Re: -Werror

["Jonathan Perkin" <jperkin%joyent.com@localhost>]

I have a lot of sympathy for both sides of this argument.  On the one
hand I think we all want to pursue the qualitative aspects of pkgsrc,
but on the other we have a responsibility to our users to provide
packages which build on their chosen pkgsrc-supported platform.

Unfortunately, we simply do not have the time to pursue a solely
qualitative approach, and if given the choice between 'no package
available at all' and 'package available with a small chance of bugs'
I'd go for the latter every single time.  Users can't test nothing.

My main concern here is that we are being told "don't do that" but
there is no direction on what we should be doing instead.  It is also
unclear whether this is strictly limited to -Werror or whether it
encompasses a larger set of constructs - after all, many of the other
hacks already in pkgsrc have similar if not higher risks attached.

Additionally, this is a practise we have already employed.  The
pkgsrc-2012Q3 announcement proudly says in its first paragraph about
the work gone into support clang, yet that work includes adding the
very same constructs which are now being argued against.

I obviously have a vested interest in this area, as such hacks are a
major reason why we have so many packages available on SunOS platforms
these days (and ultimately why Joyent now run one of the largest
pkgsrc installations in the world).  If we are to be limited in what
we are allowed to do in order to provide software for our users then
it has implications for us too.

Regards,

-- 
Jonathan Perkin  -  Joyent, Inc.  -  www.joyent.com