test-ui-color.py failure in Emacs shell-mode

[Matt Mackall]

On Wed, 2012-05-02 at 21:31 -0400, Jesse Glick wrote:
> On 05/02/2012 08:38 PM, Matt Mackall wrote:
> > Was there a point where this worked?
> 
> The test in its first revision worked.
> 
> > Can you bisect it?
> 
> Not easily - intermediate revs throw an error:
> 
>    ...
>    File ".../mercurial/httpclient/socketutil.py", line 81, in <module>
>      PROTOCOL_SSLv2 = ssl.PROTOCOL_SSLv2
> AttributeError: 'module' object has no attribute 'PROTOCOL_SSLv2'
> 
> which leads to the presumably unrelated

For future reference, this is what 'hg bisect --skip' is for.

> But by temporarily applying the patch of 494b26ad8736 where that error occurs, testing, and reverting, I can bisect down to:
> 
> changeset:   14516:842a9179132c
> user:        Idan Kamara <idankk86 at gmail.com>
> date:        Thu Jun 02 00:43:34 2011 +0300
> summary:     color: check if ui is already a subclass of colorui before wrapping it

> which is simply the changeset that adds the parts to the test that fail. So this is a bit anticlimactic.

So.. what's the real issue here? That color doesn't work in Emacs? Or
that the test is failing?

On my end, when running GNU Emacs 23 M-x term in a Gnome terminal, color
works AND the test passes.

$ env | grep TERM=                                           
TERM=eterm-color                                                                
COLORTERM=gnome-terminal     

I never use the GUIfied Emacs and don't have it installed.

-- 
Mathematics is the supreme nostalgia of our time.