Re: [RFC] minimal build option for configure

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. :-(

> PPS: what do you think of adding Suggests on miredo to the squid3 package?

ATM, Suggested packages are installed by default by aptitude, so maybe force miredo on every squid3 installation is a bit overkill. What about a note in README.Debian pointing to miredo?

Regards,

L

--
Luigi Gangitano -- <luigi_at_debian.org> -- <gangitano_at_lugroma3.org>
GPG: 1024D/924C0C26: 12F8 9C03 89D3 DB4A 9972  C24A F19B A618 924C 0C26