tests: make a stab at approximating wall-clock times
Vaguely empirical observations:
* ".py" tests are about an order of magnitude faster than ".t" tests
* dividing size by 1000 gives an approximation to wall-clock
run time (in seconds) that is not completely ridiculous.
Test update logic when there are renames
Update with local changes across a file rename
$ hg init
$ echo a > a
$ hg add a
$ hg ci -m a
$ hg mv a b
$ hg ci -m rename
$ echo b > b
$ hg ci -m change
$ hg up -q 0
$ echo c > a
$ hg up
merging a and b to b
warning: conflicts while merging b! (edit, then use 'hg resolve --mark')
0 files updated, 0 files merged, 0 files removed, 1 files unresolved
use 'hg resolve' to retry unresolved file merges
[1]