bug-gnulib
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: rerun


From: Ralf Wildenhues
Subject: Re: rerun
Date: Sat, 12 Jun 2010 09:31:42 +0200
User-agent: Mutt/1.5.20 (2009-10-28)

Hello Bruno,

* Bruno Haible wrote on Fri, Jun 11, 2010 at 09:42:50PM CEST:
> > Wouldn't build-aux/rerun be a good candidate for a new gnulib module?
> 
> No, why would it? Packages don't ship 'rerun'. Only maintainers have it in
> their PATH somewhere.
> 
> I haven't initiated moving 'git-merge-changelog' out of gnulib, where it
> doesn't belong, just to repeat the same mistake with 'rerun'.

I'm not quite sure what mistakes you are hinting at (I didn't follow the
discussion closely, except for the hint I gave), and I'm fine with rerun
not being part of gnulib, but one problem with the script you've posted
is that it does not carry a copyright statement of any kind, which means
it wouldn't be possible for somebody else to use, say, the GPL freedoms
with it.  It's non-free, and non-trivial.

You don't need a GNU or non-GNU package to free it up.  A GPL copyright
statement, even without copyright assignment, is good enough for many
uses.  If you want to allow them, that is.  But if you don't, then I'm
wondering why you even post the code on a GNU list.

> > Wrt. portability, this line:
> > 
> >     diff -u "$cachefile".old - | sed -e '1,2d' -e 
> > '/^-/{s/^-/-'"$color_blue"'/;s/$/'"$color_off"'/}' -e 
> > '/^+/{s/^+/+'"$color_red"'/;s/$/'"$color_off"'/}'
> > 
> > could profit from a quick test whether diff supports -u
> 
> The quality of the output depends on unified diff format. If the maintainer
> does not have it installed, he needs to install it first.

That doesn't mean the script can't be helpful with a vendor diff.
Graceful degradation and all that.

> >   sed -e '1,2d' \
> >       -e '/^-/{' -e 's/^-/-'"$color_blue"'/' -e 's/$/'"$color_off"'/' -e 
> > '}' \
> >       -e '/^+/{' -e 's/^+/+'"$color_red"'/' -e 's/$/'"$color_off"'/' -e '}'
> 
> No, it violates POSIX to pass several incomplete sed scripts as separate -e
> arguments.

We've discussed this on bug-gnu-utils already.  Of course an alternative
is making one script argument here and separating the {, }, and s
commands with newlines.

> > then a --version option, and the script would be in pretty good shape
> > already.  Maybe if the exit status could be nonzero if either the diff
> > is nonempty or the $command failed, that would be neat.
> 
> Feel free to improve the script in any way you like!

See above.  This sentence is not a copyright statement.

Cheers,
Ralf



reply via email to

[Prev in Thread] Current Thread [Next in Thread]