[RFC] QEMU KVM target -> Anthony

[RFC] QEMU KVM target -> Anthony

[Dâniel Fraga]

	Anthony, it has been almost a month since you announced your
patch to remove gcc3 dependency:

http://www.mail-archive.com/kvm-devel-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f@public.gmane.org/msg01386.html

	You promissed that you would change the code and submit it
again.

	I'm just a user but I can't wait to see this code merged, since
I just use gcc 4 and don't want to bother with gcc 3 details.

	Any news on that?

	Thank you very much!

-- 
http://u-br.net
Linux 2.6.20: Homicidal Dwarf Hamster
gcc (GCC) 4.1.2


-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys-and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

Re: [RFC] QEMU KVM target -> Anthony

[Anthony Liguori]

Dâniel Fraga wrote:
> 	Anthony, it has been almost a month since you announced your
> patch to remove gcc3 dependency:
>
> http://www.mail-archive.com/kvm-devel-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f@public.gmane.org/msg01386.html
>
> 	You promissed that you would change the code and submit it
> again.
>   

I'm not entirely sure it's the right thing to do.  The proper solution 
is qops.

I also don't think it's a good idea to merge things into KVM that 
wouldn't be reasonable to integrate into upstream QEMU.  I think this is 
probably the case with this one.

Regards,

Anthony Liguori

> 	I'm just a user but I can't wait to see this code merged, since
> I just use gcc 4 and don't want to bother with gcc 3 details.
>
> 	Any news on that?
>
> 	Thank you very much!
>
>   


-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys-and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

Re: [RFC] QEMU KVM target -> Anthony

[Dâniel Fraga]

On Sun, 18 Mar 2007 11:44:36 -0500
Anthony Liguori <anthony-rdkfGonbjUSkNkDKm+mE6A@public.gmane.org> wrote:

> I'm not entirely sure it's the right thing to do.  The proper
> solution is qops.
> 
> I also don't think it's a good idea to merge things into KVM that 
> wouldn't be reasonable to integrate into upstream QEMU.  I think this
> is probably the case with this one.

	Ok, so for someone who wants to compile qemu with gcc4, what's
the solution? I tried your patch, but it isn't applied correctly. Do
you have an updated version of the patch?

	Thank you.

-- 
http://u-br.net
Linux 2.6.20: Homicidal Dwarf Hamster
gcc (GCC) 4.1.2


-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys-and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

Re: [RFC] QEMU KVM target -> Anthony

[Anthony Liguori]

Dâniel Fraga wrote:
> On Sun, 18 Mar 2007 11:44:36 -0500
> Anthony Liguori <anthony-rdkfGonbjUSkNkDKm+mE6A@public.gmane.org> wrote:
>
>   
>> I'm not entirely sure it's the right thing to do.  The proper
>> solution is qops.
>>
>> I also don't think it's a good idea to merge things into KVM that 
>> wouldn't be reasonable to integrate into upstream QEMU.  I think this
>> is probably the case with this one.
>>     
>
> 	Ok, so for someone who wants to compile qemu with gcc4, what's
> the solution?

You really should use gcc3.  You'll have a much better user experience 
and honestly it's more useful to have the ability to run -no-kvm to 
troubleshoot.

Is there a particular reason why gcc3 isn't available to you?

Regards,

Anthony Liguori

>  I tried your patch, but it isn't applied correctly. Do
> you have an updated version of the patch?
>
> 	Thank you.
>
>   


-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys-and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

Re: [RFC] QEMU KVM target -> Anthony

[Dâniel Fraga]

On Sun, 18 Mar 2007 14:51:47 -0500
Anthony Liguori <anthony-rdkfGonbjUSkNkDKm+mE6A@public.gmane.org> wrote:

> You really should use gcc3.  You'll have a much better user
> experience and honestly it's more useful to have the ability to run
> -no-kvm to troubleshoot.
> 
> Is there a particular reason why gcc3 isn't available to you?

	I like to have only the latest versions of software installed
here. For me, it doesn't make sense to use gcc3 only to be able to
compile qemu, since I'll not use it on any other software.

	Well, to be honest, I think that qemu author should have
migrated to gcc4. I can't understand why someone still uses gcc3 for
development. I mean, the author could take a weekend and fix all the
code to compile with gcc4. Is it so hard? I think it's the natural order
of things.

	I think it's easier for the qemu author to port his code to
gcc4 instead of forcing the entire world to install an obsolete
compiler like gcc3 just to compile his code.

	We should get rid of old things, as fast as possible. Gcc3 has
no future, so why keep using it?

	Besides that, gcc4 offers a lot of options that gcc3
doesn't. It's faster, I can use its vectorizer and optimizes better the
code. If it takes too long, we'll have gcc 4.3 and qemu will be still
requiring gcc3. Don't you think it's absurd?

	Thank you!

-- 
http://u-br.net
Linux 2.6.20: Homicidal Dwarf Hamster
gcc (GCC) 4.1.2


-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys-and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

Re: [RFC] QEMU KVM target -> Anthony

[Jim Paris]

> 	I like to have only the latest versions of software installed
> here. For me, it doesn't make sense to use gcc3 only to be able to
> compile qemu, since I'll not use it on any other software.

Why not install them both?  Most distributions make it quite easy.
For Debian/Ubunty, "apt-get install gcc-3.3" and you'll have it in 60
seconds.  If you want to remove it when you're done, it's just as quick.

> 	Well, to be honest, I think that qemu author should have
> migrated to gcc4. I can't understand why someone still uses gcc3 for
> development. I mean, the author could take a weekend and fix all the
> code to compile with gcc4. Is it so hard? I think it's the natural order
> of things.

I don't think it's appropriate to tell the qemu authors what they
should be doing.  Perhaps the problems are more complicated than you
believe and they thought their time was better spent elsewhere.

> 	We should get rid of old things, as fast as possible.
> Gcc3 has no future, so why keep using it?
> 
> 	Besides that, gcc4 offers a lot of options that gcc3
> doesn't. It's faster, I can use its vectorizer and optimizes better the
> code. If it takes too long, we'll have gcc 4.3 and qemu will be still
> requiring gcc3. Don't you think it's absurd?

GCC 4.x also introduced many new bugs.  For a list of just the known
regressions, see:  http://preview.tinyurl.com/ywvpxv

-jim

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys-and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

Re: [RFC] QEMU KVM target -> Anthony

[Gregory Haskins]

Hi Daniel,
  Without getting into the debate about whether using the older GCC3 is absurd or not ;)  I thought I might point out that you can easily have both coexist peacefully on your system.  Heres what you do:

1) download a gcc3 tarball to your system and unpack
2) run "configure" with the suffix set to something like "-3" and a unique install path like "/usr/local/gcc-3"
3) configure KVM/QEMU build with "configure --qemu-cc=/usr/local/gcc-3/bin/gcc-3"

This is what I do and it works fine.  Hope this helps.
-Greg


>>> Dâniel Fraga <fragabr@gmail.com> 03/18/07 4:53 PM >>>
On Sun, 18 Mar 2007 14:51:47 -0500
Anthony Liguori <anthony@codemonkey.ws> wrote:

> You really should use gcc3.  You'll have a much better user
> experience and honestly it's more useful to have the ability to run
> -no-kvm to troubleshoot.
> 
> Is there a particular reason why gcc3 isn't available to you?

	I like to have only the latest versions of software installed
here. For me, it doesn't make sense to use gcc3 only to be able to
compile qemu, since I'll not use it on any other software.

	Well, to be honest, I think that qemu author should have
migrated to gcc4. I can't understand why someone still uses gcc3 for
development. I mean, the author could take a weekend and fix all the
code to compile with gcc4. Is it so hard? I think it's the natural order
of things.

	I think it's easier for the qemu author to port his code to
gcc4 instead of forcing the entire world to install an obsolete
compiler like gcc3 just to compile his code.

	We should get rid of old things, as fast as possible. Gcc3 has
no future, so why keep using it?

	Besides that, gcc4 offers a lot of options that gcc3
doesn't. It's faster, I can use its vectorizer and optimizes better the
code. If it takes too long, we'll have gcc 4.3 and qemu will be still
requiring gcc3. Don't you think it's absurd?

	Thank you!

-- 
http://u-br.net
Linux 2.6.20: Homicidal Dwarf Hamster
gcc (GCC) 4.1.2


-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys-and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV
_______________________________________________
kvm-devel mailing list
kvm-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/kvm-devel


-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys-and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV
_______________________________________________
kvm-devel mailing list
kvm-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/kvm-devel

Re: [RFC] QEMU KVM target -> Anthony

[Dâniel Fraga]

On Sun, 18 Mar 2007 18:42:48 -0400
"Gregory Haskins" <ghaskins-Et1tbQHTxzrQT0dZR+AlfA@public.gmane.org>
wrote:

> This is what I do and it works fine.  Hope this helps.

	Ok Greg, thanks for the message, but I really would like to be
able to compile qemu using gcc 4.1.2. But thanks anyway.

-- 
http://u-br.net
Linux 2.6.20: Homicidal Dwarf Hamster
gcc (GCC) 4.1.2


-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys-and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

Re: [RFC] QEMU KVM target -> Anthony

[Ed Swierk]

On 3/18/07, Dâniel Fraga <fragabr@gmail.com> wrote:
>         Well, to be honest, I think that qemu author should have
> migrated to gcc4. I can't understand why someone still uses gcc3 for
> development. I mean, the author could take a weekend and fix all the
> code to compile with gcc4. Is it so hard? I think it's the natural order
> of things.
>
>         I think it's easier for the qemu author to port his code to
> gcc4 instead of forcing the entire world to install an obsolete
> compiler like gcc3 just to compile his code.

Please read this email thread if you want to know why your assumption
is incorrect: http://lists.gnu.org/archive/html/qemu-devel/2006-10/msg00175.html

--Ed
-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys-and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV
_______________________________________________
kvm-devel mailing list
kvm-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/kvm-devel

Re: [RFC] QEMU KVM target -> Anthony

[Dâniel Fraga]

On Sun, 18 Mar 2007 18:36:54 -0400
Jim Paris <jim-XrPbb/hENzg@public.gmane.org> wrote:

> Why not install them both?  Most distributions make it quite easy.
> For Debian/Ubunty, "apt-get install gcc-3.3" and you'll have it in 60
> seconds.  If you want to remove it when you're done, it's just as
> quick.

	Just an analogy: it's this kind of thing that allow companies
like Sun and Adobe to take forever to release a 64bit plugin, for
example. Just because people can run the plugin on 32bit... I disagree
with this. I think that if we have a 64 bit machine, we should be able
to use it in pure 64bit mode and the software should follow the
hardware evolution.

	The same applies for gcc. If I have a better gcc, I should be
able to compile qemu with it, instead of using the old gcc3.

	gcc is the second most important piece of software in Linux
world, so all the software should follow gcc evolution, since there
isn't another viable compiler (icc isn't a free option).

> I don't think it's appropriate to tell the qemu authors what they
> should be doing.  Perhaps the problems are more complicated than you
> believe and they thought their time was better spent elsewhere.

	Ok, but at least I think that if someone wants to do something
right, it should put all his/her effort so it will be done right,
otherwise it will result in problems like these. And Anthony's patch
would suffice for kvm's purpose.

	But you're correct. i can't blame qemu author if he doesn't
want to support gcc4. I'm just on his hands... At the same time, those
who don't want to use gcc3 will be unable to use kvm because kvm
depends on qemu...

	I just think that if KVM uses a patched qemu, it does not mind
if we have patches for the qemu distributed in the kvm tarball, since
it will just be used together with KVM. And as far as I know, the
official qemu won't support kvm, right? So you'll have to maintain your
own qemu version separately.

> GCC 4.x also introduced many new bugs.  For a list of just the known
> regressions, see:  http://preview.tinyurl.com/ywvpxv

	Ok, but when qemu will fully support gcc4? Does it mean we will
have to use gcc3 forever?

	Thanks.

-- 
http://u-br.net
Linux 2.6.20: Homicidal Dwarf Hamster
gcc (GCC) 4.1.2


-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys-and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

Re: [RFC] QEMU KVM target -> Anthony

[Dâniel Fraga]

On Sun, 18 Mar 2007 16:08:21 -0700
"Ed Swierk" <eswierk-t4QvRbHP2kpBDgjK7y7TUQ@public.gmane.org> wrote:

> Please read this email thread if you want to know why your assumption
> is incorrect:
> http://lists.gnu.org/archive/html/qemu-devel/2006-10/msg00175.html

	I read everything. The last message by Paul Brook:

"- qemu makes assumptions about the layout of the code gcc generates.
This works by chance on older gcc. This effects all hosts, and is not a gcc bug.

- qemu reserves several registers for its own use. On architecturally crippled 
hosts (ie. x86) this means we hit really obscure gcc bugs on x86 because gcc 
runs out of registers. This is a gcc bug, but is also relatively easy to 
workaround".

	He even proposes a solution:

https://nowt.dyndns.org/

	And even this way, nobody agreed with each other...

	So I ask: if Paul Brook could fix this, why it isn't merged? A
solution should be applied. What we can't do is keep discussing forever
this and the previous message is realy annoying:

"Basically, gcc changed in a way that broke qemu.  There's been an open
bug report in gcc ever since, but the GCC developers really aren't interested in 
backwards compatability.  (Heck, gcc 4.0 breaks building bash 2.05b).  The 
qemu developers aren't interested in applying ugly patches to support gcc 4.x 
until gcc 3.x becomes so obsolete nobody ships it anymore.  (And considering 
that there are still some niche embedded boards that have hacked up versions 
of gcc 2.95 targeting them and nothing else, I wouldn't be surprised if in 
five years we have your main compiler and the compiler to build qemu, ala 
kgcc under Red Hat 7.  *shrug*)"

	It's like saying: "we don't care. Keep using gcc3". :(

-- 
http://u-br.net
Linux 2.6.20: Homicidal Dwarf Hamster
gcc (GCC) 4.1.2


-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys-and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

Re: [RFC] QEMU KVM target -> Anthony

[Anthony Liguori]

Dâniel Fraga wrote:
> On Sun, 18 Mar 2007 16:08:21 -0700
> "Ed Swierk" <eswierk-t4QvRbHP2kpBDgjK7y7TUQ@public.gmane.org> wrote:
>
>   
>> Please read this email thread if you want to know why your assumption
>> is incorrect:
>> http://lists.gnu.org/archive/html/qemu-devel/2006-10/msg00175.html
>>     
>
> 	I read everything. The last message by Paul Brook:
>
> "- qemu makes assumptions about the layout of the code gcc generates.
> This works by chance on older gcc. This effects all hosts, and is not a gcc bug.
>
> - qemu reserves several registers for its own use. On architecturally crippled 
> hosts (ie. x86) this means we hit really obscure gcc bugs on x86 because gcc 
> runs out of registers. This is a gcc bug, but is also relatively easy to 
> workaround".
>
> 	He even proposes a solution:
>
> https://nowt.dyndns.org/
>
> 	And even this way, nobody agreed with each other...
>
> 	So I ask: if Paul Brook could fix this, why it isn't merged? A
> solution should be applied. What we can't do is keep discussing forever
> this and the previous message is realy annoying:
>   

This is qops which is why I said "qops is the right solution" in my 
previous note.  It's a huge, fundamental change to QEMU so it needs some 
time being developed outside of the main tree.

I understand that it's frustrating to depend on gcc3.  Please realize 
though that if it was simple to fix than it would be fixed and more 
importantly, that the work is being done to fix the situation.

Regards,

Anthony Liguori

> "Basically, gcc changed in a way that broke qemu.  There's been an open
> bug report in gcc ever since, but the GCC developers really aren't interested in 
> backwards compatability.  (Heck, gcc 4.0 breaks building bash 2.05b).  The 
> qemu developers aren't interested in applying ugly patches to support gcc 4.x 
> until gcc 3.x becomes so obsolete nobody ships it anymore.  (And considering 
> that there are still some niche embedded boards that have hacked up versions 
> of gcc 2.95 targeting them and nothing else, I wouldn't be surprised if in 
> five years we have your main compiler and the compiler to build qemu, ala 
> kgcc under Red Hat 7.  *shrug*)"
>
> 	It's like saying: "we don't care. Keep using gcc3". :(
>
>   


-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys-and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

Re: [RFC] QEMU KVM target -> Anthony

[Dâniel Fraga]

On Sun, 18 Mar 2007 20:39:33 -0500
Anthony Liguori <anthony-rdkfGonbjUSkNkDKm+mE6A@public.gmane.org> wrote:

> This is qops which is why I said "qops is the right solution" in my 
> previous note.  It's a huge, fundamental change to QEMU so it needs
> some time being developed outside of the main tree.
> 
> I understand that it's frustrating to depend on gcc3.  Please realize 
> though that if it was simple to fix than it would be fixed and more 
> importantly, that the work is being done to fix the situation.

	Ok. I'll wait for the fix.

	Thank you and everybody who answered!

-- 
http://u-br.net
Linux 2.6.20: Homicidal Dwarf Hamster
gcc (GCC) 4.1.2


-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys-and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV