grep: avoid false -Fw match in non-UTF8 multibyte locales

For example, this command would erroneously print its input line:
  echo ab | LC_CTYPE=ja_JP.eucjp grep -Fw b
This arose when the "memrchr" search for a preceding newline failed:
in that case, MB_START was not adjusted and was initially the same
as BEG, so wordchar_prev mistakenly returned 0.
* src/kwsearch.c (Fexecute): Set MB_START also when there is no
preceding newline.
* NEWS (Bug fixes): Mention it.
* tests/mb-non-UTF8-word-boundary: New file. Test for the bug.
* tests/Makefile.am (TESTS): Add it.
Reported by NIDE, Naoyuki in https://bugs.gnu.org/38223.

4 files changed, 40 insertions, 3 deletions

diff --git a/NEWS b/NEWS
index c31c6000..4d682055 100644
--- a/NEWS
+++ b/NEWS
@@ -9,6 +9,11 @@ GNU grep NEWS                                    -*- outline -*-
 
 ** Bug fixes
 
+  grep -Fw can no longer false match in non-UTF8 multibyte locales
+  For example, this command would erroneously print its input line:
+    echo ab | LC_CTYPE=ja_JP.eucjp grep -Fw b
+  [Bug#38223 introduced in grep 3.0]
+
   The exit status of 'grep -L' is no longer incorrect when standard
   output is /dev/null.
   [Bug#37716 introduced in grep 3.2]
diff --git a/src/kwsearch.c b/src/kwsearch.c
index 42567e98..65e056c6 100644
--- a/src/kwsearch.c
+++ b/src/kwsearch.c
@@ -223,12 +223,14 @@ Fexecute (void *vcp, char const *buf, size_t size, size_t *match_size,
       if (! match_words)
         goto success;
 
+      /* We need a preceding mb_start pointer.  Use the beginning of line
+         if there is a preceding newline, else BUF.  */
+      char const *bol = memrchr (mb_start, eol, beg - mb_start);
+      mb_start = bol ? bol + 1 : buf;
+
       /* Succeed if the preceding and following characters are word
          constituents.  If the following character is not a word
          constituent, keep trying with shorter matches.  */
-      char const *bol = memrchr (mb_start, eol, beg - mb_start);
-      if (bol)
-        mb_start = bol + 1;
       if (! wordchar_prev (mb_start, beg, buf + size))
         for (;;)
           {
diff --git a/tests/Makefile.am b/tests/Makefile.am
index a70c994b..d82501e7 100644
--- a/tests/Makefile.am
+++ b/tests/Makefile.am
@@ -123,6 +123,7 @@ TESTS =						\
   mb-dot-newline				\
   mb-non-UTF8-overrun				\
   mb-non-UTF8-performance			\
+  mb-non-UTF8-word-boundary			\
   multibyte-white-space				\
   multiple-begin-or-end-line			\
   null-byte					\
diff --git a/tests/mb-non-UTF8-word-boundary b/tests/mb-non-UTF8-word-boundary
new file mode 100755
index 00000000..a29193d2
--- /dev/null
+++ b/tests/mb-non-UTF8-word-boundary
@@ -0,0 +1,29 @@
+#!/bin/sh
+# grep -Fw could false-match when using a non-UTF8 multibyte locale.
+
+# Copyright 2019 Free Software Foundation, Inc.
+
+# This program is free software: you can redistribute it and/or modify
+# it under the terms of the GNU General Public License as published by
+# the Free Software Foundation, either version 3 of the License, or
+# (at your option) any later version.
+
+# This program is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+# GNU General Public License for more details.
+
+# You should have received a copy of the GNU General Public License
+# along with this program.  If not, see <https://www.gnu.org/licenses/>.
+
+. "${srcdir=.}/init.sh"; path_prepend_ ../src
+require_JP_EUC_locale_
+
+fail=0
+
+echo ab > in || framework_failure_
+
+# This would mistakenly print its input line from grep-3.0..3.3
+returns_ 1 grep -Fw b in || fail=1
+
+Exit $fail