Null characters in branch name?

[Matt Mackall]

On Thu, 2014-08-21 at 12:23 -0700, Rick Rankin wrote:
> Interesting. That hadn't occurred to me. I did have at least one disk
> failure in the laptop I was using that the time, though I don't have a
> record of when the failure(s) occurred, so I can't correlate. It's
> certainly a possibility, though.

If you're using a laptop, you're using DRAM with typical
undetected/uncorrected memory failure rates of between once a week and
once a day per GB. Mostly these are invisible, but sometimes they result
in software crashes, and sometimes in data corruption like this.

As it's very hard to have this sort of thing happen in a Python program
(there are no pointers to corrupt and no easy ways to get NUL bytes into
strings by accident like in C), and we empirically have very little
evidence of it happening in any sort of frequent or deterministic way,
it smells a lot like hardware failure.

-- 
Mathematics is the supreme nostalgia of our time.