab *strtab_from_hashnode(const struct hashnode *hashnode)
+{
+    return hashnode ? container_of(hashnode, struct strtab, section.hashnode) : NULL;
+}
+
+/* XXX: generalize and merge with tunit_section_enter() */
+struct strtab *tunit_strtab_section_enter(struct tunit *tunit, const char *name)
+{
+    uintptr_t hashval;
+    struct strtab *strtab;
+
+    hashval = string_hash(name);
+    strtab = strtab_from_hashnode(hashtab_lookup(&tunit->sections, hashval, name));
+    if (strtab)
+	return strtab;
+
+    strtab = malloc(sizeof *strtab);
+    if (!strtab) {
+	fprintf(stderr, "%s: %s: malloc(%zu) failed: %s\n", tunit->progname, __FUNCTION__, sizeof *strtab, strerror(errno));
+	return NULL;
+    }
+
+    strtab_init(strtab, name);
+    /* XXX: undo stuff strtab_init() did which breaks .ident */
+    strtab->section.sh_type = SHT_NULL;
+
+    strtab->section.hashnode.hashval = hashval;
+    if (hashtab_insert(&tunit->sections, &strtab->section.hashnode) < 0)
+	return NULL;
+
+    return strtab;
+}
+
 /*
  * Symbols hash table.
  */

as/tunit.h

@@ -35,6 +35,7 @@ enum stmt_tag {
     /* directives */
     S_DOT_FILE,
     S_DOT_GLOBL,
+    S_DOT_IDENT,
     S_DOT_SIZE,
     S_DOT_TEXT,
     S_DOT_TYPE_FUNCTION,
@@ -47,7 +48,7 @@ struct stmt {
     struct stmt *next;
     enum stmt_tag tag;
     union {
-	struct {	/* S_DOT_FILE */
+	struct {	/* S_DOT_FILE, S_DOT_IDENT */
 	    const char *text;	/* XXX: should be pdp10_uint9_t* */
 	} string;
 	struct {	/* S_DOT_GLOBL, S_LABEL, S_DOT_SIZE, S_DOT_TYPE_FUNCTION */
@@ -63,6 +64,16 @@ struct stmt {
     } u;
 };
 
+/*
+ * Sections.
+ *
+ * There are several kinds of sections:
+ * - generic sections with an image array
+ *   these contain instructions or initialized data
+ * - strtab sections
+ *   these contain a strtab and a specialised ->output() method
+ */
+
 struct section {
     struct hashnode hashnode;
     const char *name;
@@ -80,8 +91,6 @@ struct section {
     Elf36_Word sh_entsize;		/* assigned during output */
 };
 
-void section_init(struct section *section, const char *name);
-
 struct strtab_entry {
     struct strtab_entry *next;
     const char *string;
@@ -119,13 +128,17 @@ struct tunit {
 
 int tunit_init(struct tunit *tunit, const char *progname);
 void tunit_fini(struct tunit *tunit);
+
+void section_init(struct section *section, const char *name);
 struct section *tunit_section_enter(struct tunit *tunit, const char *name);
+struct strtab *tunit_strtab_section_enter(struct tunit *tunit, const char *name);
+
+void strtab_init(struct strtab *strtab, const char *name);
+pdp10_uint36_t strtab_enter(struct tunit *tunit, struct strtab *strtab, const char *name);
+
 struct symbol *tunit_symbol_lookup(struct tunit *tunit, const char *name);
 struct symbol *tunit_symbol_enter(struct tunit *tunit, const char *name);
 
-pdp10_uint36_t strtab_enter(struct tunit *tunit, struct strtab *strtab, const char *name);
-void strtab_init(struct strtab *strtab, const char *name);
-
 /**
  * container_of - cast a member of a structure out to the containing structure
  * @ptr:	the pointer to the member.