distcc fails to distribute to IPv6 addresses

Actually, when an IPv6 address is in the hosts file, distcc fails to send jobs to any server and compiles everything locally.

Confirmed in Google Code.

Lacking an upstream release, I would be very happy if someone integrated the two patches that did not make it into the 3.1 release:
http://code.google.com/p/distcc/source/detail?r=650
http://code.google.com/p/distcc/source/detail?r=673

They are mentioned here:
http://code.google.com/p/distcc/issues/detail?id=34
http://code.google.com/p/distcc/issues/detail?id=42 seems to be a duplicate

This should be fixed in all but Hardy.

Ken Sharp (kennybobs) wrote 8 hours ago: #5
> This should be fixed in all but Hardy.

You confirm that zeroconf discovery, and manual entry in hosts files, of IPv6 addresses works on 3.1-4ubuntu2?

Upstream #34 claims to require these two commits:

 http://code.google.com/p/distcc/source/detail?r=650
 http://code.google.com/p/distcc/source/detail?r=673

while the mentioned Ubuntu version only includes the later.

Daniel Hartwig (wigs) wrote 1 minute ago: #6
> Ken Sharp (kennybobs) wrote 8 hours ago: #5
>> This should be fixed in all but Hardy.
…

or perhaps I misread you and you are rather requesting the fix to be applied to those versions? :-)

Oops, no, you're right. I misread it. Sorry for the noise.

The fix is upstream so I guess it should be ready to be packaged.

3.1-5 includes the fixes for both patches in 3.1-4ubuntu2, so no Ubuntu changes are required on this version.

This bug was fixed in the package distcc - 3.1-5

---------------
distcc (3.1-5) unstable; urgency=low

  * new maintainer (Closes: #664497)
  * update Standards-Version to 3.9.3:
    - debian/distcc.init.d: start action exits successfully if daemon
      is already running [9.3.2]; postinst will succeed on upgrades in
      this case also (Closes: #620773, LP: #822887)
  * switch to dpkg-source 3.0 (quilt) format
  * remove Build-Depends: dpatch
  * add Build-Depends: autotools-dev
  * debian/rules:
    - use debhelper compat level 9
    - reduce to "dh $@" style
    - fixes FTBFS caused by previous build-arch target (Closes: #666383)
    - enabled bindnow hardening (no PIE yet, it causes build failure)
  * debian/control:
    - added Homepage field
    - distcc: description starts lowercase
  * debian/distcc.config:
    - renamed from debian/config
    - use "set -e"
    - include missing #DEBHELPER# token
  * debian/distcc.init.d:
    - added stop levels 0, 6
    - unset TMPDIR before starting the daemon to avoid problems with
      root's value for this being unwritable by the distccd user; to
      provide distccd with a sensible TMPDIR put it in /etc/default/distcc
      (Closes: #514556)
  * debian/distcc.postinst:
    - use "set -e" instead of "#!/bin/sh -e"
    - drop extra calls to "sed ... /etc/default/distcc" for every field
      that only removed whitespace which would be removed anyway when
      updating the fields
    - always create distccd user if it does not exist (Closes: #548053)
  * debian/distccmon-gnome.menu:
    - changed title to "distcc monitor"
    - dropped incorrect hint tag
  * debian/patches:
    - 06_set-pythonpath-securely.patch: contains fix for #605168 which
      was previously applied directly to the source
    - 07_preferred-user.patch: starting the daemon as root causes it to
      change user, which should be to "distccd" in Debian
    - 08_gnome-data-public-dirs.patch: install desktop and icon files for
      distccmon-gnome /usr/share/applications and /usr/share/pixmaps
      respectively (LP: #512288)
    - 09_rename-pump.patch: rename the "pump" command to "distcc-pump" in
      all references such as help text, man pages, etc. (Closes: #594083)
    - 10_consecutive-preprocessor-options.patch: correctly count
      preprocessor options (Closes: #626926)
    - 11_lsdistcc-man.patch: add man page for lsdistcc utility
  * debian/watch:
    - added remote watch file
  * source/config.{guess,sub}:
    - update with autotools-dev during build instead of directly patching
      the source tree
  * use dpkg triggers to dynamically generate/update compiler links
    based on the ccache packaging. Thanks to Daniel Schaal
    (Closes: #651670)
  * cherry-pick upstream fixes for IPv6 support (Closes: #452835):
    - r650_ipv6-zeroconf.patch: IPv6 patch for Zeroconf and IPv6 literals
      in hosts file (Closes: #481951, LP: #593047)
    - r673_zeroconf-nodups.patch: remove duplicate hosts from the Zeroconf
      list (LP: #809534)
    - r678_distcc-v6-acl-2.patch: IPv6 support for access control
  * cherry-pick other upstream fixes:
    - r732_distccmon-gnome.patch: avoid client list growing indefinitely
  ...

These three patches are all the upstream changes to fix IPv6 issues:
- r650 is small and required for ipv6+zeroconf to work correctly;
- r673 is already included in 3.1-4ubuntu2 (oneiric and later);
- r678 may be considered optional, nice to include for full ipv6 support.

The regression potential of r650 is tiny, given the small size of the patch it is easy to verify.

Although fixing several other issues, 3.1-5 may not be suitable for a SRU due to unrelated packaging and path changes. Especially these:
    - 08_gnome-data-public-dirs.patch: install desktop and icon files for
      distccmon-gnome /usr/share/applications and /usr/share/pixmaps
      respectively (LP: #512288)
  * use dpkg triggers to dynamically generate/update compiler links
    based on the ccache packaging. Thanks to Daniel Schaal
    (Closes: #651670)

See also Bug #521165 for another small patch worth considering (also fixed in 3.1-5).

new version still not available in lucid/updates.

Daniel, do you want to do a stable release update (SRU)? The three patches seems to be SRU-able to me. A debdiff and a test case is needed for a SRU [1].

[1] https://wiki.ubuntu.com/StableReleaseUpdates

> Daniel, do you want to do a stable release update (SRU)?

No. I do not use Ubuntu.

> A debdiff and a test case is needed for a SRU

Recommend that an affected user running Lucid prepare the test case and for bug #521165 also.