Re: PATCH to libstdc++ to use __cxa_thread_atexit_impl if available
On 01/19/2013 12:22 PM, Jason Merrill wrote: I think I'll hold off checking this in until Siddhesh's patch goes into glibc. It's in now, so I'm going to go ahead and check in my patch.
Re: PATCH to libstdc++ to use __cxa_thread_atexit_impl if available
Cc'ing Carlos on this so that he's aware of it. Siddhesh Jakub Jelinek ja...@redhat.com wrote: On Sat, Jan 19, 2013 at 12:22:23PM -0500, Jason Merrill wrote: Siddhesh has a patch to implement the thread atexit functionality in glibc in order to integrate better with the dynamic loader and run the cleanups in the correct order. Once it's available there, this patch will make the copy in libsupc++ use it. The main __cxa_thread_atexit function will always live in libsupc++, however, in order to maintain ABI compatibility between releases of libstdc++. Does this configure change look right, or should it go in linkage.m4 somewhere? I think I'll hold off checking this in until Siddhesh's patch goes into glibc. Yeah, it should make it into glibc first (no idea why is it taking so long), otherwise we can't be sure it won't be implemented differently in glibc. Jakub
Re: PATCH to libstdc++ to use __cxa_thread_atexit_impl if available
On 01/22/2013 07:24 PM, Siddhesh Poyarekar wrote: Cc'ing Carlos on this so that he's aware of it. Siddhesh Jakub Jelinek ja...@redhat.com wrote: On Sat, Jan 19, 2013 at 12:22:23PM -0500, Jason Merrill wrote: Siddhesh has a patch to implement the thread atexit functionality in glibc in order to integrate better with the dynamic loader and run the cleanups in the correct order. Once it's available there, this patch will make the copy in libsupc++ use it. The main __cxa_thread_atexit function will always live in libsupc++, however, in order to maintain ABI compatibility between releases of libstdc++. Does this configure change look right, or should it go in linkage.m4 somewhere? I think I'll hold off checking this in until Siddhesh's patch goes into glibc. Yeah, it should make it into glibc first (no idea why is it taking so long), otherwise we can't be sure it won't be implemented differently in glibc. Jakub If there is a patch that needs to be reviewed immediately then Siddhesh should ping that patch and CC me so I can do a timely review. I'll look into this patch tomorrow. In general we lack qualified reviewers in glibc. We try our best to get through the incoming patches, but given the new friendliness of the community we're flooded with patches, but not with reviewers. Please have patience has the community grows up. Cheers, Carlos.
Re: PATCH to libstdc++ to use __cxa_thread_atexit_impl if available
On Sat, Jan 19, 2013 at 12:22:23PM -0500, Jason Merrill wrote: Siddhesh has a patch to implement the thread atexit functionality in glibc in order to integrate better with the dynamic loader and run the cleanups in the correct order. Once it's available there, this patch will make the copy in libsupc++ use it. The main __cxa_thread_atexit function will always live in libsupc++, however, in order to maintain ABI compatibility between releases of libstdc++. Does this configure change look right, or should it go in linkage.m4 somewhere? I think I'll hold off checking this in until Siddhesh's patch goes into glibc. Yeah, it should make it into glibc first (no idea why is it taking so long), otherwise we can't be sure it won't be implemented differently in glibc. Jakub
PATCH to libstdc++ to use __cxa_thread_atexit_impl if available
Siddhesh has a patch to implement the thread atexit functionality in
glibc in order to integrate better with the dynamic loader and run the
cleanups in the correct order. Once it's available there, this patch
will make the copy in libsupc++ use it. The main __cxa_thread_atexit
function will always live in libsupc++, however, in order to maintain
ABI compatibility between releases of libstdc++.
Does this configure change look right, or should it go in linkage.m4
somewhere?
I think I'll hold off checking this in until Siddhesh's patch goes into
glibc.
commit cba5e8c2535f5950d1c78a6c35a2f83d549f37dc
Author: Jason Merrill ja...@redhat.com
Date: Sat Jan 19 12:08:25 2013 -0500
* configure.ac: Check for __cxa_thread_atexit_impl.
* libsupc++/atexit_thread.cc (__cxa_thread_atexit): Just forward
to it if available.
* config.h.in, configure: Regenerate.
diff --git a/libstdc++-v3/config.h.in b/libstdc++-v3/config.h.in
index dd14b91..c14862d 100644
--- a/libstdc++-v3/config.h.in
+++ b/libstdc++-v3/config.h.in
@@ -652,6 +652,9 @@
/* Define to 1 if you have the `_tanl' function. */
#undef HAVE__TANL
+/* Define to 1 if you have the `__cxa_thread_atexit_impl' function. */
+#undef HAVE___CXA_THREAD_ATEXIT_IMPL
+
/* Define as const if the declaration of iconv() needs const. */
#undef ICONV_CONST
diff --git a/libstdc++-v3/configure b/libstdc++-v3/configure
index a4cf5c9..f53b9cf 100755
--- a/libstdc++-v3/configure
+++ b/libstdc++-v3/configure
@@ -26584,6 +26584,18 @@ $as_echo #define HAVE_TLS 1 confdefs.h
fi
+ for ac_func in __cxa_thread_atexit_impl
+do :
+ ac_fn_c_check_func $LINENO __cxa_thread_atexit_impl ac_cv_func___cxa_thread_atexit_impl
+if test x$ac_cv_func___cxa_thread_atexit_impl = xyes; then :
+ cat confdefs.h _ACEOF
+#define HAVE___CXA_THREAD_ATEXIT_IMPL 1
+_ACEOF
+
+fi
+done
+
+
# For iconv support.
if test X$prefix = XNONE; then
diff --git a/libstdc++-v3/configure.ac b/libstdc++-v3/configure.ac
index 66164a2..a64fee2 100644
--- a/libstdc++-v3/configure.ac
+++ b/libstdc++-v3/configure.ac
@@ -235,6 +235,8 @@ if $GLIBCXX_IS_NATIVE; then
# For TLS support.
GCC_CHECK_TLS
+ AC_CHECK_FUNCS(__cxa_thread_atexit_impl)
+
# For iconv support.
AM_ICONV
diff --git a/libstdc++-v3/libsupc++/atexit_thread.cc b/libstdc++-v3/libsupc++/atexit_thread.cc
index 95bdcf0..ce26717 100644
--- a/libstdc++-v3/libsupc++/atexit_thread.cc
+++ b/libstdc++-v3/libsupc++/atexit_thread.cc
@@ -26,6 +26,20 @@
#include new
#include bits/gthr.h
+#if HAVE___CXA_THREAD_ATEXIT_IMPL
+
+extern C int __cxa_thread_atexit_impl (void (*func) (void *),
+ void *arg, void *d);
+extern C int
+__cxxabiv1::__cxa_thread_atexit (void (*dtor)(void *),
+ void *obj, void *dso_handle)
+ _GLIBCXX_NOTHROW
+{
+ return __cxa_thread_atexit_impl (dtor, obj, dso_handle);
+}
+
+#else /* HAVE___CXA_THREAD_ATEXIT_IMPL */
+
namespace {
// One element in a singly-linked stack of cleanups.
struct elt
@@ -116,3 +130,5 @@ __cxxabiv1::__cxa_thread_atexit (void (*dtor)(void *), void *obj, void */*dso_ha
return 0;
}
+
+#endif /* HAVE___CXA_THREAD_ATEXIT_IMPL */
