Warnings are not errors for a reason, they exist as advice that may or may not apply.
And that's why it's important to release warning-free code to outsiders who can't always easily determine the true intent of the original author. I am using Boost to help support the well-respected BRL-CAD package (http://brlcad.org) whose policy is to release warning-free code, so Boost releases being warning-free would be a very good thing.
Indeed, but please bare in mind that GCC may not be the author's normal "everyday" compiler, and since -Wshadow isn't on by default (and isn't warned about by most other compilers) they may simply not be aware of the issue. That and an ever shifting number of warnings when new compiler releases come out. I suggest you file bug reports regarding any problems you find: if the fix is trivial and they come with patches even better. While I try to keep my stuff warning clean with MSVC and GCC, but after than all bets are off. Certainly for compilers like Intel's that output pages and pages of "notes" that aren't even actual warnings :-( John.