Thank you David for your answer.
Actually, it was another problem that removed my methods from the
generated code. In one of my methods, I was calling a method which use
a local variable out of the scope of my method in the generated code.
I guess the code generation detects such kind of errors and removes
the erroneous methods. It would be great to have a warning in such
cases to help the debugging.
SlangBrowser is very helpful, thanks for the link !
Antoine
Le 21-mai-09 à 13:45, David T. Lewis a écrit :
> On Thu, May 21, 2009 at 01:28:48PM +0200, Antoine Marot wrote:
>> Hi all,
>>
>> I am trying to change the method lookup process in the VM but I have
>> some troubles with the code generation.
>>
>> I modified #lookupMethodInClass: in the Interpreter class and I
>> added
>> some methods and some class variables as well. For now, I do not
>> need
>> to build my extension as an internal or external plugin.
>>
>> The generated C-code (in interp.c) includes my modified version of
>> #lookupMethodInClass: (it calls my own methods) but those methods do
>> not appear in the source at all (as well as my class variables).
>>
>> However, I traced the code generation and my methods seems to
>> treated.
>>
>> Any idea of what I am doing wrong ?
>
> There are a couple of possibilities.
>
> If your new method is not referenced anywhere, the C code will not be
> generated. You can force the code to be generated by adding it to
> a list of required methods in Interpreter class>>requiredMethodNames.
>
> If your method has been fully inlined, it will have been embedded in
> the code for other methods and will not appear as a separate function
> in the generated C code. If you want to disable this, add the line
> "self inline: false" in your method.
>
> You may find it easier to debug things like this if you add
> SlangBrowser
> to your image (
http://wiki.squeak.org/squeak/5916>. That will let you
> see the code generation for each method in a class browser.
>
> Dave
>
>
Locked
