From: Paul Eggert <eggert@cs.ucla.edu>
Date: Thu, 16 Nov 2023 19:34:55 +0000 (-0800)
Subject: uniq: fix bug with -w in multibyte locales
X-Git-Tag: v9.5~102
X-Git-Url: http://git.ipfire.org/cgi-bin/gitweb.cgi?a=commitdiff_plain;h=74b9d6a6e872553dcebfc41f8cb962d5d85d77e5;p=thirdparty%2Fcoreutils.git

uniq: fix bug with -w in multibyte locales

-w counted bytes not characters, which is wrong in multibyte locales.
This bug exists even in Fedora, which is why the recently-added
test cases from Fedora didn’t catch it.
* src/uniq.c (find_field): New arg PLEN.  All callers changed.
Compute length of field correctly in multi-byte locales.
(different): Don’t worry about check_chars; find_field now does that.
* tests/uniq/uniq.pl: Test for this bug.
---

diff --git a/src/uniq.c b/src/uniq.c
index 16cf9a0828..017b7a6d4d 100644
--- a/src/uniq.c
+++ b/src/uniq.c
@@ -240,24 +240,39 @@ newline_or_blank (mcel_t g)
 }
 
 /* Given a linebuffer LINE,
-   return a pointer to the beginning of the line's field to be compared. */
+   return a pointer to the beginning of the line's field to be compared.
+   Store into *PLEN the length in bytes of FIELD.  */
 
-ATTRIBUTE_PURE
 static char *
-find_field (struct linebuffer const *line)
+find_field (struct linebuffer const *line, idx_t *plen)
 {
   char *lp = line->buffer;
   char const *lim = lp + line->length - 1;
 
-  for (idx_t count = 0; count < skip_fields && lp < lim; count++)
+  for (idx_t i = skip_fields; 0 < i && lp < lim; i--)
     {
       lp = skip_buf_matching (lp, lim, newline_or_blank, true);
       lp = skip_buf_matching (lp, lim, newline_or_blank, false);
     }
 
-  for (idx_t s = skip_chars; lp < lim && s; s--)
+  for (idx_t i = skip_chars; 0 < i && lp < lim; i--)
     lp += mcel_scan (lp, lim).len;
 
+  /* Compute the length in bytes cheaply if possible; otherwise, scan.  */
+  idx_t len;
+  if (lim - lp <= check_chars)
+    len = lim - lp;
+  else if (MB_CUR_MAX <= 1)
+    len = check_chars;
+  else
+    {
+      char *ep = lp;
+      for (idx_t i = check_chars; 0 < i && lp < lim; i--)
+        ep += mcel_scan (lp, lim).len;
+      len = ep - lp;
+    }
+
+  *plen = len;
   return lp;
 }
 
@@ -269,11 +284,6 @@ find_field (struct linebuffer const *line)
 static bool
 different (char *old, char *new, idx_t oldlen, idx_t newlen)
 {
-  if (check_chars < oldlen)
-    oldlen = check_chars;
-  if (check_chars < newlen)
-    newlen = check_chars;
-
   if (ignore_case)
     return oldlen != newlen || memcasecmp (old, new, oldlen);
   else
@@ -349,9 +359,8 @@ check_file (char const *infile, char const *outfile, char delimiter)
       while (!feof (stdin)
              && readlinebuffer_delim (thisline, stdin, delimiter) != 0)
         {
-          char *thisfield = find_field (thisline);
-          idx_t thislen = (thisline->length - 1
-                           - (thisfield - thisline->buffer));
+          idx_t thislen;
+          char *thisfield = find_field (thisline, &thislen);
           bool new_group = (!prevfield
                             || different (thisfield, prevfield,
                                           thislen, prevlen));
@@ -379,30 +388,25 @@ check_file (char const *infile, char const *outfile, char delimiter)
     }
   else
     {
-      char *prevfield;
+      if (readlinebuffer_delim (prevline, stdin, delimiter) == 0)
+        goto closefiles;
+
       idx_t prevlen;
+      char *prevfield = find_field (prevline, &prevlen);
       intmax_t match_count = 0;
       bool first_delimiter = true;
 
-      if (readlinebuffer_delim (prevline, stdin, delimiter) == 0)
-        goto closefiles;
-      prevfield = find_field (prevline);
-      prevlen = prevline->length - 1 - (prevfield - prevline->buffer);
-
       while (!feof (stdin))
         {
-          bool match;
-          char *thisfield;
-          idx_t thislen;
           if (readlinebuffer_delim (thisline, stdin, delimiter) == 0)
             {
               if (ferror (stdin))
                 goto closefiles;
               break;
             }
-          thisfield = find_field (thisline);
-          thislen = thisline->length - 1 - (thisfield - thisline->buffer);
-          match = !different (thisfield, prevfield, thislen, prevlen);
+          idx_t thislen;
+          char *thisfield = find_field (thisline, &thislen);
+          bool match = !different (thisfield, prevfield, thislen, prevlen);
           match_count += match;
 
           if (match_count == INTMAX_MAX)
diff --git a/tests/uniq/uniq.pl b/tests/uniq/uniq.pl
index f8aff5fa53..b9653a5bf6 100755
--- a/tests/uniq/uniq.pl
+++ b/tests/uniq/uniq.pl
@@ -292,6 +292,15 @@ if ($mb_locale ne 'C')
         push @new, ["$test_name-mb", @new_t, {ENV => "LC_ALL=$mb_locale"}];
       }
     push @Tests, @new;
+
+    # Test that -w counts characters, not bytes.
+    my $trouble_with_w1 = "à\ná\n";
+    my @Locale_Tests =
+    (
+      ['w1-mb', '-w1',  {IN => $trouble_with_w1}, {OUT => $trouble_with_w1},
+        {ENV => "LC_ALL=$mb_locale"}]
+    );
+    push @Tests, @Locale_Tests;
    }
 
 # Remember that triple_test creates from each test with exactly one "IN"