bug-inetutils
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [bug-inetutils] tests/syslogd.sh and /tmp


From: Mats Erik Andersson
Subject: Re: [bug-inetutils] tests/syslogd.sh and /tmp
Date: Thu, 26 Jan 2012 02:43:20 +0100
User-agent: Mutt/1.5.18 (2008-05-17)

torsdag den 26 januari 2012 klockan 00:47 skrev Ludovic Courtès detta:
> Hi Mats,
> 
> Mats Erik Andersson <address@hidden> skribis:
> 
> > I have now pushed a rewrite of "tests/syslogd.sh". It is also possible
> > to use it as a replacement in inetutils-1.9 and inetutils-1.9.1. There
> > is a full implementation of TMPDIR sensing with due value quoting.
> > The test script works correctly with a SVR4 Bourne shell as found in
> > OpenSolaris and OpenIndiana, in addition to have been tested with bash,
> > dash, mksh, pdksh, and busybox-ash.
> 
> Thanks for working on it!
> 
> > A message from NixOS as to the suitability of this modification would
> > be most welcome.
> 
> It can be seen at <http://hydra.nixos.org/jobset/gnu/inetutils-master>.
> It???s currently failing, but I???ll let you analyze the log.  ;-)

1) The test is running as UID=30003. Therefore it is unable to erect
   an INET listening socket at 514/udp, making this untestable.
   This disables two tests, but does not influence the failure.

2) The build root is at 

    /tmp/nix-build-irjj........-87b6-dirty/

   which in itself exceeds 108 character, thus making itself impossible
   as IU_TESTDIR for creation of a UNIX socket at which syslogd is to
   receive any message. The creation attempt by syslogd fails and logger(1)
   also fails:

     ../src/logger: UNIX socket name too long

3) The value of TMPDIR

   /tmp/nix-build-....-dirty.drv-0

   is luckily 84 characters long, so the use of IU_LONG_SOCKET is succeeding,
   but somehow syslogd(8) is not recording the complaint that 
IU_EXCESSIVE_SOCKET
   is too long, the message is not found in "$IU_TEMPDIR/messages". Momentarily
   I do not see why this escapes detection.

The most important conclusion is that NixOS must make an effort to reduce the
path length for the working directory. Presently it distroys every hope of
basing collision-free UNIX socket names on it. The second component in the
chroot path contains a randomised alphanumeric string of length 32. Shorten it!

Good luck!
> 
> Thanks,
> Ludo???.

-- 
Mats Erik Andersson, fil. dr
<address@hidden>
2459 41E9 C420 3F6D F68B  2E88 F768 4541 F25B 5D41

Abonnerar på: debian-mentors, debian-devel-games, debian-perl,
              debian-ipv6, debian-qa



reply via email to

[Prev in Thread] Current Thread [Next in Thread]