Comment 11 for bug 2035061

Time calculations can be so tricky!

The patch is also changing the non-human-readable case (without -p), if I understood it correctly. I think this also needs to be tested.

I think it would be beneficial to include the upcoming fix for the other corner case you found, when uptime is exactly 60s, as that also affects the later versions. That's one less SRU for the future.

Finally, since you have such a nice case, do you think you could add that as a DEP8 test? The bind mount works in a LXD container. The upstream testsuite/uptime.test is too basic to be changed for this case, it's basically just a smoke test.