[Vm-dev] Integer overflow with BitBlt rule 20 and depth 32

[Juan Vuletich]

Eliot Miranda wrote:
>  
>
> ------------------------------------------------------------------------
>
>
>
> On Mon, Oct 26, 2009 at 1:04 PM, Juan Vuletich <juan at jvuletich.org 
> <mailto:juan at jvuletich.org>> wrote:
>
>
>     Hi Eliot,
>
>     Eliot Miranda wrote:
>
>
>         Hi Juan,
>
>
>         On Mon, Oct 26, 2009 at 11:07 AM, Juan Vuletich
>         <juan at jvuletich.org <mailto:juan at jvuletich.org>
>         <mailto:juan at jvuletich.org <mailto:juan at jvuletich.org>>> wrote:
>
>
>            Eliot Miranda wrote:
>
>
>                On Mon, Oct 26, 2009 at 6:01 AM, Juan Vuletich
>                <juan at jvuletich.org <mailto:juan at jvuletich.org>
>         <mailto:juan at jvuletich.org <mailto:juan at jvuletich.org>>
>                <mailto:juan at jvuletich.org <mailto:juan at jvuletich.org>
>         <mailto:juan at jvuletich.org <mailto:juan at jvuletich.org>>>> wrote:
>
>
>                   Hi Folks,
>
>                   I've just opened
>         http://bugs.squeak.org/view.php?id=7407 ,
>                with a
>                   description of the problem, several tests (based on
>         Henrik's
>                   scripts) and the fix I propose.
>
>                   David T. Lewis wrote:
>
>                        On Sat, Oct 24, 2009 at 10:40:17AM -0300, Juan
>                Vuletich wrote:
>                                         What worries me a bit is the
>         other changes I needed
>                to do
>                           to be able to run the Smalltalk BitBlt
>         simulation
>                and to
>                           do the translation. These are:
>                           BitBltSimulator >> #oopForPointer:   "May be
>         harmless"
>                           CArrayAccessor >> #long32At:         "Why is
>         this
>                needed?"
>                           CArrayAccessor >> #long32At:put:      "Why
>         is this
>                needed?"
>                            
>                       I am just guessing here, but I think that
>         BitBltSimulator
>                       expects to
>                       be used with an interpreter simulator, so
>         perhaps if you
>                       initialize it
>                       with a simulator there will be no need to add these
>                methods.
>
>                        sim := BitBltSimulator new setInterpreter:
>                       InterpreterSimulator new
>
>                       Dave
>                      
>                   Given that BitBltSimulation calls #isIntegerObject: it
>                looks like
>                   the ivar interpreterProxy should hold an
>         InterpreterProxy
>                (as it
>                   already does) and not an InterpreterSimulator. I did
>         not add my
>                   patch to make simulation work to Mantis, as I'm not sure
>                about them.
>
>
>                When simulating the interpreterProxy inst var should
>         hold the
>                InterpreterSimulator, not the proxy.
>                
>
>            Maybe it should be an InterpreterSimulator when simulating the
>            whole interpreter. When calling #copyBitsSimulated, it is
>         set to
>            an InterpreterProxy in #copyBitsFrom: .
>
>            Anyway, I'm asking for help on making #copyBitsSimulated work
>            again, like it should do when called from BitBltTest. If nobody
>            can help with that, I guess I'll open a Mantis issue for this
>            problem, in the hope that some day it gets fixed.
>
>
>         Well with my current VM I see no problems; all 10 tests are
>         green.  What is the bug that you see?  How can I reproduce it?
>
>         tia
>         Eliot
>
>
>     Thank you for caring about this issue!
>
>     I apologize for not being clear. Check
>     #testAlphaCompositingSimulated and
>     #testAlphaCompositing2Simulated. Both do nothing if
>     BitBltSimulation is not there. If you load VMMaker, both tests
>     give errors. The VM in use should be irrelevant, this is pure
>     Smalltalk.
>
>
> I understand that.  I have an image containing BitBltSimulation and 5 
> different VMs :)  I'm developing Cog, a faster Squeak VM.  In my image 
> with BitBltSimulation present all 10 BitBltTest tests pass.  But I 
> have done some work on the simulator.
>
> I am asking you to help me reproduce the bug and then I can export the 
> fixes from my VM to fix VMMaker.  What version of VMMaker are you using?
>

Ok. I see. I downloaded latest trunk from 
http://ftp.squeak.org/trunk/Squeak3.10.2-Trunk-091024.zip.
Opened the image, open MC Browser. Add repository:
MCHttpRepository
    location: 'http://www.squeaksource.com/VMMaker'
    user: ''
    password: ''
Opened the repository
Loaded VMMaker-dtl.145.mcz
Proceeded on the warning about Klatt and FFI.
Run the tests. Both give errors.

Thanks,
Juan Vuletich

Ps. I can't wait for Cog!

>
>
>     It is a very bad idea to have a test do nothing under default
>     conditions! (i.e. no special packages such as VMMaker loaded)
>     That's why I moved both to a new class, that should be part of
>     VMMaker, and removed that silly check for BitBltSimulation. This
>     is included in the stuff I attached to
>     http://bugs.squeak.org/view.php?id=7407 .
>
>     Cheers,
>     Juan Vuletich
>