Re: open MinGW failure

[Simon Josefsson]

Bruno Haible <address@hidden> writes:

> Removing the mingw* case in the 'case' statement is wrong, because the * case
> executes an AC_CACHE_CHECK which would yield 'no' when cross-compiling to 
> mingw
> or natively compiling to mingw via wine but yield 'yes' when natively 
> compiling
> to mingw via Cygwin. In other words: The AC_CACHE_CHECK test would yield
> different results depending on the compilation environment, therefore it
> must be bypassed through a 'mingw*' case clause.
>
> pw* should be treated like mingw*, because there's no difference regarding
> target executable format. Both can be executed under Wine.

Thanks -- here is an improved patch I will be using in my projects while
we wait for Wine bugfixing.  It may not be perfect, but it works for me.

/Simon

diff --git a/m4/open.m4 b/m4/open.m4
index d705b3a..63ab918 100644
--- a/m4/open.m4
+++ b/m4/open.m4
@@ -9,7 +9,9 @@ AC_DEFUN([gl_FUNC_OPEN],
   AC_REQUIRE([AC_CANONICAL_HOST])
   case "$host_os" in
     mingw* | pw*)
-      gl_REPLACE_OPEN
+      # The misbehaviour is only under Wine, see
+      # http://bugs.winehq.org/show_bug.cgi?id=21292
+      gl_cv_func_open_slash=no
       ;;
     *)
       dnl open("foo/") should not create a file when the file name has a
@@ -46,13 +48,13 @@ changequote([,])dnl
             ])
           rm -f conftest.sl conftest.tmp conftest.lnk
         ])
-      case "$gl_cv_func_open_slash" in
-        *no)
-          AC_DEFINE([OPEN_TRAILING_SLASH_BUG], [1],
-            [Define to 1 if open() fails to recognize a trailing slash.])
-          gl_REPLACE_OPEN
-          ;;
-      esac
+      ;;
+  esac
+  case "$gl_cv_func_open_slash" in
+    *no)
+      AC_DEFINE([OPEN_TRAILING_SLASH_BUG], [1],
+                [Define to 1 if open() fails to recognize a trailing slash.])
+      gl_REPLACE_OPEN
       ;;
   esac
 ])