Very strange bug with Cog VM

Previous Topic Next Topic
 
classic Classic list List threaded Threaded
7 messages Options
Reply | Threaded
Open this post in threaded view
|

Very strange bug with Cog VM

Stéphane Rollandin
hello,

I stumbled upon this:

1) open an image with Cog

2) open a Worspace, evaluate
        0.0 closeTo: 0
-> true (so far, so good)

3) now evaluate
        (1.0 closeTo: 1) and: [0.0 closeTo: 0]
-> false !

4) evaluate again
        0.0 closeTo: 0
-> false !

5) save the image and quit.

6) reopen the image, start again at 2...


???

Stef



Reply | Threaded
Open this post in threaded view
|

Re: Very strange bug with Cog VM

Eliot Miranda-2
Hi Stéphane,

cannot reproduce in my 4.1 derived updated image

#('0.0 closeTo: 0' '(1.0 closeTo: 1) and: [0.0 closeTo: 0]' '0.0 closeTo: 0') collect: [:s| Compiler evaluate: s] #(true true true)


2010/6/30 Stéphane Rollandin <[hidden email]>
hello,

I stumbled upon this:

1) open an image with Cog

2) open a Worspace, evaluate
       0.0 closeTo: 0
-> true (so far, so good)

3) now evaluate
       (1.0 closeTo: 1) and: [0.0 closeTo: 0]
-> false !

4) evaluate again
       0.0 closeTo: 0
-> false !

5) save the image and quit.

6) reopen the image, start again at 2...


???

Stef






Reply | Threaded
Open this post in threaded view
|

Re: Very strange bug with Cog VM

Stéphane Rollandin
> #('0.0 closeTo: 0' '(1.0 closeTo: 1) and: [0.0 closeTo: 0]' '0.0
> closeTo: 0') collect: [:s| Compiler evaluate: s] #(true true true)

I get #(true false false) in the 4.1 image released as a all-in-one Cog
zip for Windows by Andreas, on the 20th, which you can get at
http://squeakvm.org/win32/experimental/CogSqueak4.1.zip


Stef





Reply | Threaded
Open this post in threaded view
|

Re: Very strange bug with Cog VM

Stéphane Rollandin
>> #('0.0 closeTo: 0' '(1.0 closeTo: 1) and: [0.0 closeTo: 0]' '0.0
>> closeTo: 0') collect: [:s| Compiler evaluate: s] #(true true true)
>
> I get #(true false false) in the 4.1 image released as a all-in-one Cog
> zip for Windows by Andreas, on the 20th, which you can get at
> http://squeakvm.org/win32/experimental/CogSqueak4.1.zip

To be as comprehensive as possible in my report:

I get #(true true true) on that very image with a regular Squeak4.0.2 VM.

I get #(true false false) either with the Cog VM in the archive or with
the newer VM Levente provided today, based on VMMaker-oscog.17.mcz


Stef



Reply | Threaded
Open this post in threaded view
|

Re: Very strange bug with Cog VM

Eliot Miranda-2
Hi Stéphane ,

    moving to the VM list...

   we need to be specific on platforms.  I've tested on both Mac and Windows, but not yet on linux.  These VMs I've built myself from the svn tree:

svn info
Path: .
Repository UUID: fa1542d4-bde8-0310-ad64-8ed1123d492a
Revision: 2219
Node Kind: directory
Schedule: normal
Last Changed Author: eliot
Last Changed Rev: 2219
Last Changed Date: 2010-06-25 14:32:53 -0700 (Fri, 25 Jun 2010)

The build version infos are:
(Smalltalk getSystemAttribute: 1006), (String with: Character cr), (Smalltalk getSystemAttribute: 1007), (String with: Character cr), (Smalltalk getSystemAttribute: 1008)

'Mac OS X built on Jun 26 2010 08:21:33 Compiler: 4.2.1 (Apple Inc. build 5659)
CoInterpreter VMMaker-oscog.17 uuid: b0ce2f1b-0e18-48d8-a490-4b8ca8476ec7 Jun 27 2010
SimpleStackBasedCogit VMMaker-oscog.16 uuid: f8b8c00f-be77-4358-a537-ac5dd4e27747 Jun 27 2010'

 'Win32 built on Jun 27 2010 10:28:30 Compiler: 3.4.4 (cygming special, gdc 0.12, using dmd 0.125)
CoInterpreter VMMaker-oscog.17 uuid: b0ce2f1b-0e18-48d8-a490-4b8ca8476ec7 Jun 27 2010
SimpleStackBasedCogit VMMaker-oscog.16 uuid: f8b8c00f-be77-4358-a537-ac5dd4e27747 Jun 27 2010'

and for both Mac OS X & Win32 the results in a Squeak4.1 updated from trunk are
#('0.0 closeTo: 0' '(1.0 closeTo: 1) and: [0.0 closeTo: 0]' '0.0 closeTo: 0') collect: [:s| Compiler evaluate: s] #(true true true)

What's your platform and build info?

cheers,
Eliot

2010/6/30 Stéphane Rollandin <[hidden email]>
#('0.0 closeTo: 0' '(1.0 closeTo: 1) and: [0.0 closeTo: 0]' '0.0
closeTo: 0') collect: [:s| Compiler evaluate: s] #(true true true)

I get #(true false false) in the 4.1 image released as a all-in-one Cog
zip for Windows by Andreas, on the 20th, which you can get at
http://squeakvm.org/win32/experimental/CogSqueak4.1.zip

To be as comprehensive as possible in my report:

I get #(true true true) on that very image with a regular Squeak4.0.2 VM.

I get #(true false false) either with the Cog VM in the archive or with the newer VM Levente provided today, based on VMMaker-oscog.17.mcz


Stef






Reply | Threaded
Open this post in threaded view
|

Re: Very strange bug with Cog VM

Andreas.Raab
In reply to this post by Stéphane Rollandin
On 6/30/2010 2:10 PM, Stéphane Rollandin wrote:
>> #('0.0 closeTo: 0' '(1.0 closeTo: 1) and: [0.0 closeTo: 0]' '0.0
>> closeTo: 0') collect: [:s| Compiler evaluate: s] #(true true true)
>
> I get #(true false false) in the 4.1 image released as a all-in-one Cog
> zip for Windows by Andreas, on the 20th, which you can get at
> http://squeakvm.org/win32/experimental/CogSqueak4.1.zip

This image is missing the Cog floating point byte order fixes. I need to
update it.

   - A.

Reply | Threaded
Open this post in threaded view
|

Re: Very strange bug with Cog VM

Eliot Miranda-2


On Wed, Jun 30, 2010 at 2:44 PM, Andreas Raab <[hidden email]> wrote:
On 6/30/2010 2:10 PM, Stéphane Rollandin wrote:
#('0.0 closeTo: 0' '(1.0 closeTo: 1) and: [0.0 closeTo: 0]' '0.0
closeTo: 0') collect: [:s| Compiler evaluate: s] #(true true true)

I get #(true false false) in the 4.1 image released as a all-in-one Cog
zip for Windows by Andreas, on the 20th, which you can get at
http://squeakvm.org/win32/experimental/CogSqueak4.1.zip

This image is missing the Cog floating point byte order fixes. I need to update it.

Doh!  I should have suspected as much.

humble mumble,
Eliot
 

 - A.