Re: [PATCH qemu-server 1/2] tests: improve multiarch build support

[Dominik Csapak <d.csapak@proxmox.com>]

On 2/3/26 3:59 PM, Thomas Lamprecht wrote:
> Am 03.02.26 um 15:21 schrieb Dominik Csapak:
>> The CPUConfig module sets the local 'host_arch' on loading, so instead
>> to try to mock all subs that access this variable, provide a setter only
>> intended for tests (and warn when using this).
> 
> Would be nicer to get a local get_host_arch that just return's the PVE::Tools'
> method result, as we then could override the getter without having to worry
> about this method being misused in the future by accident.
> 
> Moving the CPU model initialization into a sub, like suggested in my reply
> to Fiona's patches,  and redoing that on a simulated switch of the host
> architecture would be also the slightly nicer interface IMO.
> 

ok, makes sense, i'd do that in a v2

>>
>> Also, we have to override the 'get_host_arch' in the Helpers module too.
> 
> The helper module has no such method? There is only get_vm_arch there and
> some usage of PVE::Tools' get_host_arch? Or is it because we import the one
> from PVE::Tools there?
> 

yes, exactly. The importing and using of a sub means that when mocking
we have to mock it in the module we're using it.

alternatively, we could write out 'PVE::Tools::get_host_arch' everywhere
without importing, then we'd only have to mock once
(but makes the code slightly less readable)