Re: [PATCH 3/3] test-lib: '--run' to run only specific tests

From: Junio C Hamano <gitster@pobox.com>
To: Ilya Bobyr <ilya.bobyr@gmail.com>
Cc: git@vger.kernel.org, Thomas Rast <trast@inf.ethz.ch>,
	Eric Sunshine <sunshine@sunshineco.com>,
	Ramsay Jones <ramsay@ramsay1.demon.co.uk>
Subject: Re: [PATCH 3/3] test-lib: '--run' to run only specific tests
Date: Tue, 06 May 2014 14:02:31 -0700	[thread overview]
Message-ID: <xmqqfvkmei3c.fsf@gitster.dls.corp.google.com> (raw)
In-Reply-To: <xmqqk39yeiic.fsf@gitster.dls.corp.google.com> (Junio C. Hamano's message of "Tue, 06 May 2014 13:53:31 -0700")

Junio C Hamano <gitster@pobox.com> writes:

> The need to explain better with longer description will reduce the
> likelyhood that the feature is understood and correctly used.  When
> you can write "1-2,4-", why accept "1-4 !3" and force yourself to
> explain to people why that is different from "!3 1-4"?

By the way, having said all that, I would understand if the result
is made not depend on the order of selectors given to the option.

That is:

 - Find all the positive ones and form the "positive" set; if there is
   no positive ones, then make the "positive" set include everything.

 - Find all the negative ones and form a "negative" set.  The
   "negative" set can be an empty set if there is no negative
   selector.

 - Subtract the "negative" set from the "positive" set, and run only
   the tests in the resulting set.

Then "1-4 !3" and "!3 1-4" would mean the same thing.  Explanation
of the semantics would be far simpler than "we do it from left to
right".

One reason why the orders shouldn't matter is, unlike the "print
dialog" case, we won't run tests out of order when given "1 4 3 2".
We will still run "1-4".  So it would be natural for the readers to
expect that the orders they give selectors would not matter.