Trouble to build gnu-smalltlak-2.3.4 (test suite failed)

ello,

I have try to create a RPM package for gnu-smalltalk-2.3.4.

Unfortunately two tests from the test suite are failed.

I have uploaded the following files:
http://www.herr-schmit.de/pub/gnu-smalltalk/testsuite.log
http://www.herr-schmitt.de/pub/gnu-smalltalk/gnu-smalltalk.spec
http://www.herr-schmitt.de/pub/gnu-smalltalk/gnu-smalltalk-2.3.4-1.fc8.src.rpm 


The Build was done on the x86_64 system with Fedora Core 6 installed.

It will be nice, if anyoune can help me.

Best Regards:

Jochen Schmitt


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

Jochen Schmitt wrote:
For geometry.st, the problem is that sqrt is computed with a slightly
different precision on x86_64, apparently.  I can fix it by using a
Pythagorean triple (i.e. making sure the sqrt is integer).  It's not a
big deal, I guess you can patch out the failing test (both the .st and
.ok files); Thomas, you'll probably have the same problem on Debian.

For GDBM, I'll look at it later.

Paolo


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

> For geometry.st, the problem is that sqrt is computed with a slightly
> different precision on x86_64, apparently.  I can fix it by using a
> Pythagorean triple (i.e. making sure the sqrt is integer).  It's not a
> big deal, I guess you can patch out the failing test (both the .st and
> .ok files); Thomas, you'll probably have the same problem on Debian.

Here's a patch.

Paolo

* looking for [hidden email]--2004b/smalltalk--stable--2.3--patch-54 to compare with
* comparing to [hidden email]--2004b/smalltalk--stable--2.3--patch-54
M  tests/geometry.ok
M  tests/geometry.st

* modified files

--- orig/tests/geometry.ok
+++ mod/tests/geometry.ok
@@ -36,10 +36,10 @@ Execution begins...
 returned value is Point new "<0>"
 
 Execution begins...
-'A dist: B = '136.0147050873544
+'A dist: B = '104.0
 'C dotProduct: D = '20000
 'C grid: D = '150@250
-'C normal = '-0.8320502943378438@0.5547001962252292
+'E normal * 5 = '-4@3
 'C truncatedGrid: D = '150@200
 '175@300 transpose = '300@175
 returned value is Point new "<0>"


--- orig/tests/geometry.st
+++ mod/tests/geometry.st
@@ -112,19 +112,20 @@ myAreasOutside: a
 
 "point functions"
 
-| A B C D |
+| A B C D E |
   A := 45@230.
-  B := 175@270.
+  B := 141@270.
   C := 160@240.
   D := 50@50.
+  E := 3@4.
   'A dist: B = ' print.
   (A dist: B) printNl.
   'C dotProduct: D = ' print.
   (C dotProduct: D) printNl.
   'C grid: D = ' print.
   (C grid: D) printNl.
-  'C normal = ' print.
-  (C normal) printNl.
+  'E normal * 5 = ' print.
+  (E normal * 5) rounded printNl.
   'C truncatedGrid: D = ' print.
   (C truncatedGrid: D) printNl.
   '175@300 transpose = ' print.




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

Hello Paolo,

On Tue, May 29, 2007 at 11:08:08AM +0200, Paolo Bonzini wrote:
>
> >For geometry.st, the problem is that sqrt is computed with a slightly
> >different precision on x86_64, apparently.  I can fix it by using a
> >Pythagorean triple (i.e. making sure the sqrt is integer).  It's not a
> >big deal, I guess you can patch out the failing test (both the .st and
> >.ok files); Thomas, you'll probably have the same problem on Debian.

Possibly, but since my upload was broken it FTBFS'ed anyway on amd64
arch.

> Here's a patch.

I'll give it a try, thanks!

Regards,

Thomas


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

> For GDBM, I'll look at it later.

It's a stupid 64-bit cleanliness issue...

--- ../gst-stable/examples/gdbm-c.st 2007-05-24 22:11:07.000000000 +0200
+++ examples/gdbm-c.st 2007-05-29 13:59:34.000000000 +0200
@@ -40,7 +40,7 @@
  CStruct subclass: #DatumStruct
         declaration: #(         "typedef struct {       "
             (#dPtr (#ptr #char))"       char *dptr;     "
-           (#dSize #long)      "       int   dsize;    "
+           (#dSize #int)       "       int   dsize;    "
         )                       "      } datum;         "
         classVariableNames: ''
         poolDictionaries: ''

Paolo


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

Indeed, I see this as well on x86_64. The patch fixes the geometry.st
test problem.

Thanks,

Thomas


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