Test GRUB2 compiled by MingW

Test GRUB2 compiled by MingW

[Hanzac Chen]

Hi,

After I get the GRUB2 successfully compiled by MingW (since it now can
handle binary output with its latest version of binutils), I can't boot
it after making the image according to the README.

1. boot it with QEMU, got "GRUB loading kernel..." then it seemed to be
dead.
2. boot it with Bochs, got a little bit further and displayed "no upper
memory".

BTW: When generate the kernel_syms.lst, every symbol should have a
underline at the beginning when compiling with some kind of compiler
like MingW.

Bye,
Hanzac

Re: Test GRUB2 compiled by MingW

[Marco Gerards]

Hanzac Chen <hanzac@gmail.com> writes:

Hi Hanzac,

> After I get the GRUB2 successfully compiled by MingW (since it now can
> handle binary output with its latest version of binutils), I can't boot
> it after making the image according to the README.
>
> 1. boot it with QEMU, got "GRUB loading kernel..." then it seemed to be
> dead.
> 2. boot it with Bochs, got a little bit further and displayed "no upper
> memory".

This is not normal.  Perhaps there is still a problem with the
compiler or linker or so.  MingW is just a modified gcc, right?

> BTW: When generate the kernel_syms.lst, every symbol should have a
> underline at the beginning when compiling with some kind of compiler
> like MingW.

Is there a way to fix this so everything remains working on the
supported OS'es and making things work on MingW as well?

--
Marco

Re: Test GRUB2 compiled by MingW

[Yoshinori K. Okuji]

On Friday 01 April 2005 06:49 pm, Marco Gerards wrote:
> This is not normal.  Perhaps there is still a problem with the
> compiler or linker or so.  MingW is just a modified gcc, right?

I think this is normal. IIRC, Windows does not use ELF, while GRUB 2 depends 
on ELF.

Okuji

Re: Test GRUB2 compiled by MingW

[Hanzac Chen]

Hi,

Yoshinori K. Okuji wrote:
> On Friday 01 April 2005 06:49 pm, Marco Gerards wrote:
> 
>>This is not normal.  Perhaps there is still a problem with the
>>compiler or linker or so.  MingW is just a modified gcc, right?
> 
> 
> I think this is normal. IIRC, Windows does not use ELF, while GRUB 2 depends 
> on ELF.

Yes, MingW generates COFF and if using COFF, the GCC/binutils can't use
some of GAS's directive, like section switch. (I found GRUB using that
in its module loading mechanism.)

Maybe I still don't understand it well.

How much else does GRUB 2 depend on ELF?

BYE,
Hanzac

Re: Test GRUB2 compiled by MingW

[Hanzac Chen]

Hi,

Marco Gerards wrote:
> 
> This is not normal.  Perhaps there is still a problem with the
> compiler or linker or so.  MingW is just a modified gcc, right?
OK, I see, next time I'll compile it under Linux GCC, to compare that.

I need to know if the problem is caused by the kernel unable to load its
modules. I'm trying to understand how GRUB loads its modules.

BTW: MingW uses PECOFF object format, so it may have some limitations.
>>BTW: When generate the kernel_syms.lst, every symbol should have a
>>underline at the beginning when compiling with some kind of compiler
>>like MingW.
> 
> 
> Is there a way to fix this so everything remains working on the
> supported OS'es and making things work on MingW as well?
Modifying genkernsyms.sh is enough, I think.
Anyway, this is not a big problem.

BYE,
Hanzac

Re: Test GRUB2 compiled by MingW

[Marco Gerards]

Hanzac Chen <hanzac@gmail.com> writes:

> Yes, MingW generates COFF and if using COFF, the GCC/binutils can't use
> some of GAS's directive, like section switch. (I found GRUB using that
> in its module loading mechanism.)
>
> Maybe I still don't understand it well.
>
> How much else does GRUB 2 depend on ELF?

A lot.  The modules are all ELF files and the core image has an ELF
loader.  It would be the best if you could somehow produce ELF
executables.

Otherwise another module loader should be implemented for COFF.

Thanks,
Marco