Re: [parisc-linux] Re: hppa glibc 2.3.3

Re: [parisc-linux] Re: hppa glibc 2.3.3

[Joel Soete]

> -- Original Message --
> Date: Thu, 7 Oct 2004 18:39:40 -0400
> From: Carlos O'Donell <carlos@baldric.uwo.ca>
> To: Joel Soete <soete.joel@tiscali.be>
> Cc: parisc-linux@parisc-linux.org
> Subject: Re: [parisc-linux] Re: hppa glibc 2.3.3
>
>
> > I look forward and encounter:
> > <http://sources.redhat.com/ml/glibc-cvs/2004-q4/msg00074.html>
> >
> > It seems an evidence to just change errno to rtld_errno symbol for th=
ose
> > other arch but for hppa I realy didn't encounter this variable name,
what
> > did I missed (certanly a macro but which one?)
>
> Please please please, if you are going to quote other arches, read the
> code before commenting.
>
> i386 code is ancient and manipulates errno directly. We use the generic=

> code which is always kept up to date. Notice how alpha and ia64 didn't
> need to be fixed? Neither do we :)
>
Sorry to be so lazy (just a bit tired, appologies), here are the ref to
likewise modifications:
<http://sources.redhat.com/cgi-bin/cvsweb.cgi/libc/sysdeps/unix/sysv/linu=
x/ia64/sysdep.S.diff?cvsroot=3Dglibc&r1=3D1.8&r2=3D1.9>
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
RCS file: /cvs/glibc/libc/sysdeps/unix/sysv/linux/ia64/sysdep.S,v
retrieving revision 1.8
retrieving revision 1.9
diff -u -r1.8 -r1.9
--- libc/sysdeps/unix/sysv/linux/ia64/sysdep.S	2004/05/04 16:46:06	1.8
+++ libc/sysdeps/unix/sysv/linux/ia64/sysdep.S	2004/10/04 20:59:36	1.9
@@ -30,7 +30,7 @@
 	 * be fine.  Otherwise, we would have to first load the global
 	 * pointer register from __gp.
 	 */
-	addl	r2=3D@gprel(errno),gp
+	addl	r2=3D@gprel(rtld_errno),gp
 	;;
 	st4	[r2]=3Dr8
 	mov	r8=3D-1
<http://sources.redhat.com/cgi-bin/cvsweb.cgi/libc/sysdeps/unix/alpha/sys=
dep.h.diff?cvsroot=3Dglibc&r1=3D1.25&r2=3D1.26>
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
RCS file: /cvs/glibc/libc/sysdeps/unix/alpha/sysdep.h,v
retrieving revision 1.25
retrieving revision 1.26
diff -u -r1.25 -r1.26
--- libc/sysdeps/unix/alpha/sysdep.h	2004/09/26 17:40:31	1.25
+++ libc/sysdeps/unix/alpha/sysdep.h	2004/10/04 20:59:38	1.26
@@ -82,7 +82,7 @@
 #if RTLD_PRIVATE_ERRNO
 # define SYSCALL_ERROR_LABEL	$syscall_error
 # define SYSCALL_ERROR_HANDLER			\
-	stl	v0, errno(gp)	!gprel;		\
+	stl	v0, rtld_errno(gp)	!gprel;	\
 	lda	v0, -1;				\
 	ret
 #elif defined(PIC)

pfff, 'thuth is out there' :)

Thanks for your attention,
    Joel


-------------------------------------------------------------------------=
--
Tiscali ADSL GO, 29,50 Euro/mois pendant toute une ann=E9e, profitez-en..=
..
http://reg.tiscali.be/adsl/welcome.asp?lg=3DFR




_______________________________________________
parisc-linux mailing list
parisc-linux@lists.parisc-linux.org
http://lists.parisc-linux.org/mailman/listinfo/parisc-linux

Re: [parisc-linux] Re: hppa glibc 2.3.3

[Carlos O'Donell]

On Fri, Oct 08, 2004 at 08:00:53AM +0200, Joel Soete wrote:
> Sorry to be so lazy (just a bit tired, appologies), here are the ref to
> pfff, 'thuth is out there' :)

The only problem might exist if __errno_location does not return the
proper address of rltd_errno when RTLD_PRIVATE_ERRNO is defined. It
should. If it doesn't then our definition of
SYSCALL_ERRNO_HANDLER should change in that situation (away from
DEFAULT_SYSCALL_ERROR_HANDLER which is set to __errno_location).

All we want is that SYSCALL_ERRNO_HANDLER act as a function that returns
the address of an int, into which we write the syscall error value.

Read the code in:

libc/sysdeps/generic/errno-loc.c
libc/sysdeps/generic/errno.c
libc/sysdeps/unix/sysv/linux/hppa/sysdep.h

The last file defines the syscall wrappers and their usage of
SYSCALL_ERRNO_HANDLER.

c.


_______________________________________________
parisc-linux mailing list
parisc-linux@lists.parisc-linux.org
http://lists.parisc-linux.org/mailman/listinfo/parisc-linux

Re: [parisc-linux] Re: hppa glibc 2.3.3

[Randolph Chung]

> The only problem might exist if __errno_location does not return the
> proper address of rltd_errno when RTLD_PRIVATE_ERRNO is defined. It
> should. If it doesn't then our definition of
> SYSCALL_ERRNO_HANDLER should change in that situation (away from
> DEFAULT_SYSCALL_ERROR_HANDLER which is set to __errno_location).

is there a fix for this? i was building glibc from cvs today to look at
the _dl_debug_state problem and ran into this too.... (also the problem
with the missing NULL defn for waitid)

The problem here seems to be that we are using the generic dl-sysdep.h 
and errno.c/errno-loc.c definitions, and when building rtld-errno-loc.o,
__errno_location is coded to return &errno and not &rtld_errno.

i hacked this in my build but not sure if this is the proper solution --
should we propose a change to the generic errno-loc.c? or do we need our
own? or am i on the wrong track?

i'm using something like this:

Index: linuxthreads/sysdeps/pthread/errno-loc.c
===================================================================
RCS file: /cvs/glibc/libc/linuxthreads/sysdeps/pthread/errno-loc.c,v
retrieving revision 1.3
diff -u -p -r1.3 errno-loc.c
--- linuxthreads/sysdeps/pthread/errno-loc.c	1 Apr 2003 00:52:40 -0000	1.3
+++ linuxthreads/sysdeps/pthread/errno-loc.c	11 Oct 2004 04:41:14 -0000
@@ -24,9 +24,13 @@
 #include <sysdep-cancel.h>
 
 #if ! USE___THREAD
+#if RTLD_PRIVATE_ERRNO
+extern int rtld_errno;
+#else
 #undef errno
 extern int errno;
 #endif
+#endif
 
 int *
 #if ! USE___THREAD
@@ -41,6 +45,10 @@ __errno_location (void)
       return LIBC_THREAD_GETMEM (self, p_errnop);
     }
 #endif
+#if RTLD_PRIVATE_ERRNO
+  return &rtld_errno;
+#else
   return &errno;
+#endif
 }
 libc_hidden_def (__errno_location)

randolph
-- 
Randolph Chung
Debian GNU/Linux Developer, hppa/ia64 ports
http://www.tausq.org/
_______________________________________________
parisc-linux mailing list
parisc-linux@lists.parisc-linux.org
http://lists.parisc-linux.org/mailman/listinfo/parisc-linux

Re: [parisc-linux] Re: hppa glibc 2.3.3

[Carlos O'Donell]

> is there a fix for this? i was building glibc from cvs today to look at
> the _dl_debug_state problem and ran into this too.... (also the problem
> with the missing NULL defn for waitid)

I have a fix for the missing NULL defn.

> The problem here seems to be that we are using the generic dl-sysdep.h 
> and errno.c/errno-loc.c definitions, and when building rtld-errno-loc.o,
> __errno_location is coded to return &errno and not &rtld_errno.

Yeah, I saw this yesterday while going through the tree.
 
> i hacked this in my build but not sure if this is the proper solution --
> should we propose a change to the generic errno-loc.c? or do we need our
> own? or am i on the wrong track?

IMO your hack is correct, a lot of other arches just 
write their own version of this file and change it 
depending on the defines.

I get a warning when I compile this and it doesn't 
fail the build for me.


The patch looks good.

As for the missing _dl_debug_state, I'm really starting
to think it's a linker issue.

c.

_______________________________________________
parisc-linux mailing list
parisc-linux@lists.parisc-linux.org
http://lists.parisc-linux.org/mailman/listinfo/parisc-linux

[parisc-linux] Re: hppa glibc 2.3.3

[Carlos O'Donell]

On Tue, Oct 05, 2004 at 08:47:56AM -0500, Elliott Potter wrote:
> > This was a bug fixed upstream, where the nscd thread daemon was using
> > NPTL related syscalls when it shouldn't. Pull cvs again and that should
> > fit the problem.
> 
> Just an update, this seemed to work just fine.  I'm going to play with
> it a bit in my chroot environment and see how it works.

Perfect. I'm working on a variant where I've forward ported the debian
specific patches into upstream head.

Please report back if you have any problems, or if you have success :)

e.g. Try installing Xnest + gnome, and run some stuff remotely?

c.

_______________________________________________
parisc-linux mailing list
parisc-linux@lists.parisc-linux.org
http://lists.parisc-linux.org/mailman/listinfo/parisc-linux

Re: [parisc-linux] Re: hppa glibc 2.3.3

[Carlos O'Donell]

On Thu, Oct 07, 2004 at 01:56:09PM +0200, Joel Soete wrote:
> Since the Oct 1, I try desperately to complete the glibc cvs build but without
> success:

a. Build an 'unstable' chroot. Enter the chroot.
b. apt-get build-dep glibc
c. Pull down glibc cvs head.
d. Apply the latest hppa patch set.
e. Build glibc.

I suggest the chroot method because it seems that you might have
problems with your environment. In particular I always try a chroot
build if something seems funny.

Cheers,
Carlos.


_______________________________________________
parisc-linux mailing list
parisc-linux@lists.parisc-linux.org
http://lists.parisc-linux.org/mailman/listinfo/parisc-linux

Re: [parisc-linux] Re: hppa glibc 2.3.3

[Joel Soete]

> -- Original Message --
> Date: Thu, 7 Oct 2004 11:29:00 -0400
> From: Carlos O'Donell <carlos@baldric.uwo.ca>
> To: Joel Soete <soete.joel@tiscali.be>
> Cc: parisc-linux@parisc-linux.org
> Subject: Re: [parisc-linux] Re: hppa glibc 2.3.3
> 
> 
> On Thu, Oct 07, 2004 at 01:56:09PM +0200, Joel Soete wrote:
> > Since the Oct 1, I try desperately to complete the glibc cvs build bu=
t
> without
> > success:
> 
> a. Build an 'unstable' chroot. Enter the chroot.
> b. apt-get build-dep glibc
> c. Pull down glibc cvs head.
> d. Apply the latest hppa patch set.
> e. Build glibc.
> 
> I suggest the chroot method because it seems that you might have
> problems with your environment. In particular I always try a chroot
> build if something seems funny.
> 
> Cheers,
> Carlos.
> 
mmm good idea but with exactely the same env (ie same binutils, same gcc,=

same kernel, same nfs server and client) I reach to build glibc cvs 2004
09 30 + your patch with same index (even complet test) without pb but for=

later cvs (ie 2004 10 0[1234567] it failed.

Fwiw I check and configure script: no change?
Is there some include file which could change its behaviour?

That said, I can force ranlib with --enable-ranlib but unfortunately the
build  just failed:(
I try to have a look in precompiled elf/rtld.i but I didn't yet find wher=
e
errno is requested in set_dp?
(ooh iirc something was changed regarfing __errno in upstream:
<http://sources.redhat.com/cgi-bin/cvsweb.cgi/libc/include/errno.h.diff?c=
vsroot=3Dglibc&only_with_tag=3Dfedora-branch&r1=3D1.11&r2=3D1.11.2.1>
i.e.:
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
RCS file: /cvs/glibc/libc/include/errno.h,v
retrieving revision 1.11
retrieving revision 1.11.2.1
diff -u -r1.11 -r1.11.2.1
--- libc/include/errno.h	2002/12/31 20:33:33	1.11
+++ libc/include/errno.h	2004/10/05 08:25:13	1.11.2.1
@@ -14,8 +14,8 @@
    so a single (hidden) global variable is all it needs.  */
 
 #  undef  errno
-#  define errno errno		/* For #ifndef errno tests.  */
-extern int errno attribute_hidden;
+#  define errno rtld_errno
+extern int rtld_errno attribute_hidden;

=3D=3D=3D=3D=3D=3D=3D
So certainly where I have to check?

Thanks,
    Joel


-------------------------------------------------------------------------=
--
Tiscali ADSL GO, 29,50 Euro/mois pendant toute une ann=E9e, profitez-en..=
..
http://reg.tiscali.be/adsl/welcome.asp?lg=3DFR




_______________________________________________
parisc-linux mailing list
parisc-linux@lists.parisc-linux.org
http://lists.parisc-linux.org/mailman/listinfo/parisc-linux

Re: [parisc-linux] Re: hppa glibc 2.3.3

[Joel Soete]

Carlos, 

I need more help
> <http://sources.redhat.com/cgi-bin/cvsweb.cgi/libc/include/errno.h.diff=
?cvsroot=3Dglibc&only_with_tag=3Dfedora-branch&r1=3D1.11&r2=3D1.11.2.1>
> i.e.:
> =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
> RCS file: /cvs/glibc/libc/include/errno.h,v
> retrieving revision 1.11
> retrieving revision 1.11.2.1
> diff -u -r1.11 -r1.11.2.1
> --- libc/include/errno.h	2002/12/31 20:33:33	1.11
> +++ libc/include/errno.h	2004/10/05 08:25:13	1.11.2.1
> @@ -14,8 +14,8 @@
>     so a single (hidden) global variable is all it needs.  */
> 
>  #  undef  errno
> -#  define errno errno		/* For #ifndef errno tests.  */
> -extern int errno attribute_hidden;
> +#  define errno rtld_errno
> +extern int rtld_errno attribute_hidden;
> 
> =3D=3D=3D=3D=3D=3D=3D
> So certainly where I have to check?
> 
I look forward and encounter:
<http://sources.redhat.com/ml/glibc-cvs/2004-q4/msg00074.html>

It seems an evidence to just change errno to rtld_errno symbol for those
other arch but for hppa I realy didn't encounter this variable name, what=

did I missed (certanly a macro but which one?)

Thanks again,
    Joel


-------------------------------------------------------------------------=
--
Tiscali ADSL GO, 29,50 Euro/mois pendant toute une ann=E9e, profitez-en..=
..
http://reg.tiscali.be/adsl/welcome.asp?lg=3DFR




_______________________________________________
parisc-linux mailing list
parisc-linux@lists.parisc-linux.org
http://lists.parisc-linux.org/mailman/listinfo/parisc-linux

Re: [parisc-linux] Re: hppa glibc 2.3.3

[Carlos O'Donell]

> I look forward and encounter:
> <http://sources.redhat.com/ml/glibc-cvs/2004-q4/msg00074.html>
> 
> It seems an evidence to just change errno to rtld_errno symbol for those
> other arch but for hppa I realy didn't encounter this variable name, what
> did I missed (certanly a macro but which one?)

Please please please, if you are going to quote other arches, read the
code before commenting.

i386 code is ancient and manipulates errno directly. We use the generic
code which is always kept up to date. Notice how alpha and ia64 didn't
need to be fixed? Neither do we :)

c.

_______________________________________________
parisc-linux mailing list
parisc-linux@lists.parisc-linux.org
http://lists.parisc-linux.org/mailman/listinfo/parisc-linux