[PATCH] tests: use Git's test framework

[Jason A. Donenfeld]

On Mon, Apr 1, 2013 at 4:09 PM, John Keeping <john at keeping.me.uk> wrote:
> Obviously there are a number of further changes that can be done on top
> of this to avoid the now-unnecessary wrapper functions and use a test
> prerequisite for HTML tidy but I don't want to work on those unless
> there is agreement that this is a sensible direction.

If moving forward with this would indeed result in reducing the
complexity of our own test harness, I'm okay with that. Removing our
wrapper would be quite nice. AFAIK, git's testing infrastructure is
fairly stable, so I don't anticipate using it will introduce any
unforeseen hiccups down the road.

Maybe, though, we ought to be having a larger discussion about this.
We're in the process of bringing cgit up to speed with git head. We're
now relying on git's Makefile. And now we're discussing using git's
test infrastructure as well. We're getting cozier with git.

Is this something we want to do? Or should we say, "let's just be
friends", before a child is on the way, and then a fight followed by a
nasty divorce?

My feeling is that reusing as much of git's existing infrastructure is
ideal and should be encouraged. One of the strengths of cgit has
always been that we use the actual meat of git and are tied to it in
an intimate way. In fact, down the road, if cgit is in a ripe position
for merging in some respect with git-itself, I think this is also
something I'd be in favor of. (But hey, it's only the third date;
let's not get too excited about such futures ;-).

So from this perspective, I'd be in support of relying on git's test harness.

But there might be other considerations I've overlooked; my ears are
open to dissenting opinions.