I don't think there's any benefit in constructing a
DeviceTree and then
making changes in it for something simple like that. Maybe in the future
when there are 'part' and 'fs' libblockdev's plugins (not a high
priority now, though) we could use it here.
There's not necessarily any benefit to me using devicetree for this
except for the fact that it's code that already exists.
What about using less memory to also test everything (including
zram)
works even with less RAM? From my experience the installation with less
RAM doesn't take significantly more time.
I used to have these tests consume 1 GB of memory each, and that was
enough to make them work most of the time. However I didn't want my GUI
tests to fail because of not enough memory, so that is why they are
currently so high.
I could certainly add a test case for lower memory situations, though.
In fact it would be possible to write a test case that ran over and over
again with less memory until it failed, at which point we'd know what
the minimum requirement was.
Do we have a test for the Progress hub behaving properly when spokes
are
not completed, the "Finish configuration" button appears but
insensitive, then they are completed and the "Finish configuration"
button can/should be clicked? I've never seen it broken, but I'm not
sure how much it is tested because people usually enter rootpw or user
information during package installation.
This is exactly what my first test does.
- Chris