Re: [PATCH 1/2] keyval: accept escaped commas in implied option

[Mark Cave-Ayland]

On 11/11/2020 10:53, Daniel P. Berrangé wrote:

> On Wed, Nov 11, 2020 at 05:45:20AM -0500, Paolo Bonzini wrote:
> > This is used with the weirdly-named device "SUNFD,two", so accepting it
> > is also a preparatory step towards keyval-ifying -device and the
> > device_add monitor command.  But in general it is an unexpected wart
> > of the keyval syntax and leads to suboptimal errors compared to QemuOpts:
>
> If "SUNFD,two" is the only wierdly named device, can we just rename
> it to get rid of the comma, and then put validation in QOM to forbid
> commas entirely.  eg rename it to "SUNFD-two"
>
> Just have a targetted hack in vl.c to replace any use of "SUNFD,two"
> with the new name before parsing in keyval, if we care enough about
> back compat for this niche hardware device.

There's also SUNW,tcx I can think of immediately, plus there's a few others mentioned 
in hw/sparc/sun4m.c. The reason these names have commas is simply because that's the 
name of the corresponding node in the PROM device tree.

A quick grep suggests that all these devices are in-built: the selection between 
SUNW,tcx and cg3 framebuffers is decided with the -vga command line option, so 
potentially these could be renamed without any user-visible changes.


ATB,

Mark.