Re: [edk2-devel] [PATCH] .azurepipelines: Enable CI for OvmfPkg and EmulatorPkg

["Ard Biesheuvel" <ard.biesheuvel@linaro.org>]

On Mon, 30 Mar 2020 at 22:56, Laszlo Ersek <lersek@redhat.com> wrote:
>
> On 03/30/20 19:44, Ard Biesheuvel wrote:
> > On Mon, 30 Mar 2020 at 19:11, Sean via Groups.Io
> > <sean.brogan=microsoft.com@groups.io> wrote:
> >>
> >> On Mon, Mar 30, 2020 at 10:04 AM, Ard Biesheuvel wrote:
> >>
> >> Is there any way I could contribute ArmVirtQemu to this? Or would it
> >> be easier if I provided comments/instructions?
> >>
> >> Either way.
> >> Any instructions you provide would be great.  I was going to hack something up for feedback but happy for someone else to do it.   Let me know.
> >
> > OK, so the typical invocation would be
> >
> > qemu-system-aarch64 -M virt -cpu cortex-a57 -m 1024 -net none
> > -nographic -bios .../path/to/QEMU_EFI.fd -hda
> > fat:rw:.../path/to/startup.nsh
> >
> > The only complication compared to OVMF is that there is no separate
> > serial port for debug output vs console output, so everything is going
> > to come out of the same pipe, and grep'ing the console output for
> > meaningful strings may easily result in false positives. (-pflash
> > could be used as well, but doesn't really add anything in this case,
> > and QEMU for ARM has a quirk where pflash images must be exactly 64 MB
> > in size)
>
> I'm begging you not to introduce instances of "-bios" anywhere near
> edk2. Please? :) We need to educate QEMU users, and we need to keep our
> sanity when facing bug reports. Let's not set bad examples anywhere, if
> we can manage.
>
> I think truncate(1) can do what we need for padding, without actually
> allocating those MBs.
>

truncate should work on Linux, but I have no idea how to pad an image
to a certain size on Windows, and I think the idea was to enable both?

In any case, I don't feel as strongly about this as Laszlo does: even
though -bios really shouldn't be used when you are actually installing
an OS into the VM, I don't think it is inappropriate for booting into
a shell and nothing else. Perhaps we could annotate the scripts in a
way that discourages people from adopting it?

Or if there is an easy way to make this work on both Windows and Linux
using -pflash, I obviously wouldn't mind. I just don't feel it is
essential.