Bug#1497: acct-alpha-5-7 problems
Marek Michalkiewicz writes ("Bug#1497: acct-alpha-5-7 problems"):
> - /etc/init.d/acct prints the usage message when invoked correctly
> (with "start" or "stop" as the first argument), and doesn't start
> or stop accounting (but running accton by hand works as expected).
> It works after commenting out the ". /etc/init.d/functions" line
> (which is not necessary here anyway), so I guess maybe something
> is screwed up in this file (part of sysvinit).
Almost any script for which /etc/init.d/functions *is* necessary is
broken and needs to be rewritten. The file /etc/init.d/functions
should be removed or replaced by a file containing only a comment
saying `for backward compatibility only', so that people stop using
> - trying to start accounting again when it is already running gives
> a message "Process accounting not available on this system", which
> is a bit confusing (I suggest not to redirect stderr from accton to
> /dev/null, so that one can see what is really the problem).
Redirecting stderr of a program to /dev/null is rarely a good idea.
Can't accton be persuaded to produce a sensible and non-frightening
error message when it gets EINVAL or ENOSYS or whatever the error
return is ?