[PATCH V4] run-tests: use hghave.py

[Mads Kiilerich]

Adrian Buehlmann wrote, On 06/19/2012 01:45 PM:
> # HG changeset patch
> # User Adrian Buehlmann <adrian at cadifra.com>
> # Date 1340041060 -7200
> # Node ID 60e6c72cc6df5e3c1e8abd88c6949a6382e31411
> # Parent  79902f7e27df0eeae96d4d47d5c8a12fae2a0bb7
> run-tests: use hghave.py
>
> Eliminates the start of a subprocess for #if requirements checking.
>
> The has_xxx functions in hghave.py now get the path where they should do their
> check in.
...
> diff --git a/tests/run-tests.py b/tests/run-tests.py
> --- a/tests/run-tests.py
> +++ b/tests/run-tests.py
> @@ -54,6 +54,7 @@
>   import time
>   import re
>   import threading
> +import hghave

This will introduce an import dependency from run-tests to the mercurial 
that we are testing. I'm waiting for Matt to decide on that.

Avoiding code duplication is nice, but tight coupling is bad. If we 
really wanted run-tests to be 'a mercurial extension' then it would be 
completely different. It might however be worth it.

It is also possible that we should duplicate some code and remove the 
mercurial imports from hghave.py before we start importing hghave.

> @@ -1306,6 +1301,11 @@
>   
>       COVERAGE_FILE = os.path.join(TESTDIR, ".coverage")
>   
> +    # insert root in sys.path, so mercurial can be imported in hghave
> +    rootdir = os.path.realpath(os.path.join(TESTDIR, ".."))
> +    if rootdir not in sys.path:
> +        sys.path.insert(1, rootdir)
> +

In the case where we are testing without -l (and thus will create a temp 
hg install) and haven't run 'make local' then this will point at a 
mercurial without any compiled binary extensions. That might not be what 
we want and wasn't what we did before.

/Mads