zsh-workers
 help / color / mirror / code / Atom feed
* Re: zsh problems on sparc: fixed
       [not found] <Pine.GSO.4.10.10001181434300.20378-100000@leeor.math.technion.ac.il>
@ 2000-01-18 19:59 ` Peter Stephenson
  2000-01-18 20:13   ` Chmouel Boudjnah
  0 siblings, 1 reply; 4+ messages in thread
From: Peter Stephenson @ 2000-01-18 19:59 UTC (permalink / raw)
  To: Zsh hackers list

"Zvi Har'El" wrote:
> On 17 Jan 2000, Francis GALIEGUE wrote:
> 
> > 
> > The solution: --disabe-lfs in ./configure. Strange. Patch (the
> > utility) also has problems with large fs support on sparc... Dunno why
> > but now it's solved and I can at least use my favorite shell on these
> > marvellous toys :)
> > 
> > 
> This was my first solution, but upgrading my gcc (now I have
> gcc version 2.95.2 19991024 (release)) did the trick without disabling lfs!

I've put this in the MACHINES file, but it would be nice to know a bit more
about what Solaris systems and gcc versions are and aren't affected.  I
wouldn't completely rule out an alignment problem in zsh, but without more
detailed bug tracking it's impossible to say.  Although there seems to be
some argument about whether it does fix it.

Index: Etc/MACHINES
===================================================================
RCS file: /home/pws/CVSROOT/projects/zsh/Etc/MACHINES,v
retrieving revision 1.1.1.1
diff -u -r1.1.1.1 MACHINES
--- Etc/MACHINES	1999/11/28 17:42:27	1.1.1.1
+++ Etc/MACHINES	2000/01/18 19:11:20
@@ -131,3 +131,9 @@
 	To avoid this, make sure you compile zsh without any reference
 	to /usr/ucblib in your LD_LIBRARY_PATH.  You can easily do this
 	by just unsetting LD_LIBRARY_PATH before building zsh.
+
+	Problems have been reported using --enable-lfs (the default) to
+	enable large file system and integer support on Solaris 2 with gcc.
+	Apparently upgrading to gcc version 2.95.2 fixes this.  If this
+	is not feasible, configure with --disable-lfs.  We would be
+	grateful for more detailed information.

-- 
Peter Stephenson <pws@pwstephenson.fsnet.co.uk>


^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: zsh problems on sparc: fixed
  2000-01-18 19:59 ` zsh problems on sparc: fixed Peter Stephenson
@ 2000-01-18 20:13   ` Chmouel Boudjnah
  2000-01-18 20:42     ` Peter Stephenson
  0 siblings, 1 reply; 4+ messages in thread
From: Chmouel Boudjnah @ 2000-01-18 20:13 UTC (permalink / raw)
  To: Peter Stephenson; +Cc: Zsh hackers list

Peter Stephenson <pws@pwstephenson.fsnet.co.uk> writes:

> I've put this in the MACHINES file, but it would be nice to know a bit more
> about what Solaris systems 

no solaris installed, linux :

(chmou@e250)[~]-% cat /proc/cpuinfo 
cpu             : TI UltraSparc II  (BlackBird)
fpu             : UltraSparc II integrated FPU
promlib         : Version 3 Revision 7
prom            : 3.7.3
type            : sun4u
ncpus probed    : 2
ncpus active    : 2
Cpu0Bogo        : 494.79
Cpu1Bogo        : 494.79
MMU Type        : Spitfire
State:
CPU0:           online
CPU1:           online
(chmou@e250)[~]-% rpm -q glibc
glibc-2.1.2-9mdk
(chmou@e250)[~]-% uname -a     
Linux e250.mandrakesoft.com 2.2.12-42smp #1 SMP Fri Nov 12 12:33:56 EST 1999 sparc64 unknown
(chmou@e250)[~]-% 

> and gcc versions are and aren't affected.

in our side we get bus error with :

egcs-1.1.2
gcc2.95.2

the _only_ solution is to disable lfs.

                                  --Chmouel


^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: zsh problems on sparc: fixed
  2000-01-18 20:13   ` Chmouel Boudjnah
@ 2000-01-18 20:42     ` Peter Stephenson
  2000-01-19 19:21       ` Clint Adams
  0 siblings, 1 reply; 4+ messages in thread
From: Peter Stephenson @ 2000-01-18 20:42 UTC (permalink / raw)
  To: Zsh hackers list

Chmouel Boudjnah wrote:
> Peter Stephenson <pws@pwstephenson.fsnet.co.uk> writes:
> 
> > I've put this in the MACHINES file, but it would be nice to know a bit more
> > about what Solaris systems 
> 
> no solaris installed, linux :

Ah, I've been chasing up a gum tree looking on Solaris 5.6, then, where
there's no problem.  It's presumably the same problem reported by Clint, or
a version of it.  If that's the case, try moving the 'data' element of
'struct qual' around line 108 in glob.c to be the first element of the
struct.  It shouldn't be necessary with gcc, but it's the first workaround
that comes to mind.  To be on the safe side, better rewrite struct gmatch
too --- here's a patch.  The compiler's supposed to handle this if the
struct itself is properly aligned anyway, so it's still possible the
memory allocator isn't returning 64-bit aligned quantities for some
reason which would need further investigation.

> Date: Thu, 30 Dec 1999 17:43:30 -0500
> From: Clint Adams <schizo@debian.org>
> Subject: reproducible bus error
> 
> On UltraSPARC Linux zsh (compiled 32-bit) experiences a bus error
> when trying to execute this line (1320) in Src/glob.c:
> 
> qn->data = data;
> 
> I notice that off_t data has ALIGN64.  Is that relevant or
> is something else going on here?

Index: Src/glob.c
===================================================================
RCS file: /home/pws/CVSROOT/projects/zsh/Src/glob.c,v
retrieving revision 1.2
diff -u -r1.2 glob.c
--- Src/glob.c	1999/12/03 19:12:10	1.2
+++ Src/glob.c	2000/01/18 20:37:41
@@ -41,13 +41,13 @@
 typedef struct gmatch *Gmatch; 
 
 struct gmatch {
-    char *name;
     off_t size ALIGN64;
+    off_t _size ALIGN64;
+    char *name;
     long atime;
     long mtime;
     long ctime;
     long links;
-    off_t _size ALIGN64;
     long _atime;
     long _mtime;
     long _ctime;
@@ -102,10 +102,10 @@
 typedef int (*TestMatchFunc) _((char *, struct stat *, off_t, char *));
 
 struct qual {
+    off_t data ALIGN64;		/* Argument passed to function               */
     struct qual *next;		/* Next qualifier, must match                */
     struct qual *or;		/* Alternative set of qualifiers to match    */
     TestMatchFunc func;		/* Function to call to test match            */
-    off_t data ALIGN64;		/* Argument passed to function               */
     int sense;			/* Whether asserting or negating             */
     int amc;			/* Flag for which time to test (a, m, c)     */
     int range;			/* Whether to test <, > or = (as per signum) */
-- 
Peter Stephenson <pws@pwstephenson.fsnet.co.uk>


^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: zsh problems on sparc: fixed
  2000-01-18 20:42     ` Peter Stephenson
@ 2000-01-19 19:21       ` Clint Adams
  0 siblings, 0 replies; 4+ messages in thread
From: Clint Adams @ 2000-01-19 19:21 UTC (permalink / raw)
  To: Peter Stephenson; +Cc: Zsh hackers list

> Ah, I've been chasing up a gum tree looking on Solaris 5.6, then, where
> there's no problem.  It's presumably the same problem reported by Clint, or
> a version of it.  If that's the case, try moving the 'data' element of

FWIW, what was the problem for me was solved by the 64-bit padding
issue whose thread had subjects of "Glob problem with memory"
and "64_bit logical is broken"

I have no problems on UltraSPARCs and zsh compiled with
gcc version 2.95.2 19991109 and glibc 2.1.2.  I have not
applied 9357.


^ permalink raw reply	[flat|nested] 4+ messages in thread

end of thread, other threads:[~2000-01-19 19:22 UTC | newest]

Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
     [not found] <Pine.GSO.4.10.10001181434300.20378-100000@leeor.math.technion.ac.il>
2000-01-18 19:59 ` zsh problems on sparc: fixed Peter Stephenson
2000-01-18 20:13   ` Chmouel Boudjnah
2000-01-18 20:42     ` Peter Stephenson
2000-01-19 19:21       ` Clint Adams

Code repositories for project(s) associated with this public inbox

	https://git.vuxu.org/mirror/zsh/

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 NNTP newsgroup(s).