Ernst Kloppenburg <[EMAIL PROTECTED]> wrote: > I am using 'tail' from textutils 2.0-6 on a Debian GNU/Linux system. > > When tail is used in a pipe, it behaves strangely when the option > --follow=name is given. The behaviour I found strange is illustrated by the > following example: > > 1) tail -f /var/log/syslog | wc > when wc is terminated by kill <pid>, the tail program exits the next time > a line is written to the file followed. This makes sense, because at this > time it is detekted that the pipe is broken. > > 2) tail --follow=name /var/log/syslog | wc > in this case, when wc is killed, the tail command never terminates
Thanks for the report. I can't reproduce that using the latest test release on the same type of system. Would you please try it? ftp://alpha.gnu.org/gnu/fetish/textutils-2.0.18.tar.gz ftp://ftp.funet.fi/pub/gnu/alpha/gnu/textutils/textutils-2.0.18.tar.gz You may need this patch: [!__BOUNDED_POINTERS__]: Define away __bounded, __unbounded, and __ptrvalue. ============================================================================= Index: lib/regex.c =================================================================== RCS file: /fetish/textutils/lib/regex.c,v retrieving revision 1.75 retrieving revision 1.76 diff -u -p -u -r1.75 -r1.76 --- lib/regex.c 2001/11/27 21:34:15 1.75 +++ lib/regex.c 2001/12/02 17:46:54 1.76 @@ -112,6 +112,13 @@ # define gettext_noop(String) String # endif +/* Support for bounded pointers. */ +# ifndef __BOUNDED_POINTERS__ +# define __bounded /* nothing */ +# define __unbounded /* nothing */ +# define __ptrvalue /* nothing */ +# endif + /* The `emacs' switch turns on certain matching commands that make sense only in Emacs. */ # ifdef emacs _______________________________________________ Bug-textutils mailing list [EMAIL PROTECTED] http://mail.gnu.org/mailman/listinfo/bug-textutils