Re: new snapshot available: coreutils-6.12.208-2441

[Jos Backus]

        Hi Jim,

On Sat, Sep 27, 2008 at 07:36:06PM +0200, Jim Meyering wrote:
> I think the trick to doing this right is as follows:
> 
>     When a tailed file (name) disappears, and a subsequent
>     open attempt shows the new file is nonexistent or empty,
>     continue tailing the old file descriptor.
> 
>     Only once the new file has content do we perform the final read
>     and then close the old descriptor.
> 
> Which means we'd need a more involved patch.
> Do you feel like implementing that?

I'll try because I want to see this bug fixed, but I'm going to need a hand
understanding the recheck() logic.  While it took me only a minute to create
my initial patch (which does appear to fix the problem; does it break
anything, you think?), I'm having a hard time seeing how to implement the
change you suggest.

It looks like one of the first things that need fixing is the case where we do

    touch foo
    tail -F foo &
    mv foo bar

Today this produces "foo has become inaccessible". What should happen instead
is that we should continue to tail the open fd for foo, which presumably means
getting rid of this error message and setting `ok' to true. This introduces a
change in behavior; is that okay?

-- 
Jos Backus
jos at catnook.com