Subject: resolving the Unix warnings flamewar
From: Paul Rohr (firstname.lastname@example.org)
Date: Wed Feb 09 2000 - 15:56:08 CST
OK, before this one gets too out of hand, let's focus on concrete solutions
to specific problems. It's easy to suck a lot of bandwidth arguing about
abstract principles. Coded solutions are a lot more useful, though.
The real job is to find a practical, maintainable way to reduce the number
of warnings generated on Unix, preferably to zero. (I presume we're all in
consensus that this is a Good Thing, right?) There have been a lot of
issues raised, and a lot of solutions proposed, but I'm not sure we're all
headed in a single concrete direction here.
So, here's a modest proposal.
1. As a first pass, let's focus on making the debug version of the abi tree
warning-free on Unix. If we're not there yet, this is probably just a
patch. However, if there *are* any specific coding idioms which need to be
addressed, it should be easy to come up with straightforward examples, which
we can discuss and resolve on a case-by-case basis.
2. Once that's done, the next step is to do the same thing for release
builds. At that point, we can revive the whole debate about asserts on
otherwise unused arguments. Unless anyone's interested in redesigning APIs
throughout the product to avoid passing unused arguments, I expect that
we'll be debating alternatives like -Wno-unused vs. Justin's nifty
UT_UNUSEDARG macro (not to be confused with the current UT_UNUSED macro,
which gets used a bit differently).
3. Finally, once the abi tree has been thoroughly scrubbed, we can
entertain proposals to clean the third-party libraries we use. Anyone
interested in doing this would also be taking on the task of getting the
necessary patches submitted back to the maintainers of the original code.
This archive was generated by hypermail 2b25 : Wed Feb 09 2000 - 15:50:43 CST