From 998dbee6d9494c4dee29a2abe195d79d9b1c2e08 Mon Sep 17 00:00:00 2001
From: Rob Landley <rob@landley.net>
Date: Wed, 19 Apr 2006 22:22:06 +0000
Subject: Patch from Jean Wolter:
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

it looks like the introduced support for character classes and
equivalence classes is not correct. The attached patch tries to fix
some symptoms and tries to make tr behave like gnu tr for the added
test cases. The patch

 - removes if clauses with side effects
 - fixes handling of buffer pointer (strcat added characters to the
   buffer without increasing the buffer pointer)
 - re-arranges character classes to match ASCII order

regards,
Jean
---
 coreutils/tr.c | 27 ++++++++++++++++-----------
 1 file changed, 16 insertions(+), 11 deletions(-)

(limited to 'coreutils')

diff --git a/coreutils/tr.c b/coreutils/tr.c
index 15a9d17b0..752b13bf8 100644
--- a/coreutils/tr.c
+++ b/coreutils/tr.c
@@ -116,7 +116,8 @@ static unsigned int expand(const char *arg, register unsigned char *buffer)
 			arg += 3; /* Skip the assumed a-z */
 		} else if (*arg == '[') {
 			arg++;
-			if (ENABLE_FEATURE_TR_CLASSES && *arg++ == ':') {
+			i = *arg++;
+			if (ENABLE_FEATURE_TR_CLASSES && i == ':') {
 				if (strncmp(arg, "alpha", 5) == 0) {
 					for (i = 'A'; i <= 'Z'; i++)
 						*buffer++ = i;
@@ -124,12 +125,12 @@ static unsigned int expand(const char *arg, register unsigned char *buffer)
 						*buffer++ = i;
 				}
 				else if (strncmp(arg, "alnum", 5) == 0) {
+					for (i = '0'; i <= '9'; i++)
+						*buffer++ = i;
 					for (i = 'A'; i <= 'Z'; i++)
 						*buffer++ = i;
 					for (i = 'a'; i <= 'z'; i++)
 						*buffer++ = i;
-					for (i = '0'; i <= '9'; i++)
-						*buffer++ = i;
 				}
 				else if (strncmp(arg, "digit", 5) == 0)
 					for (i = '0'; i <= '9'; i++)
@@ -140,10 +141,15 @@ static unsigned int expand(const char *arg, register unsigned char *buffer)
 				else if (strncmp(arg, "upper", 5) == 0)
 					for (i = 'A'; i <= 'Z'; i++)
 						*buffer++ = i;
-				else if (strncmp(arg, "space", 5) == 0)
-					strcat((char*)buffer, " \f\n\r\t\v");
-				else if (strncmp(arg, "blank", 5) == 0)
-					strcat((char*)buffer, " \t");
+				else if (strncmp(arg, "space", 5) == 0) {
+				    const char s[] = "\t\n\v\f\r ";
+					strcat((char*)buffer, s);
+					buffer += sizeof(s) - 1;
+				}
+				else if (strncmp(arg, "blank", 5) == 0) {
+					*buffer++ = '\t';
+					*buffer++ = ' ';
+				}
 				/* gcc gives a warning if braces aren't used here */
 				else if (strncmp(arg, "punct", 5) == 0) {
 					for (i = 0; i <= ASCII; i++)
@@ -156,13 +162,13 @@ static unsigned int expand(const char *arg, register unsigned char *buffer)
 							*buffer++ = i;
 				}
 				else {
-					strcat((char*)buffer, "[:");
-					arg++;
+					*buffer++ = '[';
+					*buffer++ = ':';
 					continue;
 				}
 				break;
 			}
-			if (ENABLE_FEATURE_TR_EQUIV && *arg++ == '=') {
+			if (ENABLE_FEATURE_TR_EQUIV && i == '=') {
 				*buffer++ = *arg;
 				/* skip the closing =] */
 				arg += 3;
@@ -173,7 +179,6 @@ static unsigned int expand(const char *arg, register unsigned char *buffer)
 				arg -= 2;
 				continue;
 			}
-			i = *arg++;
 			ac = *arg++;
 			while (i <= ac)
 				*buffer++ = i++;
-- 
cgit v1.2.3