Improving the regression testing infrastructure

[Eric Pouech]

Geoffrey Hausheer a écrit :
> 
> > >    True. So what we need is to add a function in the
> > testing framework
> > > so that tests can get argc/argv/envp. Seems like a
> > reasonable approach.
> > I was more thinking adding some wt_argc and wt_argv, with init in main
> > so that we don't have to deal with the portability issues of
> > the __argc,
> > __ ARGC...
> 
> Is there a reason we can't just use environment variables
> rather than argc/argv?  I would think that would be less
> of a hassle than using argc/argv.

on the same address space across several threads, no need for
env variables, all the threads share the same address space

when between two processes, then the argc/argv is passed anyway
(arguments
to main)
the only place where you could need then is for testing in child
that you get the correct strings (see exchanges with François around
this)