Re: [Help-smalltalk] GNU Smalltalk 2.2e released

[Michael Fellinger]

On Monday 04 December 2006 18:19, Paolo Bonzini wrote:
> Given the presence of a few serious bugs in gst 2.2e (a few of which
> regressions), I spun a further pre-release, which is available at
>
> ftp://alpha.gnu.org/gnu/smalltalk/smalltalk-2.2e.tar.gz
>
> All bugs reported have been fixed, with the exception of the
> "mysterious" OrderedCollectionANSITest failure.
>
> In particular, this includes:
> - method verification failure on asynchronous C call-outs
> - #roundTo: and #rounded implemented erroneously
> - GMP interface not 64-bit clean
> - occurrences of DESTDIR in the installed image (Jochen, please confirm
> that we are fine now!)
> - failure to build from source on Solaris 10
>
> I expect 2.3 to be released end of this week.
>
> Paolo

I'm sorry that i haven't found the time to inspect the prior bug and apply the 
patches (also due to some problem with applying them, patch mostly said they 
were applied already? | i'm really not good at handling manual diffs... i'm a 
lazy SCM-user)
however, i just tried to build/test the current version again and this time i 
get 3/105 errors:


FAIL: ScaledDecimalANSITest
ScaledDecimalANSITest: 70 run, 69 passed, 1 failed, 0 errors
Set (ScaledDecimalANSITest>>#testXroundToX )

FAIL: FractionANSITest
FractionANSITest: 70 run, 69 passed, 1 failed, 0 errors
Set (FractionANSITest>>#testXroundToX )

FAIL: FloatANSITest
FloatANSITest: 81 run, 80 passed, 1 failed, 0 errors
Set (FloatANSITest>>#testXroundToX )


This is on a standard 32bit-architecture (i686)
I can only guess that it has something to do with the tweaking of 
Float-round-stuff... (very wild guess ;) however, that is what i've found.
it would be very helpful if the testcases gave a little bit more informative 
output, like where the test fails, what the input and output was and what the 
error is... right now all i can tell is that there is a problem. :|

Well, hope that helps, i'll now also test x86_64:

here we got the three ANSI problems from 32bit but also a fourth:


./run-test: line 9:   452 Segmentation fault      $top_builddir/gst -rI 
$top_builddir/gst.im ${base}.st >$build_base.log 2>&1
FAIL: intmath.st


the instructions with which i've built this are:

./configure --prefix=/usr --disable-emacs --infodir=/usr/share/info
make || return 1
make check || return 1
make DESTDIR=$startdir/pkg/ install

nothing out of the ordinary i hope :)
oh, and i couldn't find whether this intmath.st left a log somewhere, i 
suppose it doesn't.
On the bright side, there is no problem with OrderedCollectionANSITest 
anymore.

oh, and the system i'm running is Archlinux, in case you are interested, the 
PKGBUILD can be found at:
http://aur.archlinux.org/packages/smalltalk/smalltalk/PKGBUILD

Please tell me again if you need me to run something or if i made some obvious 
mistake... i'm quite sure i don't compile with the JIT, which i'd like to do 
though... will check the ./configure --help later.

^manveru