Luigi Gangitano wrote:
Hi Bastian, Adam,
you are saying that squid3's configure should fail if optional features cannot
be built when requested via --enable- switches.
Actually I've seen squid3 behavior in many different projects, so was not
expecting it to do anything different.
Yes Squid upstream believe in that behaviour as well. A fix is underway
to correct this bug. Unfortunately it's deep in the autoconf code, so a
simple patch won't be able to be passed down.
3.1.2 will have the correct build pass/fail you require when
--enable-ipv6 is specified. Might be 10-15 days for 3.1.2 release though.
The upstream fix will not enable Squid-3.1 to build on the failing
buildd, only an update to make it support v6 in the kernel will do that.
But this will at least make the broken builds finish in FTBS.
Il giorno 20/apr/2010, alle ore 19.04, Bastian Blank ha scritto:
The build of Debian packages needs to be stable. It must not use state
into count that depends on the environment except for build-depends and
build-conflicts.
squid3 uses a test to detect if ipv6 is supported on the current system,
which may return false. Also it must support both split-stack and
unified-stack on Linux.
I'm not sure what Bastian is talking about requiring split-stack for.
The test which fails is the one which determines if v6 is able to build
at all and is the pre-dependency test for detecting which stack type to
build. spit-stack in Squid currently fails hard under upstream bug 2753.
Setting to RC because the behaviour is not sufficient fixed.
Ack, and agreed.
Amos Jeffries
Squid Proxy Project
--
To UNSUBSCRIBE, email to [email protected]
with a subject of "unsubscribe". Trouble? Contact [email protected]
