Re: [PATCH] memmem, strstr, strcasestr: fix bug with long periodic needle

[Bruno Haible]

Eric Blake wrote on 2010-10-05:
> @@ -24,9 +63,9 @@ dnl Additionally, check that memmem is efficient and 
> handles empty needles.
>  AC_DEFUN([gl_FUNC_MEMMEM],
>  [
>    AC_REQUIRE([gl_FUNC_MEMMEM_SIMPLE])
> -  if test $ac_cv_have_decl_memmem = yes; then
> +  if test $HAVE_MEMMEM = 1 && test $REPLACE_MEMMEM = 0; then
>      AC_CACHE_CHECK([whether memmem works in linear time],
> 

I'm getting this output from a testdir's configure on a Linux/x86 system:

  ./configure: line 19790: test: =: unary operator expected
  checking for memmem... (cached) yes

Later on, the test-memmem program eats a lot of CPU time because REPLACE_MEMMEM
has been set to 0 where 1 would have been appropriate.

This fixes it.


2010-11-20  Bruno Haible  <address@hidden>

        memmem: Fix autoconf test.
        * m4/memmem.m4 (gl_FUNC_MEMMEM): Test HAVE_DECL_MEMMEM, not HAVE_MEMMEM.

--- m4/memmem.m4.orig   Sat Nov 20 19:41:12 2010
+++ m4/memmem.m4        Sat Nov 20 19:37:47 2010
@@ -1,4 +1,4 @@
-# memmem.m4 serial 18
+# memmem.m4 serial 19
 dnl Copyright (C) 2002, 2003, 2004, 2007, 2008, 2009, 2010 Free Software
 dnl Foundation, Inc.
 dnl This file is free software; the Free Software Foundation
@@ -64,7 +64,7 @@
 AC_DEFUN([gl_FUNC_MEMMEM],
 [
   AC_REQUIRE([gl_FUNC_MEMMEM_SIMPLE])
-  if test $HAVE_MEMMEM = 1 && test $REPLACE_MEMMEM = 0; then
+  if test $HAVE_DECL_MEMMEM = 1 && test $REPLACE_MEMMEM = 0; then
     AC_CACHE_CHECK([whether memmem works in linear time],
       [gl_cv_func_memmem_works_fast],
       [AC_RUN_IFELSE([AC_LANG_PROGRAM([[