Openwall GNU/*/Linux - a small security-enhanced Linux distro for servers
[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Date: Sun, 29 Apr 2012 06:31:20 +0400
From: Solar Designer <solar@...nwall.com>
To: owl-dev@...ts.openwall.com
Subject: Re: new gcc and packages

Hi Mesut,

On Thu, Apr 26, 2012 at 05:26:29PM +0300, Mesut Can G?rle wrote:
> I packaged new gcc(4.7.0). I updated the patch and applied Solar Designer's
> recommended patch. I use new gcc (4.7.0) and try to make buildworld.

Thank you!

> Failing packages for both x86_64 and i686 :
> 
> -kernel
> 
> This issue is talked on http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50876.
> They provide a patch for this. But first the patch must be confirmed.

Your build logs show:

gcc: error: unrecognized command line option '-melf_i386'

gcc: error: unrecognized command line option '-melf_x86_64'

While this might be caused by the bug you referenced, I am not sure that
this is the case.  Have you tried applying that patch?  If so, was the
error gone?

> -nmap
> 
> Needs OpenSSL 0.9.8 or later for SHA-256 support. Default OpenSSL on
> Openwall is 0.9.7. But when I use OpenSSL in the cvs, nmap is successfully
> built.

Actually, we moved to OpenSSL 1.0.0 in Owl-current over a year ago, so
I am puzzled by this comment - did you start with this testing on some
mix of Owl 3.0 and Owl-current?

> All packages (except kernel) can be built when dependencies are matched.

Again, the reference to having to match their dependencies suggests that
you were not building on Owl-current.  I think the test builds will need
to be redone on cleanly installed Owl-current (e.g., using one of our
precreated OpenVZ templates of the Owl-current userland - maybe
chroot'ing into that tree if you don't have Owl as your "host" Linux
distro and you don't use OpenVZ there).

> All related files can be accessible from http://mesutcang.net23.net/owl/gcc.

Thanks.  I looked at some build logs there, and also at gcc.spec.diff.
Here are some comments on the latter:

As discussed before, you're switching from gcc-core and gcc-g++ tarballs
to the larger cumulative gcc tarball.  This change is not acceptable for
committing into our tree.

The gcc-4.7.0-owl-wrong-code-generation.diff would need to have
something other than "owl" in its filename: we'd need to credit the
patch's original author (or project/team/company).  That said, since it
is not clear whether we'd be committing an update to 4.7.0 at all (maybe
we'd instead update to 4.7.1+ right away), we don't have to bother with
this right now.  That patch file would be gone with 4.7.1+ anyway.

--enable-languages=c,c++ was discussed before (needed because of the
source tarballs change, OK to have either way), but why are you adding
--disable-bootstrap and removing bootstrap-lean?  (This is merely a
question.  The change may well be appropriate, but we need to know why
it is made.)

I understand that some of the above might require substantial time (such
as for re-doing the test builds), so I am not really asking you to do
that.  Please just provide answers that you readily have.  I understand
that you need to re-focus on your GSoC project with PacketFence now, so
we're likely to proceed further with the gcc update on our own - but
it's helpful to have the full set of readily-available answers from you
such that we don't waste time on things that you've already tried.

Overall, I am happy to know that only the kernel did not build.

Thanks again,

Alexander

Powered by blists - more mailing lists

Your e-mail address:

Powered by Openwall GNU/*/Linux - Powered by OpenVZ