Re: colorful tests

[Benoit SIGOURE]

On Nov 4, 2007, at 7:17 PM, Ralf Wildenhues wrote:

> Hello Benoit,
>
> * Benoit SIGOURE wrote on Sun, Nov 04, 2007 at 06:50:08PM CET:
> > On Nov 4, 2007, at 6:41 PM, Ralf Wildenhues wrote:
> > > I took the various bits of code from the parallel tests check.mk  
> > > from
> > > Akim and tried to mangle that suitable for inclusion in Automake.
> > > This is what I ended up with.  (More topics to follow.)
> > > What do you think about it?
> >
> > It looks great.
>
> Thanks for looking at it.  Another question that came to me:
>
> While `parallel-tests' (the next Automake option to be implemented)
> is certainly needed (as the chance is large that the new code will
> break at least some existing test suites), I'm not sure color-tests
> is even needed.  We could just make color the default for everyone
> (given capable terminal of course).  But maybe developers with exotic
> terminal color settings will be upset.
>
> Of course the step of making color-tests the default could also be
> done some time later.  (Maybe including also a no-color-tests for
> die-hards.)
>
> Further, user's should have a say in this, too.  Do you think it's
> acceptable to tell them  "Use `TERM=dumb make check' if you don't
> want colorized output, or pipe output through `cat'"?  Or should
> we have (gasp!) yet another configure switch, --disable-color-tests?

I think we should leave it as-is for some time.  Since the option is  
used in Automake itself, it give a chance to all the people  
developing/testing/using bleeding edge Automake to test this  
feature.  If everyone's happy with it in, say, a couple of months, we  
can safely make it a default.
Now I think that the no-color-tests could be useful for those out  
there that are allergic to fancy colors and stuff, so I think it'll  
be worth implementing it (which is easy).
I don't like the idea of yet another configure switch...  configure  
scripts are already large enough and since this option would only  
change a small aesthetic aspect of the package, it's not worth the  
price.  OTOH, we can indeed document the TERM=dumb trick or maybe add  
another variable, but that's probably not the cleverest thing to do.   
Indeed, what if the testsuite needs a working terminal for some  
reason?  We could use another variable, say, NOCOLOR, and people  
would do `make check NOCOLOR=whatever'.  We simply need to throw a  
test x$(NOCOLOR) = x in the `&&' pipeline that decides whether colors  
must be used or not.

--
Benoit Sigoure aka Tsuna
EPITA Research and Development Laboratory

PGP.sig
Description: This is a digitally signed message part