Re: potiential problems with parm decls.

[Kenneth Zadeck <zadeck@naturalbridge.com>]

Mark Mitchell wrote:
> Kenneth Zadeck wrote:
>
>   
>> How are you (we) handling PARM_DECLS?
>> Honza has informed my that the PARM_DECLS in the function's
>> DECL_ARGUMENTS needs to be "==" to the set of PARM_DECLS accessed in the
>> function.
>>     
>
> The code I wrote doesn't set DECL_ARGUMENTS, on the theory that
> parameters are essentially local variables.  I think it would be best if
> you set these up because the only information I have is types, and that
> may not be enough.  But, if that's impractical, then I will see what I
> can do.
>
> Also as for:
>
>   
How do you represent the function_decls for the functions that do not
have bodies?  Do these have parm_decls in them, or only a list of
types?  If they have parm_decls, i could see a problem.  If they only
have types, then there is no problem with me doing the parm decls for
the functions i see.


>> Mark claims that everything on his end should work except for static
>> initializers so we may be ready for some larger scale testing as soon as
>> the above bug is tracked down.
>>     
>
> It's more that I don't know what else is broken, rather than that I
> think it's all perfect. :-)  I don't want to give the impression I think
> it's all in tip-top shape.
>   
That is certainly true for my code as well.
> Also, with respect to static initializers, I think I can make it work in
> the direction I started, but if you think it's easy to serialize
> initializers using your existing machinery, that might be an easier
> path.  How hard would it be to leverage your serialization code to put
> static initializers for variables in their own sections, so that the
> initializer for "i" would be in some ".lto.i" section in the same way
> that the body of "f" is in ".lto.f"?  I'm happy to work on this; I'm
> just looking for some guidance as to what you think about it.
>
> Thanks,
>
>   
My only concern here is that you said at the summit that these were not
in gimple.  If they are in gimple, then it should not be a big deal for
me to co-opt the existing machinery to put each of them in its own
section. 


Kenny