Re: strstr is linear test fails on my machine with glibc 2.13

[Bruno Haible]

Eric Blake wrote:
> Known glibc bug when using glibc 2.11 or newer on an SSE4.2-enabled chip:
> http://sourceware.org/bugzilla/show_bug.cgi?id=12100

Unless it is fixed, let's update the comments and cross-compilation guess.


2011-07-19  Bruno Haible  <address@hidden>

        strstr: Update cross-compilation guess.
        * m4/strstr.m4 (gl_FUNC_STRSTR): On glibc > 2.12 with x86 or x86_64
        CPUs, guess no, in view of glibc
        BZ #12100 <http://sourceware.org/bugzilla/show_bug.cgi?id=12100>.
        Suggested by Eric Blake. Reported by Reuben Thomas.

--- m4/strstr.m4.orig   Tue Jul 19 23:42:30 2011
+++ m4/strstr.m4        Tue Jul 19 23:37:10 2011
@@ -1,4 +1,4 @@
-# strstr.m4 serial 14
+# strstr.m4 serial 15
 dnl Copyright (C) 2008-2011 Free Software Foundation, Inc.
 dnl This file is free software; the Free Software Foundation
 dnl gives unlimited permission to copy and/or distribute it,
@@ -93,13 +93,15 @@
     return result;
     ]])],
         [gl_cv_func_strstr_linear=yes], [gl_cv_func_strstr_linear=no],
-        [dnl Only glibc > 2.12 and cygwin > 1.7.7 are known to have a
-         dnl bug-free strstr that works in linear time.
+        [dnl Only glibc > 2.12 on processors without SSE 4.2 instructions and
+         dnl cygwin > 1.7.7 are known to have a bug-free strstr that works in
+         dnl linear time.
          AC_EGREP_CPP([Lucky user],
            [
 #include <features.h>
 #ifdef __GNU_LIBRARY__
  #if ((__GLIBC__ == 2 && __GLIBC_MINOR__ > 12) || (__GLIBC__ > 2)) \
+     && !(defined __i386__ || defined __x86_64__) \
      && !defined __UCLIBC__
   Lucky user
  #endif
-- 
In memoriam Paul Schneider 
<http://en.wikipedia.org/wiki/Paul_Schneider_(pastor)>