On Thu, 2020-05-07 at 10:52 +0200, Kamil Paral wrote:
On Thu, May 7, 2020 at 8:48 AM Lukas Ruzicka
<lruzicka(a)redhat.com> wrote:
> Hello,
> I agree, that user switching as described in this proposal is something
> that should work really well, because it can obviously change the user
> experience in some cases.
> I only have a problem with the following statement:
>
> The switching mechanism must correctly attempt the requested operation.
>
> As I understand it, it might be enough if the operation is "attempted"
> only? This sounds vague and weak to me. I think we need something stronger
> here, like "it performs the requested operation" or something similar.
>
It's similar to e.g. shutdown criterion [1] where the "shutdown mechanism
must correctly request a shutdown from the system firmware" or storage
resize criterion [2] where "installer mechanism for resizing storage
volumes must correctly attempt the requested operation".
The important word in all these cases is "correctly". It implies that the
software part, the calling part, must be implemented correctly. But because
it deals with hardware and low level firmware and drivers, it admits that
that part might not work correctly in all configurations. Which is then
expanded in the next sentence.
I simply got my inspiration from our other criteria when writing this one.
It doesn't have to stay this way. But I think the "correctly
attempt/request" wording is quite fitting here and it's not weak (at least
not towards our software stack, which is the part which we can control
reasonably well).
I actually would agree with Lukas here. We use the "attempt" wording as
a kind of weasel formula when there's a part of the stack we don't want
to support. That's why we use it in the resize case: there's a lot of
complexity to disk resizing and it's a thing that just doesn't always
work, so we don't want to block on the resize itself but rather block
on anaconda correctly requesting it. Shutdown is similar: we are
blocking on what's under our control, the OS side of things. If there's
a system firmware bug that prevents it shutting down properly, that
isn't our problem.
I think we don't need the weasel form in this criterion as we actually
do want to block on user switching *fully working*. So I think I'd
agree with Lukas (and Brandon) that this can be changed to "perform",
but keeping the reference to conditional violations if you like - but
I'd change the link to this part of the blocker bug FAQ:
https://fedoraproject.org/wiki/Blocker_Bug_FAQ#What_about_hardware_and_lo...
which is what we usually use as the target for such references.
--
Adam Williamson
Fedora QA Community Monkey
IRC: adamw | Twitter: AdamW_Fedora | XMPP: adamw AT happyassassin . net
http://www.happyassassin.net