Jan Nikl

Thanks @dylan-copeland , actually you may see the CI in work here, including the test with single precision (the last one) :wink: . All of the tests have passed 👍...

I do not know, if `MFEM_PRECISION` is `SINGLE` or `DOUBLE`, I would stick to that...

I mean there is not a single way how to abbreviate "single" 😄 , it should be "sngl", but in three letters that "sgl" is the most suggestive, I think...

Well, that is another option 😄 . It does not need to be 3 letters, but something short and of the same length ideally as it appears in the test...

Ok, here we go. Just @v-dobrev , based on the discussion on Slack, the preference was to use tags instead of branches for github-actions to prevent the situation when `master`...

Done, ready for merging :wink:

I think the problem might be that `DistanceTo` sums totally different numbers, it would be better to sum the relative errors, i.e., instead of `(x_i-y_i)^2` sum `(x_i-y_i)^2/y_i^2` :wink:

Yes, that would explain why it happens only in this context. I presumed these basic methods are safe, but MFEM can always surprise 😄 . It goes quite far beyond...

That is ok :wink:, so I did that in a17333cb1918d5550b01b1fdb706c69e450e5436 . I also merged in `master` due to the recent changes of the tests in #4262 . This should safe...

@tzanio , would you merge it in `next` again to see if we are fine or not?