Re: [PATCH v2 06/15] rev-list: make --count work with --objects

From: Jeff King <peff@peff.net>
Date: Feb 18, 2020, 19:59 UTC
Message-ID: <20200218195901.GB21774@coredump.intra.peff.net>
In-Reply-To: <3230c8c9-5208-3161-324e-c0ce948ee5b3@gmail.com>

On Tue, Feb 18, 2020 at 01:05:35PM -0500, Derrick Stolee wrote:

Show 20 quoted lines

> On 2/14/2020 1:22 PM, Jeff King wrote:
> > +test_expect_success 'rev-list --count' '
> > +	count=$(git rev-list --count HEAD) &&
> > +	git rev-list HEAD >actual &&
> > +	test_line_count = $count actual
> > +'
> > +
> > +test_expect_success 'rev-list --count --objects' '
> > +	count=$(git rev-list --count --objects HEAD) &&
> > +	git rev-list --objects HEAD >actual &&
> > +	test_line_count = $count actual
> > +'
> 
> I suppose these tests work, although I would probably
> prefer precomputed explicit expected counts instead
> of asking rev-list for the correct answer to a
> rev-list command. This is still fine, because we test
> that the non-count versions return the correct results,
> but I would hate for a bug to affect both modes equally
> and cause these tests to pass.

I was hoping it would make the tests a bit less brittle. And as you note, we should be checking the enumeration results themselves more carefully in other tests, so I find it fairly unlikely to see a bug that doesn't trigger _any_ test failure.

-Peff