It could also probably be improved further, we've been detecting python
_after_ looking for python-based tools rst2*. Or we can keep unrelated
checks, I don't think it hurts.

The fundamental problem still exists, in that we have no way to
tickle the threads stuck in accept(2), so for now this remains
a debugging feature.

If this (=c00080.vtc) explodes on some platform, start looking
at the Waiter_Destroy() implementation.

a29fca70 requires subprocess.check_output
which was added with Python 2.7

gcov numbers.

Say so if feature +short_panic supresses output.

free slots, one for the data we want to write and one for the
end of chunk bits.

Fixes: #2207

inside struct ws.

clear that you are not supposed to use them as pointers.

Spotted by coverity.

While here rearrange variables to be more consistent.