I've looked into the bug and it seems you are right. However, the comment of Craig Small worries me a little bit:
Show quoted text> Thats a big call to say it won't have a problem on any system. The problem
> is lots of systems do dumb things to timers. So to change something like
> that it will need to be tested a lot first.
>
> Things that suspend come to mind and systems that use stolen time are
> another.
> Ill try to find where btime was used and why. there may be a good reason
> for it.
Given the system specs, the calculation is correct. That btime is not local to
the VM, but to the underlying host should not be fixed in the downstream
software (my opinion).
For this conclusion I thought of a scenario where two different virtual server
software packages exist, one using btime as time when the (underlying) host started,
the other using uptime. Here I cannot decide which implementation is the correct one.
Additionally, until I cannot foresee the effect of this change, I'll keep this bug on hold.
One important factor will be the decision of the procps team.
In any case, many thanks for submitting this bug.
Joachim
Am Sa 05. Nov 2016, 20:45:00, jidanni@jidanni.org schrieb:
Show quoted text