Re: [Fwd: dlopen regression in 1.7? (or is it just me?)]

[Christopher Faylor <cgf-use-the-mailinglist-please at cygwin dot com>]

On Wed, Aug 12, 2009 at 10:47:46AM +0200, Corinna Vinschen wrote:
>Index: lib/_cygwin_crt0_common.cc
>===================================================================
>RCS file: /cvs/src/src/winsup/cygwin/lib/_cygwin_crt0_common.cc,v
>retrieving revision 1.17
>diff -u -p -r1.17 _cygwin_crt0_common.cc
>--- lib/_cygwin_crt0_common.cc	7 Jul 2009 20:12:44 -0000	1.17
>+++ lib/_cygwin_crt0_common.cc	12 Aug 2009 08:42:32 -0000
>@@ -62,6 +62,8 @@ struct per_process_cxx_malloc __cygwin_c
> /* Set up pointers to various pieces so the dll can then use them,
>    and then jump to the dll.  */
> 
>+extern int __dynamically_loaded;
>+
> int __stdcall
> _cygwin_crt0_common (MainFunc f, per_process *u)
> {
>@@ -114,27 +116,32 @@ _cygwin_crt0_common (MainFunc f, per_pro
>   u->realloc = &realloc;
>   u->calloc = &calloc;
> 
>-  /* Likewise for the C++ memory operators - if any.  */
>-  if (newu && newu->cxx_malloc)
>+  if (!__dynamically_loaded)
>     {
>-      /* Inherit what we don't override.  */
>-#define CONDITIONALLY_OVERRIDE(MEMBER) \
>-      if (!__cygwin_cxx_malloc.MEMBER) \
>-	__cygwin_cxx_malloc.MEMBER = newu->cxx_malloc->MEMBER;
>-      CONDITIONALLY_OVERRIDE(oper_new);
>-      CONDITIONALLY_OVERRIDE(oper_new__);
>-      CONDITIONALLY_OVERRIDE(oper_delete);
>-      CONDITIONALLY_OVERRIDE(oper_delete__);
>-      CONDITIONALLY_OVERRIDE(oper_new_nt);
>-      CONDITIONALLY_OVERRIDE(oper_new___nt);
>-      CONDITIONALLY_OVERRIDE(oper_delete_nt);
>-      CONDITIONALLY_OVERRIDE(oper_delete___nt);
>+      /* Likewise for the C++ memory operators - if any. 
>+         Only do that for statically linked DLLs, otherwise the cxx_malloc
>+	 pointer could potentially point into volatile memory (dlclose). */
>+      if (newu && newu->cxx_malloc && !__dynamically_loaded)

I think the second !_dynamically_loaded is redundant isn't it?

cgf