Shipping your own `run-tests.py` is a bad practice as the custom version will
slowly diverge from important fixes and upgrade from the upstream one. Using the
core one is simple and we should use it.
I'm not convinced by the arguments used in the re-installation (e221a52),
- It is easy to run the test against any version of Mercurial including the
installed one using `--with-hg=`,
- If you have people contributing to this level of extensions, having them
clone Mercurial is fairly cheap. If this is really seen a blocker, completing
the Mercurial's upstream plan to have run-tests.py installable as a developer
tools seems a cheap way to move forward instead,
- If you own a complex installation with multiple Mercurial related dependencies,
you already have a way to handle version dependency in this context and your
contributor can use this very same infrastructure to make sure they run at the
right version. And they probably should use it already anyway to make sure
they are testing the stack properly.
This changesets also install a README explaining how to obtains the test runner
and runs it against various Mercurial version.