Re: [PATCH 00/13] misc: Rename 'softmmu' -> 'system'

[Philippe Mathieu-Daudé]

On 4/10/23 15:49, Paolo Bonzini wrote:
> On Wed, Oct 4, 2023 at 3:41 PM Claudio Fontana <cfontana@suse.de> wrote:
> > On 10/4/23 14:37, Thomas Huth wrote:
> > > On 04/10/2023 14.33, Daniel P. Berrangé wrote:
> > > > Am I mis-understanding what you mean by 'finishes' here, as
> > > > I see many references to softmmu remaining
> > > > In particular under configs/
> > > >
> > > > I was also hoping it meant that we'd be changing configure
> > > > to allow
> > > >
> > > >       configure --target-list=x86_64-system
> > > >       configure --target-list=x86_64-vm
> > > >
> > > > for less typing
> > >
> > > Maybe we should also bikeshed about the naming first... "system" is a quite
> > > overloaded word in this context already, and "vm" sounds rather like
> > > hardware-accelerated stuff ... what about using something like "sysemu"? Or
> > > "fullsys" for "full system emulation" (in contrast to "user space"-only
> > > emulation)?
>
> I agree that changing other remnants should be done right
> after this patch, for example $softmmu in configure. Changing
> all targets is a very large and very user-visible change, it is
> required but it should be planned very well.

As usual I should have been more verbose in my cover.

This series focus on the C code (and travis) to avoid misuse
of 'softmmu'.

Yes I agree it would be nice to also rename the binaries, but
this is a buildsys change (even if we use symlinks/aliases for
some deprecation period). This is not a tiny patch, and requires
thoughts.

On one hand I'd rather not rush renaming binaries -- annoying
users -- without a clear long term plan (which I'm not clear about).

On another hand I wouldn't delay Richard user-mode work.
Renaming binaries seems orthogonal to me, and could be done
later IMO.

> As to the actual target names, I think system is the only
> consistent choice since we have --enable/--disable-system
> (as pointed out by Claudio) and qemu-system-*.  sysemu
> may make a little more sense in the codebase (we have
> include/sysemu after all), but maybe that ship has sailed
> since we have many occurrences of "system", for example
> system_ss and other related sourcesets.
>
> Paolo