I am running distcc 3.0 i686-pc-linux-gnu on a gentoo system with GCC 4.1.2. On any compile, or running 'distcc --show-hosts' I get an "ERROR: invalid tcp port specification in "<ipv6 address here>""

Because of this, distcc fails to work with zeroconf on my network (or any network where zeroconf reports IPv6 adresses).

The best solution is to either fix the address parsing so that zeroconf's IPv6 addresses parse correctly, or to force avahi to give only IPv4 addresses.

Thanks for the bug report. A patch would be even nicer!

I'm working on one, but figured I'd submit a bug-report just in case it's been fixed already.

Sure, good idea. There's no fix yet, as far as I know..

May be worth looping in Lennart Poettering, the original author of avahi support in distcc.

Sure, good idea. There's no fix yet, as far as I know..

May be worth looping in Lennart Poettering, the original author of avahi support in distcc.

I've fixed the problem by adding IPv6 literal host support to the TCP host parser and making the zeroconf host file writer handle IPv6 Hosts properly. I also update the requisite parts of the distcc manpage.

Known/suspected remaining bugs: * If avahi returns the same host on both IPv4 and IPv6, it will be entered into the host list twice. * This patch doesn't pay attention to the IPv6 configure flags, so there may be problems there that usually won't be encountered when a user without IPv6 support tries to use an IPv6 literal in his host file.

The patch looks good to me.

I've added you as a member of the distcc project on code.google.com, so you can go ahead and commit it into the repository, if you want.

Committed patch to SVN

Can we mark this one as fixed now?

Probably, though I think I may have missed some configure flags. But it's fixed the bugged problem.

As a followup to the 'known/suspected bugs' (sorry, wasn't sure whether I should open a new issue), the attached patch removes hosts that have duplicate service names.

So, e.g. my laptop that has both IPv4 and IPv6 addresses on both its wired and wireless interfaces, instead of showing up four times, only shows up once.

The deduplication patch needs more work, it should basically concatenate all ips reported for a host and then select one based on if it can actually reach it (route, ipv6 vs ipv4 etc)

While this might add support for avahi ipv6, but what about the "allow" masks? How can you compile when --allow is mandatory?

Ie i can see my hosts but i can't connect to them due to acl rules?

Or have i missed a patch?

This aint fixed... not in Debian anyway.

Indeed it is not. Very annoying bug.

As Comment 13 in Issue 42 points out, this bug was fixed in revision 650 and revision 673 but for some reason got left out of the 3.1 release. It is also missing from the current trunk. Perhaps someone did a poor job merging branches at some point and accidentally wiped out the fix?

In any case, since this issue is not getting any response here, I have submitted the patch that fixes this to Ubuntu. It is now available in oneiric as distcc_3.1-4ubuntu2:

https://bugs.launchpad.net/ubuntu/+source/distcc/+bug/809534/comments/3

The bug was fixed after the release of distcc 3.1. The lack of a time machine prevented the fix from being included in the distcc 3.1 release :) The bug is fixed at head and will be included in the next release of distcc. The real problem here is that it's been a long time since the last release. Must be time to do a new release soon...

Yes it would be good to release, as many people have problems with ipv6 and duplicates hosts...

I built from trunk, (after trying the build packages, which happen to not be built with zeroconf support) and while zeroconf now works some of the time, it still doesn't work all of the time (intermittent during a long compile with same parse error)

This problem is not fixed.

The patch was applied to head some time ago. So if it is still not working reliably, another patch will be required. I suggest you open a new bug for the new issue, and if you have a fix for it please attach the patch to the new bug.

Issue 118 has been merged into this issue.