CentOS virt - Oct 2007 - Re: [kvm-devel] [ kvm-Bugs-1807620 ] KVM's --disable-gcc-check doesn't work

Gregory Haskins wrote:
> On Thu, 2007-10-04 at 19:27 +0200, Farkas Levente wrote:
> 
>> ok but now as qemu code was imported into kvm, then it's probably
would
>> be better to witch gcc-4.x?
> 
> Sure.  Are you volunteering? ;)  I'm sure both upstream QEMU
developers,
> KVM developers, and the community using either would be most
> appreciative.  I know I would be.
here i mean that packaging kvm for fedora/redhat/centos using gcc-4.x in
stead of gcc-3.x. if currently there is no reason to use gcc-3.x than i
change all of my spec file.

ps. anyway it's planed to be temporary or permanent to use a qemu fork
for kvm and not try to propagate changes back to the upstream qemu?

-- 
  Levente                               "Si vis pacem para bellum!"

On Thu, 2007-10-04 at 21:49 +0200, Farkas Levente wrote:
> Gregory Haskins wrote:
> > On Thu, 2007-10-04 at 19:27 +0200, Farkas Levente wrote:
> > 
> >> ok but now as qemu code was imported into kvm, then it's
probably would
> >> be better to witch gcc-4.x?
> > 
> > Sure.  Are you volunteering? ;)  I'm sure both upstream QEMU
developers,
> > KVM developers, and the community using either would be most
> > appreciative.  I know I would be.
> 
> here i mean that packaging kvm for fedora/redhat/centos using gcc-4.x in
> stead of gcc-3.x. if currently there is no reason to use gcc-3.x than i
> change all of my spec file.
You *could*, sure.  I have done this for local builds here.  But if you
go that route I would recommend making a patch to KVM so it doesn't fall
back into QEMU mode automatically (today if it can't open the kvm module
it will assume "-no-kvm" like behavior).  Otherwise you will have a
bunch of support calls about why its not working properly should someone
cause the system to fall back.
> 
> ps. anyway it's planed to be temporary or permanent to use a qemu fork
> for kvm and not try to propagate changes back to the upstream qemu?
Ill defer to Avi here, though as I understand it: Things that benefit
the upstream source get pushed...things that are KVM only don't.

Regards,
-Greg
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
URL:
<http://lists.centos.org/pipermail/centos-virt/attachments/20071004/bd1f822b/attachment-0004.sig>

Farkas Levente wrote:
> ps. anyway it's planed to be temporary or permanent to use a qemu fork
> for kvm and not try to propagate changes back to the upstream qemu?
>   
The plan is to submit everything to upstream qemu.

-- 
Any sufficiently difficult bug is indistinguishable from a feature.

Carlo Marcelo Arenas Belon wrote:
> And considering also there might be other yet unknown bugs because of gcc4,
it
> is most likely better to stick to use gcc-3.x, unless there is really no
other
> option.
> 
> For cases where no compat-gcc package is provided (like in OpenSUSE) then
the
> following is IMHO the next best approach to force users to get gcc-3.x
> compiler installed from source just to be able to compile kvm.
This will probably become more of a problem over time.  GCC 3 is
slated to be removed from the next Debian release:
  http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=440425
  http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=440430
> PS. please let me know what you think about it and if interested so I can
> provide the 3 patches (or more if more changes are needed) in a [PATCH] 
> series for easy git-am'ing
> 
> +  if test $check_gcc = "no" ; then
> +    if gcc -v 2>&1 | grep -q 'gcc *version
*4\.[0-3]\.[0-9]'; then
> +      echo "#define USE_GCC4 1" >> $config_h
> +    fi
> +  fi
Something like

#if __GNUC__ >= 4

in the code would be simpler.

-jim