We always bring XBoard to a state where we can do a 'make' without getting
any warnings. When Arun encounters issues on the build server that do not
crop up when I compiled in my own Ubuntu 10.04, he reports them to me, and
we fix them too.
I develop on a 32-bit laptop, though, and XBoard was originally developed
on x86, so it contained many issues with pointer-int conversions when
compiled for 64-bit, and initially I created new issues of that type
because I was not sufficiently 64-bit aware. (Originally I only worked on
WinBoard, and virtually no WinBoard user compiles his own binary, so as
long as it compiled on my 32-bit system and I released the resulting
binary, there never were any complaints.)
As to warnings that occur when you do more strict checking than the
Makefile does, I can only say that some normally requested by -Wall where
intentionally disabled, because I considered them counter-productive. I am
not going to write error-prone and difficult to understand code just
because to silence some stupid, pedantic compiler... Silencing them with a
compiler option does the job quite effectively!
It would be good to clean up the code base in this respect. But it is
indeed a concern in which branch to do this, which needs careful
consideration. The branch that is currently developed the furthest is the
'aliennew' branch in my hgm.nubati.net repository. It forks off from
master, but unfortunately master has already evolved since the fork to a
point where it is no longer trivial to rebase aliennew to the head of
master, although it is in principle my intention to keep it there. I am
also not sure if aliennew is fully operational as XBoard compile; I only
used it to release the 'WinBoard Alien Edition'.