* Symbol redefinition bugs
@ 2005-11-18 19:19 Andreas Schwab
2005-11-20 12:46 ` Jeff Bailey
0 siblings, 1 reply; 6+ messages in thread
From: Andreas Schwab @ 2005-11-18 19:19 UTC (permalink / raw)
To: libc-hacker
The current binutils revealed some symbol redefinition bugs in the
ppc/ppc64 port.
Andreas.
2005-11-18 Andreas Schwab <schwab@suse.de>
* sysdeps/powerpc/powerpc32/fpu/s_lround.S: Remove useless alias.
* sysdeps/unix/sysv/linux/powerpc/powerpc32/socket.S: Handle
NO_WEAK_ALIAS.
* sysdeps/unix/sysv/linux/powerpc/powerpc64/socket.S: Likewise.
--- libc/sysdeps/powerpc/powerpc32/fpu/s_lround.S.~1.4.~ 2005-06-20 11:05:09.000000000 +0200
+++ libc/sysdeps/powerpc/powerpc32/fpu/s_lround.S 2005-11-18 19:24:09.000000000 +0100
@@ -79,7 +79,6 @@ ENTRY (__lround)
b .L9
END (__lround)
-strong_alias (__lround, __lround)
weak_alias (__lround, lround)
strong_alias (__lround, __lroundf)
--- libc/sysdeps/unix/sysv/linux/powerpc/powerpc32/socket.S.~1.4.~ 2005-06-20 11:05:20.000000000 +0200
+++ libc/sysdeps/unix/sysv/linux/powerpc/powerpc32/socket.S 2005-11-18 18:25:51.000000000 +0100
@@ -44,7 +44,11 @@
#define stackblock 20
#ifndef __socket
-#define __socket P(__,socket)
+# ifndef NO_WEAK_ALIAS
+# define __socket P(__,socket)
+# else
+# define __socket socket
+# endif
#endif
.text
@@ -114,4 +118,6 @@ ENTRY(__socket)
PSEUDO_END (__socket)
+#ifndef NO_WEAK_ALIAS
weak_alias (__socket, socket)
+#endif
--- libc/sysdeps/unix/sysv/linux/powerpc/powerpc64/socket.S.~1.5.~ 2005-01-10 10:41:04.000000000 +0100
+++ libc/sysdeps/unix/sysv/linux/powerpc/powerpc64/socket.S 2005-11-18 19:10:23.000000000 +0100
@@ -41,12 +41,12 @@
#define stackblock 80 /* offset to socket parm area. */
-#ifndef socket
-/* If this is just socket.S leave it alone! */
-#else
#ifndef __socket
-#define __socket P(__,socket)
-#endif
+# ifndef NO_WEAK_ALIAS
+# define __socket P(__,socket)
+# else
+# define __socket socket
+# endif
#endif
.text
@@ -120,4 +120,6 @@ ENTRY(__socket)
cfi_endproc
PSEUDO_END (__socket)
+#ifndef NO_WEAK_ALIAS
weak_alias (__socket, socket)
+#endif
--
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE Linux Products GmbH, Maxfeldstraße 5, 90409 Nürnberg, Germany
PGP key fingerprint = 58CA 54C7 6D53 942B 1756 01D3 44D5 214B 8276 4ED5
"And now for something completely different."
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: Symbol redefinition bugs
2005-11-18 19:19 Symbol redefinition bugs Andreas Schwab
@ 2005-11-20 12:46 ` Jeff Bailey
2005-11-20 14:38 ` Andreas Schwab
2005-11-21 21:01 ` Roland McGrath
0 siblings, 2 replies; 6+ messages in thread
From: Jeff Bailey @ 2005-11-20 12:46 UTC (permalink / raw)
To: Andreas Schwab; +Cc: libc-hacker
On ven, 2005-11-18 at 20:19 +0100, Andreas Schwab wrote:
> The current binutils revealed some symbol redefinition bugs in the
> ppc/ppc64 port.
>
> Andreas.
>
> 2005-11-18 Andreas Schwab <schwab@suse.de>
>
> * sysdeps/powerpc/powerpc32/fpu/s_lround.S: Remove useless alias.
>
> * sysdeps/unix/sysv/linux/powerpc/powerpc32/socket.S: Handle
> NO_WEAK_ALIAS.
> * sysdeps/unix/sysv/linux/powerpc/powerpc64/socket.S: Likewise.
Done this way, readelf -a showed that the various symbols went from being
WEAK to being GLOBAL. I've been testing the following patch instead for
upload to Ubuntu. (identical readelf -a output on ppc; .free, .realloc
and .malloc are GLOBAL instead of WEAK on ppc64 and I don't know why yet)
2005-11-17 Jeff Bailey <jbailey@ubuntu.com>
* sysdeps/powerpc/powerpc32/fpu/s_lround.S: Extra strong_alias
confuses new binutils
* sysdeps/unix/sysv/linux/bind.S: bind should be the alias,
not __bind.
* sysdeps/unix/sysv/linux/getsockname.S: getsockname should
be the alias, not __getsockname.
* sysdeps/unix/sysv/linux/listen.S: listen should be the alias,
not __listen.
* sysdeps/unix/sysv/linux/setsockopt.S: setsockopt should
be the alias, not __setsockopt.
* sysdeps/unix/sysv/linux/powerpc/powerpc32/socket.S: Don't
set weak_alias if NO_WEAK_ALIAS is set.
diff -ru build-tree/glibc-2.3.5/sysdeps/powerpc/powerpc32/fpu/s_lround.S glibc-2.3.5/sysdeps/powerpc/powerpc32/fpu/s_lround.S
--- sysdeps/powerpc/powerpc32/fpu/s_lround.S 2004-06-30 18:29:12.000000000 -0400
+++ powerpc/powerpc32/fpu/s_lround.S 2005-11-17 18:35:46.000000000 -0500
@@ -88,7 +88,6 @@
b .L9
END (__lround)
-strong_alias (__lround, __lround)
weak_alias (__lround, lround)
strong_alias (__lround, __lroundf)
diff -ru build-tree/glibc-2.3.5/sysdeps/unix/sysv/linux/bind.S glibc-2.3.5/sysdeps/unix/sysv/linux/bind.S
--- sysdeps/unix/sysv/linux/bind.S 2002-05-14 20:17:25.000000000 -0400
+++ sysdeps/unix/sysv/linux/bind.S 2005-11-17 20:21:57.000000000 -0500
@@ -2,4 +2,4 @@
#define NARGS 3
#define NO_WEAK_ALIAS 1
#include <socket.S>
-weak_alias (bind, __bind)
+weak_alias (__bind, bind)
diff -ru build-tree/glibc-2.3.5/sysdeps/unix/sysv/linux/getsockname.S glibc-2.3.5/sysdeps/unix/sysv/linux/getsockname.S
--- sysdeps/unix/sysv/linux/getsockname.S 2002-05-14 20:17:58.000000000 -0400
+++ sysdeps/unix/sysv/linux/getsockname.S 2005-11-17 20:34:14.000000000 -0500
@@ -2,4 +2,4 @@
#define NARGS 3
#define NO_WEAK_ALIAS 1
#include <socket.S>
-weak_alias (getsockname, __getsockname)
+weak_alias (__getsockname, getsockname)
diff -ru build-tree/glibc-2.3.5/sysdeps/unix/sysv/linux/listen.S glibc-2.3.5/sysdeps/unix/sysv/linux/listen.S
--- sysdeps/unix/sysv/linux/listen.S 2002-05-14 20:18:37.000000000 -0400
+++ sysdeps/unix/sysv/linux/listen.S 2005-11-17 20:49:01.000000000 -0500
@@ -2,4 +2,4 @@
#define NARGS 2
#define NO_WEAK_ALIAS 1
#include <socket.S>
-weak_alias (listen, __listen)
+weak_alias (__listen, listen)
diff -ru build-tree/glibc-2.3.5/sysdeps/unix/sysv/linux/powerpc/powerpc32/socket.S glibc-2.3.5/sysdeps/unix/sysv/linux/powerpc/powerpc32/socket.S
--- sysdeps/unix/sysv/linux/powerpc/powerpc32/socket.S 2003-07-11 18:46:12.000000000 -0400
+++ sysdeps/unix/sysv/linux/powerpc/powerpc32/socket.S 2005-11-17 20:29:56.000000000 -0500
@@ -116,4 +116,6 @@
PSEUDO_END (__socket)
+#ifndef NO_WEAK_ALIAS
weak_alias (__socket, socket)
+#endif
diff -ru build-tree/glibc-2.3.5/sysdeps/unix/sysv/linux/setsockopt.S glibc-2.3.5/sysdeps/unix/sysv/linux/setsockopt.S
--- sysdeps/unix/sysv/linux/setsockopt.S 2002-05-14 20:19:09.000000000 -0400
+++ sysdeps/unix/sysv/linux/setsockopt.S 2005-11-17 20:49:48.000000000 -0500
@@ -2,4 +2,4 @@
#define NARGS 5
#define NO_WEAK_ALIAS 1
#include <socket.S>
-weak_alias (setsockopt, __setsockopt)
+weak_alias (__setsockopt, setsockopt)
^ permalink raw reply [flat|nested] 6+ messages in thread* Re: Symbol redefinition bugs
2005-11-20 12:46 ` Jeff Bailey
@ 2005-11-20 14:38 ` Andreas Schwab
2005-11-21 21:01 ` Roland McGrath
1 sibling, 0 replies; 6+ messages in thread
From: Andreas Schwab @ 2005-11-20 14:38 UTC (permalink / raw)
To: Jeff Bailey; +Cc: libc-hacker
Jeff Bailey <jbailey@raspberryginger.com> writes:
> * sysdeps/unix/sysv/linux/bind.S: bind should be the alias,
> not __bind.
Did you test that on other architectures?
Andreas.
--
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE Linux Products GmbH, Maxfeldstraße 5, 90409 Nürnberg, Germany
PGP key fingerprint = 58CA 54C7 6D53 942B 1756 01D3 44D5 214B 8276 4ED5
"And now for something completely different."
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: Symbol redefinition bugs
2005-11-20 12:46 ` Jeff Bailey
2005-11-20 14:38 ` Andreas Schwab
@ 2005-11-21 21:01 ` Roland McGrath
2005-11-22 3:55 ` Jeff Bailey
1 sibling, 1 reply; 6+ messages in thread
From: Roland McGrath @ 2005-11-21 21:01 UTC (permalink / raw)
To: Jeff Bailey; +Cc: Andreas Schwab, libc-hacker
> Done this way, readelf -a showed that the various symbols went from being
> WEAK to being GLOBAL.
Which symbols? Show the exact details you are talking about.
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: Symbol redefinition bugs
2005-11-21 21:01 ` Roland McGrath
@ 2005-11-22 3:55 ` Jeff Bailey
2005-11-22 4:33 ` Ulrich Drepper
0 siblings, 1 reply; 6+ messages in thread
From: Jeff Bailey @ 2005-11-22 3:55 UTC (permalink / raw)
To: Roland McGrath; +Cc: Andreas Schwab, libc-hacker
On lun, 2005-11-21 at 13:01 -0800, Roland McGrath wrote:
> > Done this way, readelf -a showed that the various symbols went from being
> > WEAK to being GLOBAL.
>
> Which symbols? Show the exact details you are talking about.
This is on powerpc-linux using readelf -a.
The "old" build, is CVS HEAD glibc as of 2005-11-17 using binutils
2.16.1. The "new" build is CVS HEAD glibc from today, using binutils
CVS from 20051117.
old:
617: 000d85ac 44 FUNC WEAK DEFAULT 10 socket@@GLIBC_2.0
954: 000d807c 44 FUNC WEAK DEFAULT 10 bind@@GLIBC_2.0
1014: 000d8550 52 FUNC WEAK DEFAULT 10 setsockopt@@GLIBC_2.0
1253: 000d81c0 40 FUNC WEAK DEFAULT 10 listen@@GLIBC_2.0
2184: 000d8160 44 FUNC WEAK DEFAULT 10
getsockname@@GLIBC_2.0
new:
617: 000d861c 44 FUNC WEAK DEFAULT 10 socket@@GLIBC_2.0
954: 000d80ec 44 FUNC GLOBAL DEFAULT 10 bind@@GLIBC_2.0
1014: 000d85c0 52 FUNC GLOBAL DEFAULT 10 setsockopt@@GLIBC_2.0
1253: 000d8230 40 FUNC GLOBAL DEFAULT 10 listen@@GLIBC_2.0
2184: 000d81d0 44 FUNC GLOBAL DEFAULT 10
getsockname@@GLIBC_2.0
With the patch I offered, it keeps these symbols as weak on ppc. I had
tested it against ppc and x86_64.
x86_64 also had these symbols as weak before, but i386 did not:
944: 000b90f0 57 FUNC GLOBAL DEFAULT 11 bind@@GLIBC_2.0
(etc)
If you want bind, setsockopt, listen, and getsockname as weak, I'll redo
the patch against current CVS and resubmit.
Tks,
Jeff Bailey
^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2005-11-22 4:33 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2005-11-18 19:19 Symbol redefinition bugs Andreas Schwab
2005-11-20 12:46 ` Jeff Bailey
2005-11-20 14:38 ` Andreas Schwab
2005-11-21 21:01 ` Roland McGrath
2005-11-22 3:55 ` Jeff Bailey
2005-11-22 4:33 ` Ulrich Drepper
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).