Size tests fix

[vpodzime]

This should fix the recent failures of tests because of libbytesize's translations being finally in place.

[dwlehman]

Wow, how did I not notice that libbytesize doesn't use ROUND_HALF_UP (or even support it)?

My brain struggles mightily with the idea that round_to_nearest('10.3') would return 11 as opposed to 10 given that 10.3 is more than twice as close to 10 than to 11.

Since we are dealing with size quantities here I can see an argument for rounding up in certain situations, but it is still a mental hurdle for me. If the size represents a quantity used, rounding up seems to make sense (so that we do not under-estimate how much we have taken). But what if the size represents a quantity of free space? Rounding up seems to be a bad idea for this (because it leads us to over-estimate how much we have available). And what if we're relating two complementary sizes? ROUND_HALF_UP makes those kinds of relations sensible in general even when rounding.

Consider a 100 GiB drive with a 44.2 GiB partition and 55.8 GiB of free space. If someone wanted to get a general feel for the situation, for example to display in a UI, here's what we'd see:

With ROUND_UP:

  Disk Size: 100 GiB
  Used: 45 GiB
  Free: 56 GiB

With ROUND_HALF_UP:

  Disk Size: 100 GiB
  Used: 44 GiB
  Free: 56 GiB

Which one do you think does a better job of representing the truth?

[vpodzime]

Originally I didn't want to set any default because with storage, it's always a bug to rely on some default rounding. That's also why we always specify the rounding mode in Blivet.

As for the ROUND_HALF_UP I wasn't able to come up with an example of where it could be useful as using such value would always lead to issues. Even when showing to the user in the example above -- the user could tell themselves "Okay, I have 56 GiB of free space so I'm gonna create a 56GiB partition." which would fail. Thus the sizes should have been rounded up (used) and down (free) to prevent such issues.

So what I'm gonna do is:

• add ROUND_HALF_UP to rounding modes in libbytesize
• remove the default rounding mode in round_to_nearest() in libbytesize
• set the default rounding mode to ROUND_HALF_UP in round_to_nearest() in Blivet and add a warning to the docstring that relying on the default rounding mode is always a bug

[vpodzime]

The tests are gonna fail here because they require the new libbytesize. I'll build and install it on our jenkins slaves once storaged-project/libbytesize#15 gets merged.

[dwlehman]

I would be okay with making rounding mode a positional/mandatory argument for blivet-3.0.

[vpodzime]

I would be okay with making rounding mode a positional/mandatory argument for blivet-3.0.

I had the very same idea.

[vpodzime]

Jenkins, please test this.

[vpodzime]

Jenkins, please test this again.

[vpodzime]

Ran the check manually on the Jenkins slave and all tests passed.