GNU Smalltalk 2.2d released

This should be the last release candidate for GNU Smalltalk 2.3.  In
fact no change should be necessary between 2.2d and 2.3.

The release is available from:
ftp://alpha.gnu.org/gnu/smalltalk/smalltalk-2.2d.tar.gz

The changes from 2.2c are:

- An exception to the GPL is included in all the virtual machine's
source files.  The exception allows Smalltalk programs to be combined
with the virtual machine, and with any library bindings that bear the
same exception, without the need to distribute the Smalltalk program
under the GNU GPL too.  This of course does not invalidate any reason
why the Smalltalk program would need to be distributed under the GNU GPL
(including using GPLed class libraries, or dynamically linking GPLed
libraries such as GNU readline).

- RegexResults objects are created from C code rather than Smalltalk
code.  This should improve speed, even though it was done mostly for
didactic reasons.

- C call-out do not necessarily create a new context.  This also causes
performance improvements.

- Brad Watson's ncurses bindings are included.

- A severe bug in ObjectMemory class>>#quit: (latent since forever, but
exposed recently) was fixed.

- A couple of packaging fixes, needed for Jochen Schmitt's Fedora RPMs.

Thanks to everybody who contributed to this release.

Paolo


_______________________________________________
help-smalltalk mailing list
[hidden email]
http://lists.gnu.org/mailman/listinfo/help-smalltalk

On Wednesday 29 November 2006 18:56, Paolo Bonzini wrote:
> This should be the last release candidate for GNU Smalltalk 2.3.  In
> fact no change should be necessary between 2.2d and 2.3.

sorry to destroy your dreams, but i just tried to compile:

FAIL: OrderedCollectionANSITest
=======================================
1 of 105 tests failed
Please report to [hidden email]
=======================================

this is a normal 32bit machine and the previous release worked fine... any
information i could give you to shed some more light on this issue?

from the ANSI.log:
OrderedCollectionANSITest: 97 run, 96 passed, 1 failed, 0 errors
Set (OrderedCollectionANSITest>>#testXaddXafterIndexX )


^manveru


_______________________________________________
help-smalltalk mailing list
[hidden email]
http://lists.gnu.org/mailman/listinfo/help-smalltalk

Uhm, the release was tested on i686-pc-linux-gnu (with and without JIT),
and on powerpc-apple-darwin (only without JIT).  So for now, you haven't
destroyed my dreams yet. :-)

Try this:

| oc |
oc := OrderedCollection with: 1 with: 2 with: 3 with: 4.
[ oc add: 0 afterIndex: 5. self halt ] on: Error do: [ :e | e return ].
[ oc add: 0 afterIndex: -1. self halt ] on: Error do: [ :e | e return ].
(oc add: 777 afterIndex: 0) printNl.
oc printNl.
(oc add: 888 afterIndex: 5) printNl.
oc printNl.
(oc add: 999 afterIndex: 1) printNl.
oc printNl!

The output should be:

777
OrderedCollection (777 1 2 3 4 )
888
OrderedCollection (777 1 2 3 4 888 )
999
OrderedCollection (777 999 1 2 3 4 888 )

Paolo


_______________________________________________
help-smalltalk mailing list
[hidden email]
http://lists.gnu.org/mailman/listinfo/help-smalltalk

On Wednesday 29 November 2006 21:33, Paolo Bonzini wrote:
my system:
2.6.18-ARCH #1 SMP PREEMPT i686 Pentium 4 CPU 3.20GHz GNU/Linux

and yes, the code runs fine:

st> | oc |
st> oc := OrderedCollection with: 1 with: 2 with: 3 with: 4.
st> [ oc add: 0 afterIndex: 5. self halt ] on: Error do: [ :e | e return ].
st> [ oc add: 0 afterIndex: -1. self halt ] on: Error do: [ :e | e return ].
st> (oc add: 777 afterIndex: 0) printNl.
st> oc printNl.
st> (oc add: 888 afterIndex: 5) printNl.
st> oc printNl.
st> (oc add: 999 afterIndex: 1) printNl.
st> oc printNl!
777
OrderedCollection (777 1 2 3 4 )
888
OrderedCollection (777 1 2 3 4 888 )
999
OrderedCollection (777 999 1 2 3 4 888 )
OrderedCollection new: 16 "<-0x4c638cc8>"


so, what is the problem here?

^manveru


_______________________________________________
help-smalltalk mailing list
[hidden email]
http://lists.gnu.org/mailman/listinfo/help-smalltalk

Is it better to drop the configure and make output on to a web page,
and then to post a url (and possibly the workaround) to the list,
or to post the whole thing to the list?

--
Stremler


_______________________________________________
help-smalltalk mailing list
[hidden email]
http://lists.gnu.org/mailman/listinfo/help-smalltalk

Stewart Stremler wrote:
> Is it better to drop the configure and make output on to a web page,
> and then to post a url (and possibly the workaround) to the list,
> or to post the whole thing to the list?

I don't mind, either way is fine.

Paolo


_______________________________________________
help-smalltalk mailing list
[hidden email]
http://lists.gnu.org/mailman/listinfo/help-smalltalk

> so, what is the problem here?

No clue.  But you can try applying the attached patches one at a time
and see which one makes the error appear

(BTW, I already destroyed my dreams myself, as the Fedora guy Jochen
Schmitt reported that the permissions are broken on some installed files
and this needs a fix in kernel/VFS.st too).

Also, if you're using the JIT, please tell me.

Paolo

_______________________________________________
help-smalltalk mailing list
[hidden email]
http://lists.gnu.org/mailman/listinfo/help-smalltalk

begin  quoting Paolo Bonzini as of Thu, Nov 30, 2006 at 09:18:14AM +0100:
> Stewart Stremler wrote:
> >Is it better to drop the configure and make output on to a web page,
> >and then to post a url (and possibly the workaround) to the list,
> >or to post the whole thing to the list?
>
> I don't mind, either way is fine.

Okay.

First attempt to compile on a Solaris 10[1] machine:

http://www.stremler.net/temp/GST/try1.txt

My initial attempt to avoid the issue was to insert "#define _AVL_H"
-- the guard define in /usr/include/sys/avl.h -- in gstpriv.h, and
then compiling again...

http://www.stremler.net/temp/GST/try2.txt

This fails with linking problems. But hey, that's different!

I've not made significant progress beyond this; randomly excluding
things from the configure change where the compile crashes with link
problems, but that's about it.

I think I might try setting CC to the Sun compiler, just to see what
happens.  Any other suggestions?

[1] lancre> uname -a
SunOS lancre 5.10 Generic_118833-24 sun4u sparc SUNW,UltraAX-i2

--
Stewart Stremler


_______________________________________________
help-smalltalk mailing list
[hidden email]
http://lists.gnu.org/mailman/listinfo/help-smalltalk

> First attempt to compile on a Solaris 10[1] machine:
>
> http://www.stremler.net/temp/GST/try1.txt
>
> My initial attempt to avoid the issue was to insert "#define _AVL_H"
> -- the guard define in /usr/include/sys/avl.h -- in gstpriv.h, and
> then compiling again...

Yes, this is completely dumb for Sun.  Including unprotected headers
called avl_destroy and so on.  Blah.

> http://www.stremler.net/temp/GST/try2.txt
>
> This fails with linking problems. But hey, that's different!

Yes.  I wonder what they mean, I'll do a web search tomorrow.

> I think I might try setting CC to the Sun compiler, just to see what
> happens.  Any other suggestions?

No, Sun compiler does not compile GST probably.

Paolo


_______________________________________________
help-smalltalk mailing list
[hidden email]
http://lists.gnu.org/mailman/listinfo/help-smalltalk

> http://www.stremler.net/temp/GST/try2.txt
>
> This fails with linking problems. But hey, that's different!

Apparently it is a bug in version 5.0 Alpha 03/27/98 of the Sun
assembler.  This bug has been fixed in the final 5.0 version of the
assembler.

Paolo


_______________________________________________
help-smalltalk mailing list
[hidden email]
http://lists.gnu.org/mailman/listinfo/help-smalltalk

> First attempt to compile on a Solaris 10[1] machine:
>
> http://www.stremler.net/temp/GST/try1.txt
>
> My initial attempt to avoid the issue was to insert "#define _AVL_H"
> -- the guard define in /usr/include/sys/avl.h -- in gstpriv.h, and
> then compiling again...

This patch should fix the problem.

--- orig/configure.ac
+++ mod/configure.ac
@@ -92,6 +92,19 @@ dnl ------------------------------ C COM
  AC_SYS_LARGEFILE
  AC_C_INLINE
  AC_C_RESTRICT
+
+dnl Test for broken solaris include file.  Should be moved to gnulib maybe?
+AC_MSG_CHECKING([for broken sys/avl.h])
+AC_PREPROC_IFELSE([#include <sys/avl.h>
+  #ifndef _AVL_H
+  would be useless anyway :-(
+  #endif], [
+  AC_MSG_RESULT(yes)
+  AC_DEFINE(_AVL_H, 1,
+    [Define to 1 if, like Solaris, your system has a sys/avl.h header that
+pollutes the name space.])], [
+  AC_MSG_RESULT(no)])
+
  GST_C_HIDDEN_VISIBILITY
  GST_C_WARN_LONG_DOUBLE
  GST_C_GOTO_VOID_P

Paolo


_______________________________________________
help-smalltalk mailing list
[hidden email]
http://lists.gnu.org/mailman/listinfo/help-smalltalk

>>> This fails with linking problems. But hey, that's different!
>> Apparently it is a bug in version 5.0 Alpha 03/27/98 of the Sun
>> assembler.  This bug has been fixed in the final 5.0 version of the
>> assembler.
>
> Hm. Okay. It seems that's not something they include in the standard
> patches then. Joy.

But it might also be a gas bug.
http://sources.redhat.com/bugzilla/show_bug.cgi?id=144

Reportedly the compilation should be fine

1. with Sun as + Sun ld,
2. with GNU as + GNU ld,
3. with GNU as + Sun ld + Sun grep (??)
4. passing to make CFLAGS="-gstabs -O2"

Paolo


_______________________________________________
help-smalltalk mailing list
[hidden email]
http://lists.gnu.org/mailman/listinfo/help-smalltalk

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hi,

I've installed gst with JIT enabled and this time(2.2d) make check does
passed all test cases. But with preemption enabled, it does not work
which I don't care.

Thanks.

Paolo Bonzini wrote:

>
> Also, if you're using the JIT, please tell me.
>
> Paolo
>

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.2.2 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFFb4HLQqspS1+XJHgRAg4EAJ9YFcETmBW4rbeyesoHturljvzxHwCfRUPo
dwYnCqW/P8ea0ewxIVUV/Q0=
=x4gQ
-----END PGP SIGNATURE-----


_______________________________________________
help-smalltalk mailing list
[hidden email]
http://lists.gnu.org/mailman/listinfo/help-smalltalk

> I've installed gst with JIT enabled and this time(2.2d) make check does
> passed all test cases. But with preemption enabled, it does not work
> which I don't care.

Thanks.  Preemption is not a supported feature anyway (unlike the JIT).

Paolo



_______________________________________________
help-smalltalk mailing list
[hidden email]
http://lists.gnu.org/mailman/listinfo/help-smalltalk