On Wed, 2003-06-18 at 04:12, Kevin Pilch-Bisson wrote:
> I disagree! This option gives useful information sometimes. Just because it
> can't work on _some_ platforms is no reason to disable it on _all_ of them.
I say we punt the option entirely. It can provide some useful
information, but so can lint and valgrind, and we don't angst about how
we aren't always running our code through lint and under valgrind. And
playing with compiler options like this is always a recipe for pain. We
know it works on "Linux," and it's probably even true that it works on
all Linux distributions today, but who's to say the next version of
glibc won't contain a bug in this area? Who's to say it won't change
philosophy and start demanding that programs complied with -std=c89
actually be C89 programs? (A C89 program only uses C89 headers and only
calls C89 functions, so it can't do much of use besides text
processing.)
>From reading
http://philip.greenspun.com/politics/litigation/philosophy.html
I learned that if you want to avoid needless conflict, you should try to
do things the same way as everyone else does, unless you have a very
good reason not to. It's inconceivable that Red Hat will ever come out
with a release with a compiler which fails to build "Hello, world," but
it's quite possible that their compiler will fail to build it if you
pass some unusual flag.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Wed Jun 18 19:11:39 2003