Re: [Qemu-devel] Re: [PATCH] Fix checksum writing in signboot.sh

[Anthony Liguori]

Alexander Graf wrote:
> On 02.08.2009, at 13:15, Avi Kivity <address@hidden> wrote:
>
> > On 08/02/2009 01:25 PM, Filip Navara wrote:
> > > > While the patch is good, the code is unreadable.  Can we mandate 
> > > > python for
> > > > such tricks?
> > >
> > > No, please, no! Throwing additional tools at the problem is only going
> > > to make it worse for Windows users. I'm not happy with using sh script
> > > as it already added dependency on coreutils, but at least that's easy
> > > to install. Python is a nightmare compared to that.
> >
> > Is Python really so difficult to install under Windows?  How many 
> > times do you have to click 'Next'?
> >
> > Note that Windows users can usually use prebuilt binaries, so the 
> > 'Next' nightmare only affects a small number of Windows developers.
> >
> > > BTW, for years in ReactOS we had a way to build host tools with host
> > > CC and these tools were written in plain ordinary C. This worked great
> > > for both Windows and Linux builds and also for cross-compiling.
> >
> > But then you have to write those tools in C, which is annoying.
>
> Right. In fact we just switched from C to sh for portability reasons.

The problem is with cross compilers.  Our build system is based around a 
single tool chain and we only do feature probing, sanity checking, 
cflags modifications, etc. on the target tool chain.  If we build and 
run a C program using the host compiler (which is needed in order to be 
able to run the program), things get complicated quickly.

sh is preferred because it's a minimal dependency.  I would be concerned 
about perl or python for the main build because those tools aren't 
available by default for windows.  For something like a rom where we 
ship a default binary, as long as we detected the appropriate tools and 
disabled the build, I think it would be more reasonable.

> I really think we should just make the current code work as is and be 
> done. The script is pretty small and really readable IMHO.

We're going to have to revisit this for pc-bios since it depends on perl 
and it has a similar rom signing tool (biossums).  It's far more 
sophisticated though and it's currently implemented in C.  It may make 
sense to rewrite that tool in python/perl and have a single tool used 
for all of our roms.

We don't need to do this now though.

Regards,

Anthony Liguori