Luigi Gangitano wrote:
> Il giorno 28/lug/2010, alle ore 11.43, Amos Jeffries ha scritto:
>
>> Luigi Gangitano wrote:
>>> Il giorno 20/lug/2010, alle ore 02.40, Amos Jeffries ha scritto:
>>>> On Mon, 19 Jul 2010 18:17:20 +0200, Luigi Gangitano <luigi_at_debian.org>
>>>> wrote:
>>>>> Sorry again for bothering you, guys, but next debian release will ship
>>>>> with 3.1.3 (and all of it's bugs) if I can't fix this on alpha.
>>>>>
>>>>> Debian policy states that building result should not depend on
>>>>> build-machine configuration (and enabling ipv6 locally is considered so)
>>>>> and build-machine admins will not fix this for us.
>>>> You mean ia64? I'm getting the impression they need --disable-ipv6. If it
>>>> were not for the fact that 3.1.1 initially built just fine with these same
>>>> checks and had IPv6 working.
>>> ia64, right! :-) (have other issues on alphas). Until 3.1.4 build were fine since there was no build time stack-test. Headers are correct on build machines. I'd prefer to void --disable-ipv6 since this is only a build-time issue. Once built with IPv6 support, it would easily work on every other Debian ia64.
>>> See https://buildd.debian.org/fetch.cgi?pkg=squid3;ver=3.1.5-2;arch=ia64;stamp=1278518977 for build logs.
>>>> I'm working on the alternative now. Adding an enable_ipv6 run-time option
>>>> and moving the stack-tests into Squid startup. Until we have split-stack
>>>> support we are forced to default to auto-disabling underneath certain users
>>>> every time they start Squid. Which was the initial policy violation IIRC.
>>> This would be a lot better! As for another intermediate solution, could we add an --with-ipv6-stack=*** option to configure that works around stack-test?
>> Hi Luigi,
>> The 3.1.5.1 beta package contains another attempt at getting this to work.
>>
>> A configurable enable_* option proved intractable since v6 needs to be known prior to parsing the config file and we can't permit changing it mid-config. Instead ...
>>
>> The v6 capability probing is now done on the user installed machine on startup. The --enable-ipv6 option decides on whether to build the probe code now or whether to hard-code it to produce off.
>
> Thanks for this fix.
>
>> PS: it's a beta because of this v6 change being new and also a few changes in the build process that I'm not entirely certain of being stable yet. If it gets past the buildd testing okay there should be no unexpected surprises.
>
> In the meanwhile the next debian release has been fixed, so I'm now trying to push 3.1.6 to it. :-(
>
Sorry should have pointed this out to you end of last week:
http://www.squid-cache.org/Versions/v3/3.1/changesets/squid-3.1-10063.patch
As far as I know its the last of the critical fixes missed out of 3.1.6.
3.1.7 is due out in some few days now, but had a lot of portability
fixes for other OS and HTTP/1.1 polishing which I don't think will get
past the Debian minimal-change criteria.
Amos
-- Please be using Current Stable Squid 2.7.STABLE9 or 3.1.6 Beta testers wanted for 3.2.0.1Received on Tue Aug 17 2010 - 07:17:46 MDT
This archive was generated by hypermail 2.2.0 : Tue Aug 24 2010 - 12:00:05 MDT