[GNU Smalltalk 3.2.91] testsuite: 7 8 31 40 59 63 69 70 72 74 79 83 86 89 90 92 94 104 118 124 failed

Addendum:

I meant to attach the output of the "otool -L gst" for each of the gst
instances on my system:


- pob

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

An update to this - that won't surprise you!

I downloaded 3.2.91 source to my old Ubuntu 14.04 system and compiled it
there, and 132 of the tests ran successfully to completion, so no issue
with the released source, as expected.

The simple tests that fail on OSX, viz

> 9.0 sqrt
> 7.0 sqrt
> 180 degreesToRadians

complete successfully in the Linux REPL.

- pob



On Tue, Nov 7, 2017 at 12:45 AM, Patrick O'Brien <[hidden email]> wrote:
_______________________________________________
help-smalltalk mailing list
[hidden email]
https://lists.gnu.org/mailman/listinfo/help-smalltalk

> On 19. Nov 2017, at 06:26, Patrick O'Brien <[hidden email]> wrote:

Hey!


> An update to this - that won't surprise you!
>
> I downloaded 3.2.91 source to my old Ubuntu 14.04 system and compiled it
> there, and 132 of the tests ran successfully to completion, so no issue
> with the released source, as expected.

could you include the testsuite.log as well? 3.2 working and 3.2.91 is a
bit suspicious. Do you have this[1] commit? Could you try master?

holger

[1] http://git.savannah.gnu.org/cgit/smalltalk.git/commit/libgst/interp.inl?id=72ada189aba0283c551ead16635c1983968080b8


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

> On 28. Nov 2017, at 19:00, Patrick O'Brien <[hidden email]> wrote:
>
> Holger, thanks for the reply.


Hey!


> Let me know if you need more information (as I may be making invalid assumptions based on what I have written here.)


could you try "master". As you are using "3.2.91" tarball and a new compiler I think you run into the situation the compiler optimizing the integer overflow check away. The below is fixing this.

regards

        holger

> [1] http://git.savannah.gnu.org/cgit/smalltalk.git/commit/libgst/interp.inl?id=72ada189aba0283c551ead16635c1983968080b8
>
>
> <pob_test1.zip>


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

> On 3. Dec 2017, at 22:47, Patrick O'Brien <[hidden email]> wrote:
>
> Holger,

Hi!

it does look like the compiler "optimizing" the overflow check away. Please try my patch and make sure that your clang is new enough and that configure is ran. Otherwise go to the git version and test there. If that fails please include the git commit you are using, and the output of "cc --version".

have a nice weekend

        holger


> Last week, before I replied to your first email, I did apply that patch you sent me to the current 3.2.91 tarball, and that change did NOT improve the test results on OSX Sierra.
>
> *****
>
> I will by the middle of this coming week use the HEAD of the master branch, compile locally, and test with that, and then report back.


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