Re: tail bug - starting in 7.5 "tail --follow=name" behaves like "tail --follow=name --retry"

[Jim Meyering]

Tim Underwood wrote:
> Going from Ubuntu 10.04 (coreutils 7.4) to Ubuntu 10.10 (coreutils 8.5) I
> noticed that "tail --follow=name" started behaving like "tail --follow=name
> --retry".  It looks like the issue started in 7.5 and still exists in 8.9.
>
> To reproduce:
>
> tail --follow=name this_file_does_not_exist
>
> The expected behavior is:
>
> $ ./coreutils-7.4/src/tail --follow=name this_file_does_not_exist
> ./coreutils-7.4/src/tail: cannot open `this_file_does_not_exist' for
> reading: No such file or directory
> ./coreutils-7.4/src/tail: no files remaining
>
> The wrong behavior is:
>
> $ ./coreutils-7.5/src/tail --follow=name this_file_does_not_exist
> ./coreutils-7.5/src/tail: cannot open `this_file_does_not_exist' for
> reading: No such file or directory
> <tail waits forever for this_file_does_not_exist to become available which
> is what --retry is for>
>
> Here's the wrong behavior on 8.9:
>
> $ ./coreutils-8.9/src/tail --follow=name this_file_does_not_exist
> ./coreutils-8.9/src/tail: cannot open `this_file_does_not_exist' for
> reading: No such file or directory
> <waits forever...>
>
> My specific use case is to tail a log file until it gets deleted at which
> point tail should exit allowing my script to continue.

Thanks for the report.
That's a bug even in the very latest coreutils-8.11.
We'll look into it soon.