Re: story - part 3

[dichro-fefff94a@eris.rcpt.to (Mikolaj J. Habryn)]

>>>>> "k" == kiewicz  <Arkadiusz> writes:

    k> Maybe second parametr for directory should be regexp to filter
    k> files with aren't configuration files (and we want filter out
    k> these files) ...  In this case regexp rule should looks like
    k> this ".*(~|.rpmsave|.rpmorig|.rpmnew)$".

  Done - you're quite right, the regex(7) manpage has nothing on
negation in it.

    k> 2) directories in directory specified in config file (ie
    k> /etc/rlinetd.d/) should be ignored. For example I had:

  Also done.

    k> funny :-) How it can use negative size of memory ?!

  It's the difference in memory usage across the reload - I use it to
track memory leaks in the parser :) What you just saw is a reshuffle
of the arena that wound up freeing a block.

    k> 4) If getaddrinfo() is available in system library configure
    k> shouldn't check for libinet6.a (because it's linking with
    k> libinet6.a even if we compile on new glibc system).

  Okay.

    k> rl_warn() tries to send some messages to syslog but file
    k> descriptor was closed (closelog()) before so it can't send this
    k> message to syslogd (am I right ?)

  Nope - syslog(3) will reopen the connection to syslogd. The reason
that I close it is that it might have been sitting on fds 0-2, and the 
dupping would mean that it would wind up writing the log message out
the connected file descriptor.

    k> Is there C function to test if all libraries needed to run
    k> specified program are available ? 

  Yep - exec(2) :) Seriously, the short answer is no. It's something
we won't know until the dynamic linker has done it's stuff. This is
likeliest to show up when people play chroot games - they need either
a static linked executable or a full set of libraries in the jail.

    k> 7) Support for socket type (raw, seqpacket) will be implemented
    k> in rlinetd ?  (inetd && xinetd have this). 

  Haven't looked at this yet - what uses it, just out of curiosity?

    k> Also access_times,
    k> env and passenv options in xinetd is interesting ;-)

  Environment fiddling has been on the todo list since day one :)
I'll think about access_times as well - should be fairly easy to do
once there's some timer code in place.

    k> ps. I hope that you don't have enough of my reports 8-)

  Not even slightly, keep them coming :)

m.