[IPython-dev] Tag and branch 0.10 soon?

[Fernando Perez]

Hi all,

On Thu, Apr 23, 2009 at 11:59 AM, Fernando Perez <fperez.net at gmail.com> wrote:

> Well, as I said, there's nothing to be ashamed of on your part, you
> did what you had to do to make it work.  The 'right' way to do this
> will come when in the refactoring, we isolate ipython instances
> correctly so they don't have global side effects all over the place
> like they do now.  At that point the hacks you had to resort to, and
> which cause massive problems in the test suite, will just become
> unnecessary.

OK, here's a first stab at this:

https://code.launchpad.net/~fdo.perez/ipython/brian-bugfixes

Brian has already merged a bunch of stuff, and I'm building on top of
that.  Now there's a new test runner (ugly, I know, but a clean
solution will be for later...) called 'iptestall', that for now I'm
not even putting a public script for.  But if you cd to the testing/
subdir, you can run:

uqbar[testing]> ./iptestall.py
*****************************************************************************
IPython test set: kernel
...
----------------------------------------------------------------------
Ran 3 tests in 0.014s

OK

[...]

*****************************************************************************
Ran 10 test sets in 37.278s

OK

###

I'm not parsing the output of each subprocess so we lose the total
test count, but that can be fixed later.

At least this has the benefit of being much more robust.  For now I
added the modules and packages to be scanned by hand, it could be
automated later though I'm not sure it's worth automating (more
brittleness).  I tried to do something very simple but that would
*just work*.

If a test fail, you get this kind of message (I broke one deliberately):

*****************************************************************************
Ran 10 test sets in 36.350s

ERROR - 1 out of 10 test sets failed.
----------------------------------------
Runner failed: tests
You may wish to rerun this one individually, with:
iptest -v IPython.tests


And you can now run that one by hand with the old iptest that gives
you access to --pdb and all the nose goodies.  If the trial one is the
one with an error, the printed message will show that too.

Cheers,

f