Re: [Gnu-arch-users] Re: Re: Re: hack request

[Paul Pelzl]

On Sat, Oct 02, 2004 at 01:30:37PM +0200, Nikolai Weibull wrote:
> * Phil Frost <address@hidden> [Oct 02, 2004 03:50]:
> > > This should be done fine.  Vim does some contents checking when it
> > > fails to determine filetype based on extension (which is rather hard
> > > to do on stdin) and it works fine for diff's here.  I don't know how
> > > long contents checking for diff contents has been around, but it's
> > > in 6.3 for sure,
> 
> > It works here on the output of say, diff -u or svn diff, but not tla. I
> > assume it's because of the extra information tla outputs at the top of
> > the file.
> 
> Aha!  Of course.  OK, I'll try to bring this up with Bram.  Perhaps we
> can simply increase the number of lines checked.  I only checked with a
> standard diff...silly me,
>       nikolai

Will that scale?  You can get rather a lot of non-diff information at
the beginning of the output if you have a big changeset.  Sometimes I'll
look at the output of "tla delta --diffs" for revisions that are 50 or
more patches apart, and that can easily generate 500 lines of meta info
for a moderate size source tree.

Of course, if the context checking algorithm would look at some lines at
the end of a file as well as the beginning, there should be no
problem...

Paul