Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20110912164317.GC3376@openwall.com>
Date: Mon, 12 Sep 2011 20:43:17 +0400
From: Solar Designer <solar@...nwall.com>
To: owl-dev@...ts.openwall.com
Subject: Re: gcc 4.6.x and kernel-2.6.18-ovz

Vasiliy,

On Mon, Sep 12, 2011 at 04:11:14PM +0400, Vasiliy Kulikov wrote:
> and I suspect we'll get more of weird build bugs (and runtime bugs
> perhaps) during RHEL6 kernel updates too as RHEL6 has gcc 4.4.5 (DW says
> so), but we'll have 4.6.1.

Yet I think we need to try this combination.  I am optimistic about us
being able to get it to work reliably with little effort.  Of course, I
could be wrong.

I dislike using non-latest-stable gcc in Owl 4.0, and there's no other
kernel version we can use currently while continuing to include OpenVZ
support.

> Kir Kolyshkin suggested to use the same gcc version as RHELx has:
> 
> http://bugzilla.openvz.org/show_bug.cgi?id=1801
> 
> "The rule of thumb here is this: RHELx kernels are best to be compiled by gcc
> from RHELx, otherwise we can get bad results (and it happened before)."

Indeed.  Yet we're quite successful at using those RHEL5'ish kernels
with gcc 3.4.5 instead of 4.1.2.  Yes, we're patching the kernel for
that, but those changes are trivial and not too numerous.

I think 4.1.2 vs. 3.4.5 is a more significant difference than 4.4.5 vs.
4.6.x.  So the latter will likely be even easier for us.

> Probably we should use gcc 4.4.x for Owl 4.0?

No, I'd rather not, unless we run into serious difficulties using
RHEL6/OpenVZ kernels with gcc 4.6.x.

> Probably even not patch
> the 2.6.18 kernel and suggest users to upgrade gcc and the kernel at the
> same time.

This is acceptable if we have to.  But I'm afraid it'd delay
introduction of gcc 4.6.x into Owl-current a lot.  Perhaps even the
bootloader on CD change will need to happen first, then.

> Or patch the kernel and humble with possible runtime bugs
> until we upgrade it to RHEL6/OpenVZ, but this more time consuming.

This is preferable as a development-only milestone, if it is not too
time-consuming.  Perhaps try to spend a couple of hours and see if
you're able to get the kernel to build?  If we choose to use this
approach, we'll need to inform our users of in-deep-development status
of Owl-current.

Thanks,

Alexander

Powered by blists - more mailing lists

Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.