[haiku-commits] r39299 - haiku/trunk/src/system/runtime_loader
- From: ingo_weinhold@xxxxxx
- To: haiku-commits@xxxxxxxxxxxxx
- Date: Thu, 4 Nov 2010 17:22:11 +0100 (CET)
Author: bonefish
Date: 2010-11-04 17:22:11 +0100 (Thu, 04 Nov 2010)
New Revision: 39299
Changeset: http://dev.haiku-os.org/changeset/39299
Modified:
haiku/trunk/src/system/runtime_loader/elf.cpp
Log:
unload_library(): Added comment about the exit hook handling.
Modified: haiku/trunk/src/system/runtime_loader/elf.cpp
===================================================================
--- haiku/trunk/src/system/runtime_loader/elf.cpp 2010-11-04 16:21:20 UTC
(rev 39298)
+++ haiku/trunk/src/system/runtime_loader/elf.cpp 2010-11-04 16:22:11 UTC
(rev 39299)
@@ -612,7 +612,18 @@
if (status == B_OK) {
while ((image = get_disposable_images().head) != NULL) {
- // call image fini here...
+ // Call the exit hooks that live in this image.
+ // Note: With the Itanium ABI this shouldn't really be
done this
+ // way anymore, since global destructors are registered
via
+ // __cxa_atexit() (the ones that are registered
dynamically) and the
+ // termination routine should call __cxa_finalize() for
the image.
+ // The reason why we still do it is that hooks
registered with
+ // atexit() aren't associated with the image. We could
find out
+ // there which image the hooks lives in and register it
+ // respectively, but since that would be done always,
that's
+ // probably more expensive than calling
+ // call_atexit_hooks_for_range() only here, which
happens only when
+ // libraries are unloaded dynamically.
if (gRuntimeLoader.call_atexit_hooks_for_range) {
gRuntimeLoader.call_atexit_hooks_for_range(
image->regions[0].vmstart,
image->regions[0].vmsize);
Other related posts:
- » [haiku-commits] r39299 - haiku/trunk/src/system/runtime_loader - ingo_weinhold
