[squeak-dev] FFI & Windows 7: troubled

Casimiro de Almeida Barreto casimiro.barreto at gmail.com
Thu Apr 18 04:52:44 UTC 2013


On 17-04-2013 17:09, Frank Shearar wrote:
> On 17 April 2013 19:44, Casimiro de Almeida Barreto
> <casimiro.barreto at gmail.com> wrote:
>> On 16-04-2013 18:15, Frank Shearar wrote:
>>> On 16 April 2013 21:14, Casimiro de Almeida Barreto
>>> <casimiro.barreto at gmail.com> wrote:
>>>> On 16-04-2013 12:34, Frank Shearar wrote:
>>>>> On 16 April 2013 16:26, Casimiro de Almeida Barreto
>>>>> <casimiro.barreto at gmail.com> wrote:
>>>>>> On 16-04-2013 06:41, Frank Shearar wrote:
>>>>>>> On 15 April 2013 20:07, Casimiro de Almeida Barreto
>>>>>>> <casimiro.barreto at gmail.com> wrote:
>>>>>>>> Hello,
>>>>>>>>
>>>>>>>> It seems that FFI-win32 is broken: during load it'll complain about
>>>>>>>> classSide message (meta object). And there's indeed no classSide stuff.
>>>>>>>>
>>>>>>>> MCMethodDefinition>>actualClass
>>>>>>>>     ^Smalltalk at: className
>>>>>>>>         ifPresent: [ :class | classIsMeta ifTrue: [ class classSide ]
>>>>>>>> ifFalse: [ class ]]
>>>>>>> How did you try to load FFI? Into what version of Squeak?
>>>>>>>
>>>>>>> frank
>>>>>>>
>>>>>>>
>>>>>> Using monticello browser. Version 4.4 All-in-one updated from trunk (all
>>>>>> in one). Latest version of FFI from sources.squeak.org.
>>>>> I ask, because I know the Installer (head) version off SqueakMap loads
>>>>> cleanly into an up-to-date 4.4, even though a bunch of tests fail [1].
>>>>> Are you trying to load the FFI-Win32 package without first loading
>>>>> FFI-Kernel and FFI-Pools?
>>>>>
>>>>> Installer squeakMap
>>>>>   update;
>>>>>   install: 'FFI (head)'
>>>>>
>>>>> does the same thing as
>>>>>
>>>>> (Installer repository: 'http://source.squeak.org/FFI')
>>>>>     install: 'FFI-Pools';
>>>>>     install: 'FFI-Kernel';
>>>>>     install: 'FFI-Tests'
>>>>>
>>>>> but note that you _must_ have these packages loaded _before_ you load FFI-Win32.
>>>>>
>>>>> frank
>>>>>
>>>>> [1] http://build.squeak.org/job/ExternalPackages-Squeak4.4/39/#showFailuresLink
>>>>>
>>>>>> CdAB
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>> No.
>>>>
>>>> 1st: FFI-Pools
>>>> 2nd: FFI-Kernel
>>>> 3rd: FFI-win32
>>>>
>>>> But the problem is at MCMethodDefinition, at the message actualClass. If
>>>> the test for classIsMeta is true, then it tries the message classSide.
>>>> Such message does not exist. Was it deprecated ???
>>> What's the value of className in MCMethodDefinition >> #actualClass?
>>>
>>> frank
>>>
>>>
>> Ok, this morning I tried:
>>
>> (Installer repository: 'http://source.squeak.org/FFI')
>>     install: 'FFI-Pools';
>>     install: 'FFI-Kernel';
>>     install: 'FFI-Tests';
>>     install: 'FFI-Win32'
>>
>> and the same problem happened while loading FFI-Win32.
> I'll try reproduce the problem tomorrow, when I'm back in the office.
> (I've no Windows machines at home.) One thing that you might well need
> to do - but you haven't mentioned the symptom that it causes - is to
> set the preference that allows underscores in selectors:
>
>     Scanner prefAllowUnderscoreSelectors: true.
>
> (The symptom is moaning about "WS" things, because FFI-Win32 has a
> bunch of WS_FOO things.)
>
> frank
>
>> CdAB
>>
>>
>>
>>
>
Yeah. Problem was wicked underscores... In linux I allow all underscores
(compiler), but as I don't use Windows in my day to day life, I just
forgot to adjust the settings properly.

Thnx.



More information about the Squeak-dev mailing list