Re: [Qemu-devel] [PATCH 1/4] machine: conversion of QEMUMachineInitArgs to MachineState

[Marcel Apfelbaum <marcel.a@redhat.com>]

On Fri, 2014-05-16 at 20:33 +0200, Andreas Färber wrote:
> Am 16.05.2014 16:39, schrieb Igor Mammedov:
> > On Wed,  7 May 2014 17:42:57 +0300
> > Marcel Apfelbaum <marcel.a@redhat.com> wrote:
> > 
> >> Total removal of QEMUMachineInitArgs struct. QEMUMachineInitArgs's fields
> >> are copied into MachineState. Removed duplicated fields from MachineState.
> >>
> >> All the other changes are only mechanical refactoring, no semantic changes.
> >>
> >> Signed-off-by: Marcel Apfelbaum <marcel.a@redhat.com>
> >> ---
> >>    - I am perfectly aware that patches touching a lot of files
> >>      are not desirable, but this one is a very simple replacement
> >>      patch:
> >>        QEMUMachineInitArgs -> MachineState
> >>        args -> ms
> >>    - This is the simplest way to get rid of QEMUMachineInitArgs fast.
> >>
> > Patch doesn't apply cleanly on current master
> 
> patch -p1 applied all hunks ...
> 
> > 
> > [...]
> > 
> >>  
> >>  #define PC_Q35_MACHINE_OPTIONS \
> >> diff --git a/hw/i386/xen_machine_pv.c b/hw/i386/xen_machine_pv.c
> >> index 9adb57f..fb7a817 100644
> >> --- a/hw/i386/xen_machine_pv.c
> >> +++ b/hw/i386/xen_machine_pv.c
> >> @@ -28,11 +28,11 @@
> > ^^^ file was moved
> 
> ... and prompted for the hw/xenpv/xen_machine_pv.c path.
> 
> > I've reviewed all targets, it looks good.
> > +1 to afaerber's comment s/ms/machine/ 
> 
> 4 "args->" were missed in exynos4_boards.c debug output.
> 
> for f in hw/*/*.c; do
>   sed -i 's/*ms\\)/*machine\\)/g' $f
>   sed -i 's/*ms,/*machine,/g' $f
>   sed -i 's/ ms->/ machine->/g' $f
>   sed -i 's/(ms->/(machine->/g' $f
>   sed -i 's/!ms->/!machine->/g' $f
>   sed -i 's/(ms)/(machine)/g' $f
>   sed -i 's/(ms,/(machine,/g' $f
>   sed -i 's/ ms,/ machine,/g' $f
>   sed -i 's/, ms)/, machine)/g' $f
> done
> git checkout -- hw/ppc/ppc4xx_pci.c # false "(ms," positive
> 
> highbank.c had a conflicting "machine" argument -> machine_id.
> 
> e500.c needed tweaking due to passing ms in a struct. And it was
> shallow-copying *ms, which must not be done for QOM. Passing the pointer
> instead.

Thanks for the help, Andreas!
Marcel

> 
> Andreas
>