svn commit: r187332 - head/tools/regression/usr.bin/jot

[Robert Watson]

On Sat, 17 Jan 2009, Stanislav Sedov wrote:

>>> I understand, this is a bikesched, but I really don't see a reason. You 
>>> can't build FreeBSD on windows anyway.
>>
>> Many of us would *very* much like to be able to cross-build FreeBSD from 
>> both Windows and Mac OS X, which would be highly desirable for embedded 
>> systems and appliance shops.  The first obstacle to making that work is 
>> that you can't even check out our source code correctly on those platforms, 
>> so fixing that is an important priority so that the remainder can be worked 
>> on.
>
> I think this should be rather worked out in these platforms themselves (e.g. 
> by renaming files in the checkout files, and so on) withouth trashing our 
> tree. While some changes could be understandable, the one I've replied 
> completely broked heirarchy of jot regressions tests: before the names of 
> files were coreesponding to the options used in the test, now you can't say 
> exactly what options used by looking at the filename. This looks like a 
> gratitous change and regression to me.

I agree that breaking regression tests is bad.  However, I think that 
requiring files to be renamed on checkout, and then presumably modifying the 
build infrastructure to use the modified names, etc, is much worse than 
changing the regression test not to rely on case-sensitive file systems.  I 
think a reasonable goal should include allowing our source code to be 
distributed to and (in the long term) built on both Windows (with cygwin) and 
Mac OS X in an unmodified form.  There are lots of obstacles between us and 
having that happen, but few of those will be as easy to address as tweaking 
filenames and regresion tests.

Robert N M Watson
Computer Laboratory
University of Cambridge