because this breaks compatibility with former versions of APR (someone
may actually be counting on buffer being flushed before exec) I'd
suggest adding
APR_DECLARE(void) apr_pool_cleanup_for_exec_ex(apr_int_32 flags)

Which will have a flag to enable one of these behaviours.
+    cleanup_for_exec = flags & APR_CLEANUP_FOR_EXEC_FLUSHES_BUFFERS;
     cleanup_pool_for_exec(global_pool);
+    cleanup_for_exec = 0;


nitpicking:

maybe add an APR_INLINE here:
+APR_DECLARE(int) APR_INLINE apr_cleanup_is_for_exec()
+{
+    return cleanup_for_exec;
+}


and a closed parens to the comment.
@@ -2112,7 +2119,9 @@
      * I can do about that (except if the child decides
      * to go out and close them )
      */




On 5/11/07, Bojan Smojver <bojan@rexursive.com> wrote:
> On Sat, 2007-04-28 at 09:02 +1000, Bojan Smojver wrote:
>
> > I guess 41119 will be left the way it is? Should we just document and
> > close?
>
> Here is that hackish patch (also attached to the bug report).
>
> What's the final verdict on this behaviour? Document and close? Or
> patch?
>
> --
> Bojan
>
>