From 0834bb45471baf76fe11345ac6893304cd971420 Mon Sep 17 00:00:00 2001
From: "jan.nijtmans" <nijtmans@users.sourceforge.net>
Date: Wed, 10 Mar 2021 15:39:15 +0000
Subject: Fix [4c591fa487]: [string compare] EIAS violation

---
 generic/tclCmdMZ.c |  2 +-
 generic/tclInt.h   |  2 +-
 generic/tclUtf.c   | 17 ++++++++---------
 tests/utf.test     | 20 +-------------------
 4 files changed, 11 insertions(+), 30 deletions(-)

diff --git a/generic/tclCmdMZ.c b/generic/tclCmdMZ.c
index ee30a7a..c895039 100644
--- a/generic/tclCmdMZ.c
+++ b/generic/tclCmdMZ.c
@@ -2755,7 +2755,7 @@ TclStringCmp(
 		s1 = (char *) Tcl_GetUnicode(value1Ptr);
 		s2 = (char *) Tcl_GetUnicode(value2Ptr);
 		if (
-#ifdef WORDS_BIGENDIAN
+#if defined(WORDS_BIGENDIAN) && (TCL_UTF_MAX != 4)
 			1
 #else
 			checkEq
diff --git a/generic/tclInt.h b/generic/tclInt.h
index 5da21b0..9fec41e 100644
--- a/generic/tclInt.h
+++ b/generic/tclInt.h
@@ -4534,7 +4534,7 @@ MODULE_SCOPE void	TclDbInitNewObj(Tcl_Obj *objPtr, const char *file,
  *----------------------------------------------------------------
  */
 
-#ifdef WORDS_BIGENDIAN
+#if defined(WORDS_BIGENDIAN) && (TCL_UTF_MAX != 4)
 #   define TclUniCharNcmp(cs,ct,n) memcmp((cs),(ct),(n)*sizeof(Tcl_UniChar))
 #else /* !WORDS_BIGENDIAN */
 #   define TclUniCharNcmp Tcl_UniCharNcmp
diff --git a/generic/tclUtf.c b/generic/tclUtf.c
index 57e58c1..6e6d386 100644
--- a/generic/tclUtf.c
+++ b/generic/tclUtf.c
@@ -1592,25 +1592,24 @@ Tcl_UniCharNcmp(
     const Tcl_UniChar *uct,	/* Unicode string ucs is compared to. */
     unsigned long numChars)	/* Number of unichars to compare. */
 {
-#ifdef WORDS_BIGENDIAN
-    /*
-     * We are definitely on a big-endian machine; memcmp() is safe
-     */
-
-    return memcmp(ucs, uct, numChars*sizeof(Tcl_UniChar));
-
-#else /* !WORDS_BIGENDIAN */
     /*
      * We can't simply call memcmp() because that is not lexically correct.
      */
 
     for ( ; numChars != 0; ucs++, uct++, numChars--) {
 	if (*ucs != *uct) {
+#if TCL_UTF_MAX == 4
+	    /* special case for handling upper surrogates */
+	    if (((*ucs & 0xFC00) == 0xD800) && ((*uct & 0xFC00) != 0xD800)) {
+		return 1;
+	    } else if (((*uct & 0xFC00) == 0xD800)) {
+		return -1;
+	    }
+#endif
 	    return (*ucs - *uct);
 	}
     }
     return 0;
-#endif /* WORDS_BIGENDIAN */
 }
 
 /*
diff --git a/tests/utf.test b/tests/utf.test
index 8985313..f4d9134 100644
--- a/tests/utf.test
+++ b/tests/utf.test
@@ -1332,25 +1332,7 @@ test utf-19.1 {TclUniCharLen} -body {
 test utf-20.1 {TclUniCharNcmp} ucs4 {
     string compare [string range [format %c 0xFFFF] 0 0] [string range [format %c 0x10000] 0 0]
 } -1
-test utf-20.2 {[4c591fa487] TclUniCharNcmp/TclUtfNcmp} ucs2 {
-    set one [format %c 0xFFFF]
-    set two [format %c 0x10000]
-    set first [string compare $one $two]
-    string range $one 0 0
-    string range $two 0 0
-    set second [string compare $one $two]
-    expr {($first == $second) ? "agree" : "disagree"}
-} agree
-test utf-20.2.1 {[4c591fa487] TclUniCharNcmp/TclUtfNcmp} {utf16 knownBug} {
-    set one [format %c 0xFFFF]
-    set two [format %c 0x10000]
-    set first [string compare $one $two]
-    string range $one 0 0
-    string range $two 0 0
-    set second [string compare $one $two]
-    expr {($first == $second) ? "agree" : "disagree"}
-} agree
-test utf-20.2.2 {[4c591fa487] TclUniCharNcmp/TclUtfNcmp} ucs4 {
+test utf-20.2 {[4c591fa487] TclUniCharNcmp/TclUtfNcmp} {
     set one [format %c 0xFFFF]
     set two [format %c 0x10000]
     set first [string compare $one $two]
-- 
cgit v0.12