[kaffe] warning hunt

[Dalibor Topic]

Hi Nektarios,

Nektarios K. Papadopoulos wrote:
> 
> 
> Adam Heath wrote:
> 
>> I've checked in your patch.  Be careful doing fixes to verify.c(and 
>> friends).
>> I've got several fixes locally, but  not checked in, as there are several
>> cascading issues.  I can cope if you send me changes, tho, that are 
>> small and
>> isolated.
> 
> 
> Thanks and sorry for the reckless copy-paste in the prototypes in
> stackTrace.h
> 
> I fixed them and moved them in stackTrace.c (stackTrace.diff attached)
> 
> I also fixed several 'unreachable-code' and a few 'sign-compare' in
> code-analyse.c (code-analyse.diff attached).

thanks for the patches! I've checked them in. I believe it would be 
better to fix sign compare problems by changing the types involved to a 
more reasonable one where possible, but I didn't investigate if it was 
possible in this case.

> I also tried to resolve the 13 warnings:
> 'passing arg xxx of `xxxxxxxxx' with different width due to prototype'
> 
> but I just got confused! It seems that this warning is raised for any
> prototyped function with argument that require conversion other than the
> default. Quoting the gcc man page:
> [[[[
> Warn if a prototype causes a type conversion that is different from
> what would happen to the same argument in the absence of a prototype.
> ]]]]
> Passing literals or variables of the exact type (i.e. constIndex) to
> arg 1 of getField for example still raises the warning.
> Maybe we should just ignore this warnings?

Ignoring confusing warnings sounds like a good policy. I've talked to 
Adam about it, and we've agreed to leave it enabled since it should help 
us pinpoint failures on 64 bit systems, and similar beasts.

keep up the good work,
dalibor topic