Hi,

At the moment ATF provides a binding for C/C++ that simplifies the  
creation of test programs that follow the ATF interface (command line  
and output format).  This "binding" also takes care of many run-time  
stuff, such as the handling of signals, the creation of temporary  
directories, etc.

However, this is currently all written in C++.  There are multiple  
macros that hide almost all details of this fact -- you use the  
macros instead of dealing with classes, exceptions, etc. --, but I  
think that there can be some resistance in using it to write tests  
for features that are plain C.  (I'm specially afraid of this because  
of the sudden interest on pcc, which does not support C++.)

What would you think of a plain C binding so that test programs do  
not need to rely at all on C++?  This could restrict all C++ code to  
the ATF tools and not propagate to tests written by other people  
(e.g. in our use case, tests that live in the NetBSD source tree).   
Sure, implementing this will take some time (/me hates C for not  
having anything to deal with data collections nor strings), but it  
seems a good idea to me at the moment, specially for the long run.

Any comments?

Thanks,

-- 
Julio M. Merino Vidal <jmmv84@gmail.com>