Re: [PATCH] firmware: don't build hvmloader if it is not needed

[Stefano Stabellini <sstabellini@xxxxxxxxxx>]

On Mon, 15 Feb 2021, Jan Beulich wrote:
> On 13.02.2021 14:50, Marek Marczykowski-Górecki wrote:
> > On Fri, Feb 12, 2021 at 06:05:40PM -0800, Stefano Stabellini wrote:
> >> If rombios, seabios and ovmf are all disabled, don't attempt to build
> >> hvmloader.
> > 
> > What if you choose to not build any of rombios, seabios, ovmf, but use
> > system one instead? Wouldn't that exclude hvmloader too?
> 
> Even further - one can disable all firmware and have every guest
> config explicitly specify the firmware to use, afaict.

I didn't realize there was a valid reason for wanting to build hvmloader
without rombios, seabios, and ovfm.


> > This heuristic seems like a bit too much, maybe instead add an explicit
> > option to skip hvmloader?
> 
> +1 (If making this configurable is needed at all - is having
> hvmloader without needing it really a problem?)

The hvmloader build fails on Alpine Linux x86:

https://gitlab.com/xen-project/xen/-/jobs/1033722465



I admit I was just trying to find the fastest way to make those Alpine
Linux builds succeed to unblock patchew: although the Alpine Linux
builds are marked as "allow_failure: true" in gitlab-ci, patchew will
still report the whole battery of tests as "failure". As a consequence
the notification emails from patchew after a build of a contributed
patch series always says "job failed" today, making it kind of useless.
See attached.

I would love if somebody else took over this fix as I am doing this
after hours, but if you have a simple suggestion on how to fix the
Alpine Linux hvmloader builds, or skip the build when appropriate, I can
try to follow up.

Otherwise for now it might be best to just temporarily remove the Alpine
Linux x86 builds from gitlab-ci.

Any thoughts?

Attachment: patchew.email
Description: Text document