perly.c: Fix EBCDIC bug

The code was changing the value in a structure on EBCDIC platforms.  It
turns out that that structure element is used later under quite limited
circumstances.  The changed value is actually only needed during an
array look-up, so we no longer store the change.

1 files changed, 10 insertions, 10 deletions

diff --git a/perly.c b/perly.c
index 44d6ce68e4..f3243d45c7 100644
--- a/perly.c
+++ b/perly.c
@@ -320,15 +320,6 @@ Perl_yyparse (pTHX_ int gramtype)
     if (parser->yychar == YYEMPTY) {
 	YYDPRINTF ((Perl_debug_log, "Reading a token: "));
 	parser->yychar = yylex();
-
-/* perly.tab is shipped based on an ASCII system; if it were to be regenerated
- * on a platform that doesn't use ASCII, this translation back would need to be
- * removed */
-#  ifdef EBCDIC
-	if (parser->yychar >= 0) {
-	    parser->yychar = NATIVE_TO_UNI(parser->yychar);
-	}
-#  endif
     }
 
     if (parser->yychar <= YYEOF) {
@@ -336,7 +327,16 @@ Perl_yyparse (pTHX_ int gramtype)
 	YYDPRINTF ((Perl_debug_log, "Now at end of input.\n"));
     }
     else {
-	yytoken = YYTRANSLATE (parser->yychar);
+        /* perly.tab is shipped based on an ASCII system, so need to index it
+         * with characters translated to ASCII.  Although it's not designed for
+         * this purpose, we can use NATIVE_TO_UNI here.  It returns its
+         * argument on ASCII platforms, and on EBCDIC translates native to
+         * ascii in the 0-255 range, leaving everything else unchanged.  This
+         * jibes with yylex() returning some bare characters in that range, but
+         * all tokens it returns are either 0, or above 255.  There could be a
+         * problem if NULs weren't 0, or were ever returned as raw chars by
+         * yylex() */
+        yytoken = YYTRANSLATE (NATIVE_TO_UNI(parser->yychar));
 	YYDSYMPRINTF ("Next token is", yytoken, &parser->yylval);
     }