[PATCH V4] test-walk: enable for Windows

[Matt Mackall]

On Thu, 2012-06-14 at 10:05 +0200, Adrian Buehlmann wrote:
> On 2012-06-12 12:19, Mads Kiilerich wrote:
> > Moving 'glob:glob' might however lead to reductio ad absurdum. It is 
> > possible that 'glob:glob' doesn't test anything relevant that isn't 
> > tested elsewhere. It just don't think the case should be removed 
> > silently without reasonable arguments.
> 
> The absurdity is indeed not insignifcant. At least it appears so on my end.
> 
> The net result is that test-walk.t is still not tested on Windows. We
> can aim for perfection for quite some more time if we want. But if the
> cost is not having this test enabled on Windows, then I think it's not
> worth that cost.
> 
> FWIW, I tried to remove the glob:glob case in one of my early tries of
> this patch because it is simply an illegal filename on Windows (colon is
> a reserved character).
> 
> This part of the (continued) discussions appears a bit odd to me, given
> how much dedication people are obviously investing into having warnings
> about non-portable filenames elsewhere.
> 
> But obviously, your and my mileage seem to vary.
> 
> And yes, off the top of my head I can't think of a reason why a
> glob:glob file by itself should be a relevant testcase. But I indeed
> haven't digged deep enough into the functions that are tested here.

I think we need to test -somewhere- that we can pass an explicit
filename of "glob:glob" through the command-line parsing and file
matching framework. In other words, test that we can indeed name files
that look like patterns. As far as I can tell, that's not even happening
in this test though.

I'm not terribly concerned with the other side of things, as the magic
we do with filenames on the walking side shouldn't care about this sort
of thing.

-- 
Mathematics is the supreme nostalgia of our time.