httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Paul Richards <p.richa...@elsevier.co.uk>
Subject Re: indent
Date Thu, 10 Oct 1996 13:23:34 GMT
"Roy T. Fielding" <fielding@kiwi.ICS.UCI.EDU> writes:

> > I think the 2.0 tree is going to become radically different very quickly
> 
> I don't.  More importantly, I want to be able to see the actual differences.

When I say radically different I mean in terms of lines of diffs. I
can't see it happenening any other way with the threading code going
in and the other things that are planned. See the differences between
what exactly? If we find bugs post 1.2 (which we will) and we need to
track down where they occured it's going to be a bloody pain to
backtrace through the cvs changes across indent runs to find the
cause. When new bugs suddenly appear (and in beta test this is more
relevant than usual) the fastest way to find the change that caused
the error is to backtrack through cvs versions (say by day) to find
the commit that caused the bug. If this backtrack is littered with
indent runs across different file this becomes a nightmare. If you can
step back across the formatting commit in one step this isn't such a
problem.

> That won't work.  First off, it is pointless to reformat the code after
> people are using it outside the developers -- the reason to reformat it
> is so that everyone can read it, not just us.  Second, large scale use
> of indent just doesn't work -- the person doing the reformatting has to
> go through it by hand and fix things when it is fresh in their minds,
> not several weeks later when they finally get around to it.  Third, only
> one file should be changed at a time so that the changes can be tested
> in isolation.

You can apply indent individually to files and then test Apache, a
good idea, just don't commit them one at a time. Commit the finished
result not all the intermediate stages.

> The point here is to make the code readable, not just run indent on it.
> And yes, that does include adding and replacing comments (and noting
> areas that need further comment).  There are a number of files for which
> running indent would make them less readable than they are now.

I agree, but checkpointing each stage is important. It's difficult to
get complete snapshots of Apache out of cvs unless you tag the sources
at particular points. Dates don't work as well as people think unless
you know the exact time of the event your looking for.

Making this change as one commit is essential and that means your
proposal isn't viable since at this point too many people have
work-in-progress that would get badly messed up.

-- 
  Paul Richards. Originative Solutions Ltd.  (Netcraft Ltd. contractor)
  Elsevier Science TIS online journal project.
  Email: p.richards@elsevier.co.uk
  Phone: 0370 462071 (Mobile), +44 (0)1865 843155

Mime
View raw message