How reliable is GCC-3.3.1 wrt building mipsel-linux kernel?

How reliable is GCC-3.3.1 wrt building mipsel-linux kernel?

[David Daney]

The subject line kind of says it all.

We are running linux 2.4.18 on a mips4Kc core (ATI Xilleon 225) and find 
it to be quite stable when compiled with gcc 2.96/binutils 2.11.92.0.10

When the kernel is compiled with gcc 3.3.1/binutils 2.14.90.0.5 it also 
seems to be quite stable, except for when one certian driver is used 
(basically an mpeg decoder driver).  Under certian conditions the system 
seems to "freeze" (no messages printed anywhere and only a hard reset 
will recover).

Yeah that is a good bug report...

But my main question is this:  Have other people experienced 
miscompilation (ie bad code generation) with gcc 3.3.1?

One thing I am aware of is that if -fno-common is not used, bad code is 
generated for accessing some large structures.  But I we use -fno-common 
for all compilation.

I am trying to figrue out if I should be looking more at bugs in the 
driver, or if I should give up on gcc 3.3.1 and be done with it.

Thanks in advance for any insight.

David Daney.

Re: How reliable is GCC-3.3.1 wrt building mipsel-linux kernel?

[Daniel Jacobowitz]

On Tue, Nov 18, 2003 at 05:40:31PM -0800, David Daney wrote:
> The subject line kind of says it all.
> 
> We are running linux 2.4.18 on a mips4Kc core (ATI Xilleon 225) and find 
> it to be quite stable when compiled with gcc 2.96/binutils 2.11.92.0.10
> 
> When the kernel is compiled with gcc 3.3.1/binutils 2.14.90.0.5 it also 
> seems to be quite stable, except for when one certian driver is used 
> (basically an mpeg decoder driver).  Under certian conditions the system 
> seems to "freeze" (no messages printed anywhere and only a hard reset 
> will recover).
> 
> Yeah that is a good bug report...
> 
> But my main question is this:  Have other people experienced 
> miscompilation (ie bad code generation) with gcc 3.3.1?
> 
> One thing I am aware of is that if -fno-common is not used, bad code is 
> generated for accessing some large structures.  But I we use -fno-common 
> for all compilation.
> 
> I am trying to figrue out if I should be looking more at bugs in the 
> driver, or if I should give up on gcc 3.3.1 and be done with it.
> 
> Thanks in advance for any insight.

We haven't had problems here.

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer

Re: How reliable is GCC-3.3.1 wrt building mipsel-linux kernel?

[Ralf Baechle]

On Tue, Nov 18, 2003 at 05:40:31PM -0800, David Daney wrote:

> The subject line kind of says it all.
> 
> We are running linux 2.4.18 on a mips4Kc core (ATI Xilleon 225) and find 
> it to be quite stable when compiled with gcc 2.96/binutils 2.11.92.0.10
> 
> When the kernel is compiled with gcc 3.3.1/binutils 2.14.90.0.5 it also 
> seems to be quite stable, except for when one certian driver is used 
> (basically an mpeg decoder driver).  Under certian conditions the system 
> seems to "freeze" (no messages printed anywhere and only a hard reset 
> will recover).
> 
> Yeah that is a good bug report...
> 
> But my main question is this:  Have other people experienced 
> miscompilation (ie bad code generation) with gcc 3.3.1?

Quite frequently using a new, possibly more agressive compiler triggers
bugs in the kernel code ...

  Ralf

Re: How reliable is GCC-3.3.1 wrt building mipsel-linux kernel?

[David Daney]

Ralf Baechle wrote:

>On Tue, Nov 18, 2003 at 05:40:31PM -0800, David Daney wrote:
>  
>
>><...>
>>But my main question is this:  Have other people experienced 
>>miscompilation (ie bad code generation) with gcc 3.3.1?
>>    
>>
>
>Quite frequently using a new, possibly more agressive compiler triggers
>bugs in the kernel code ...
>
>  Ralf
>  
>
That's the whole point of my question.

Which options have other people used with gcc 3.3.1 with good results?

David Daney.

Re: How reliable is GCC-3.3.1 wrt building mipsel-linux kernel?

[Jan-Benedict Glaw]

[-- Attachment #1: Type: text/plain, Size: 1052 bytes --]

On Wed, 2003-11-19 16:29:43 -0800, David Daney <ddaney@avtrex.com>
wrote in message <3FBC0AF7.90600@avtrex.com>:
> Ralf Baechle wrote:
> >On Tue, Nov 18, 2003 at 05:40:31PM -0800, David Daney wrote:

> Which options have other people used with gcc 3.3.1 with good results?

It's a question of what you call "good results". If there are bugs in
the kernel sources which only show up with a really aggressive HEAD
compiler, then IMHO it's a good result to see the compiled kernel crash,
just because there actually *is* a bug.

Companies however tend to accept a slower/more bloated/whatever software
(produced by an older compiler) in order not to start hunting down the
remaining (and possibly hard to find) bugs...

MfG, JBG

-- 
   Jan-Benedict Glaw       jbglaw@lug-owl.de    . +49-172-7608481
   "Eine Freie Meinung in  einem Freien Kopf    | Gegen Zensur | Gegen Krieg
    fuer einen Freien Staat voll Freier Bürger" | im Internet! |   im Irak!
   ret = do_actions((curr | FREE_SPEECH) & ~(NEW_COPYRIGHT_LAW | DRM | TCPA));

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 189 bytes --]