[kaffe] Re: jit breakage due to missing function

[Dalibor Topic]

Dalibor Topic wrote:
> Hi Kiyo,
> 
> Kiyo Inaba wrote:
> 
>> Hi Dalibor,
>>
>> You said,
>>
>>> It seems that the configure script messes something up in the 
>>> detection of getaddrinfo. I'll look into it.
>>
>>
>>
>> And I think this is not the only problem for Solaris. At least, jni.c
>> shall be fixed to coexist with not only JIT3 but also JIT. The jit
>> version does not have function 'resetConstants'.
>>
>> Am I right?
> 
> 
> Yes, of course.
> 
> Currently the build for sparc-solaris (and all other platforms that have 
> jit, but no jit3) fails because the jit engine doesn't have a function 
> called resetConstants like the jit3 engine does.
> 
> What we could do would be
> a) #ifdef the resetConstants call to only happen if defined TRANSLATOR 
> and JIT3.
> or
> b) backport the resetConstants implementation to jit engine
> or
> c) provide a 'do-nothing' implementation of resetConstants for jit engine
> or
> d) separate the engine spefic bits of jni.c into jnisomething.c/h in 
> engine directories.
> 
> I'd propose going with a) followed by c and d, if someone volunteers to 
> do it. I could do step a) myself ;)

Step A: Done. Jit engine should build fine again (did for me on 
i386-linux). Kiyo, thanks again for pointing out it was broken!

cheers,
dalibor topic