[OpenAFS-devel] warnings fix

[Marcus Watts]

Russ writes:
...
> Bug fix for a patch from Simon that didn't quite do the right thing on
> Windows.
...

Right, I stand corrected.  But this doesn't really help.

...
> Yeah, in general I would warn everyone doing OpenAFS development that the
> new tools are so much better that the pace of development is likely to
> increase.  In the short term, a lot -- in the long term, not as much as at
> first, but hopefully still significantly.
> 
> This means that maintaining a large monolithic patch set outside of the
> tree is going to require constant and significant rebasing.  You want to
> get self-contained portions of it into the tree as soon as possible so
> that everyone shares the work instead of having to keep it current outside
> the tree.
...

Then we really do have a problem.

The current process is adapted to small changes.  It deals with
medium changes inefficiently.  It fails to address big changes at all.

There was a discussion earlier about why osd was done to 1.4.x instead
of 1.5.x or cvs head -  this is directly related, and I've hit
the very same issue with rxk5.

		for the		for any real users
		gatekeepers
cvs head	best		worst
1.5.x		2nd best	mainly for windows
1.4.x		worst		only acceptable choice

I made an early foray with cvs head and rxk5.  I gave up
because my builds broke so often I spent all my time working
on those bugs rather than on rxk5.  The above change history
that Russ detailed shows this problem exists today just as much
with gerrit; so the problem here is not with the technology.

With most of my recent rxk5 work, I stuck to 1.5.x releases
as "stable points".  This was helpful except when the volume
code broke.  :-(

> 
> Also, we may be doing some global cleanups in the future.  We'll try to
> warn everyone in advance and discuss that on the list, particularly if it
> will cause significant problems for pending out-of-tree patches.

As it happens, there are 2 easy global cleanups that I would desperately
*like* to see because it will solve significant problems for me with rxk5.

/1/ remove trailing white space from lines.

/2/ get rid of "register" declarations.

Ideally, also, run everything through "unexpand -a".

If it would help, I'm willing to write a perl script that
would do this. ( Actually, I'm probably stuck writing this
even if nobody else does value it. )

I am *not* in favour of automatic code formatting.  There's
multiple places in the afs source where some well meaning person
added extra comments to make something clearer, and some automatic
code formatter mangled it.

					-Marcus