re GNATS gcj/33 (gcj mangles composed characters)

	Fix for PR gcj/33:
	* jv-scan.c (help): Document --encoding.
	(options): Added `encoding' entry.
	(OPT_ENCODING): New define.
	(main): Handle --encoding.
	Include <langinfo.h> if nl_langinfo exists.
	* lang-options.h: Document --classpath, --CLASSPATH, --main, and
	--encoding.
	* jcf-parse.c Include <langinfo.h> if we have nl_langinfo.
	(parse_source_file): Correctly call java_init_lex.  Added `finput'
	argument.  Use nl_langinfo to determine default encoding.
	* java-tree.h (current_encoding): Declare.
	* parse.y (java_parser_context_restore_global): Don't restore
	`finput'.
	(java_parser_context_save_global): Don't set `finput' field.
	(java_pop_parser_context): Don't restore `finput'.  Free old lexer
	if required.
	* lang.c (current_encoding): New global.
	(lang_decode_option): Recognize `-fencoding='.
	(finish_parse): Don't close finput.
	* parse.h (struct parser_ctxt): Removed `finput' and
	`unget_utf8_value' fields.  Added `lexer' field.
	(java_init_lex): Fixed declaration.
	* lex.c (java_new_lexer): New function.
	(java_destroy_lexer): Likewise.
	(java_read_char): Added `lex' argument.  Handle iconv case.
	(java_read_unicode): Added `lex' argument.  Count backslashes in
	lexer structure.
	(java_init_lex): Added `finput' and `encoding' arguments.  Set
	`lexer' field in ctxp.
	(BAD_UTF8_VALUE): Removed.
	(java_lex): Handle seeing UEOF in the middle of a string literal.
	* lex.h: Include <iconv.h> if HAVE_ICONV defined.
	(java_lexer): New structure.
	(UNGETC): Removed.
	(GETC): Removed.
	(DEFAULT_ENCODING): New define.
	(java_destroy_lexer): Declare.

From-SVN: r36377

1 files changed, 37 insertions, 4 deletions

diff --git a/gcc/java/lex.h b/gcc/java/lex.h
index d4754ab..cf29aa1 100644
--- a/gcc/java/lex.h
+++ b/gcc/java/lex.h
@@ -35,6 +35,13 @@ extern int   lineno;
 /* A Unicode character, as read from the input file  */
 typedef unsigned short unicode_t;
 
+#ifdef HAVE_ICONV
+#include <iconv.h>
+#endif /* HAVE_ICONV */
+
+/* Default encoding to use if no encoding is specified.  */
+#define DEFAULT_ENCODING "UTF-8"
+
 /* Debug macro to print-out what we match  */
 #ifdef JAVA_LEX_DEBUG
 #ifdef JAVA_LEX_DEBUG_CHAR
@@ -96,12 +103,38 @@ typedef struct _java_lc {
   int col;
 } java_lc;
 
+typedef struct java_lexer
+{
+  /* The file from which we're reading.  */
+  FILE *finput;
 
-#define JAVA_LINE_MAX 80
+  /* Number of consecutive backslashes we've read.  */
+  int bs_count;
+
+  /* If nonzero, a value that was pushed back.  */
+  unicode_t unget_value;
+
+#ifdef HAVE_ICONV
+  /* The handle for the iconv converter we're using.  */
+  iconv_t handle;
 
-/* Macro to read and unread bytes */
-#define UNGETC(c) ungetc(c, finput)
-#define GETC()    getc(finput)
+  /* Bytes we've read from the file but have not sent to iconv.  */
+  char buffer[1024];
+
+  /* Index of first valid character in buffer, -1 if no valid
+     characters.  */
+  int first;
+
+  /* Index of last valid character in buffer, plus one.  -1 if no
+     valid characters in buffer.  */
+  int last;
+#endif /* HAVE_ICONV */
+} java_lexer;
+
+/* Destroy a lexer object.  */
+extern void java_destroy_lexer PARAMS ((java_lexer *));
+
+#define JAVA_LINE_MAX 80
 
 /* Build a location compound integer */
 #define BUILD_LOCATION() ((ctxp->elc.line << 12) | (ctxp->elc.col & 0xfff))