Re: [Boost-users] local variable shadow

From: Robert Ramey (ramey_at_[hidden])
Date: 2007-02-19 13:01:39

Once one starts adding patches to suppress warnings, you come upon the fact 
that compilers have different sets of warnings.   And some warnings come 
with templated code.  If one "patches" this, he can end up with and 
#ifdef/#endif mess.  Worst of all, this is not apparent to the person who 
does the patching as he is concerned with one specific compiler so then the 
next person has to clean things up.  Finally, In some cases there is a case 
to be made for shadow names.  That is sometimes there is a global concept to 
be overloaded locally and this idea is expressed with name shadowing. 
Without doing this, one ends up with more names to keep track of in his 
head.

And then there are warnings that a compiler emits which are totally bogus 
for the circumstance at hand.  In one case of the serialization library I 
had to apply such a patch just because it was easier than explaining time 
and again why this warning was bogus for the particular case - its a warning 
not an error for a reason.  This was just so people could get a "clean 
compile"  That is a compile certified as "clean".

So, though I see the appeal of the idea of applying patches to suppress 
warnings and I also see the appeal of making innocuos changes to make a 
clean compile (so real errors are highlighted), I'm not totally convinced 
its a good idea for code which is to be compiled on wildly varying platforms 
and abi environments.  I believe that the work in propogating this to all 
the build environments is underestimated.

Robert Ramey

Chris Saunders wrote:
> I find local variable shadow warnings helpful, but when I include
> various boost libraries, I have to remove a lot of shadowing within
> the boost functions themselves. Note that I'm not refering to name
> conflicts with my own code.
>
> How can I submit patches to remove these type of warnings?
>
> Thanks, -Chris