Re: [PATCH setup 0/3] Setup replacement for incver_ifdep

[Jon Turney <jon dot turney at dronecode dot org dot uk>]

On 26/11/2015 10:11, Corinna Vinschen wrote:
> On Nov 24 19:44, Achim Gratz wrote:
> > Jon Turney writes:
> > > > So, should we try to guard against that (installations on a USB stick
> > > > are probably the only practical occurence these days)?  I wouldn't mind
> > > > if we just unconditionally rebuild on FAT(32).
> > >
> > > Thinking this over, it doesn't seem that hard to use a hash to
> > > determine if the directory has changed.
> >
> > As I said before, if you're going that route I'd rather do the same
> > thing as for autorebase and maintain a list of info files.  This way you
> > could remove and add individual info files rather than just rebuilding
> > everything, just the way post-install works on Linux in this case.
> >
> > > How about the attached?
> > >
> > > (This will add a dependency on diffutils for cmp)
> >
> > You could read the md5sums into variables instead and just compare the
> > resulting strings within bash.  Otherwise we'd have to add diffutils to
> > the Base category.

I've made this change.

> > > > Otherwise, it looks good and thanks for writing up a script.
> > >
> > > #!/bin/bash
> >
> > Use /bin/dash if possible.
>
> I trust both of you to do the right thing.  My question here is only,
> can we get a solution soon so we can get rid of the old upset method for
> the info files?  Achim, how long would it take to create the same
> solution for info you're using for rebase?  Would it make sense to use
> Jon's solution in the interim?

As an interim solution, I've adopted _update-info-dir, and updated it to 
contain this script.

This removes the last use of autodep, which unblocks some further 
improvements to upset.

Future work: I can't see any reason why this script now needs an 
independent existence, so it could be absorbed by the info package.