Re: [PATCHv2 1/2] git p4: Fixing script editor checks

From: Junio C Hamano <gitster@pobox.com>
Date: Apr 11, 2012, 19:06 UTC
Message-ID: <7vpqbejdf4.fsf@alter.siamese.dyndns.org>
In-Reply-To: <CAE5ih7_X=4gewga8fMzaEvowsJsA1Ta9PQ2bGixx5eVvykputA@mail.gmail.com>

Luke Diamand <luke@diamand.org> writes:

Show 13 quoted lines

> On Wed, Apr 11, 2012 at 7:14 PM, Junio C Hamano <gitster@pobox.com> wrote:
>> Luke Diamand <luke@diamand.org> writes:
>>
>>> If P4EDITOR is defined, the tests will fail when "git p4" starts an
>>> editor.
>>
>> Is that a problem specific to tests, or should "git p4" itself unset that
>> environment? If it is a problem specific to tests, would it be a better
>> fix to add "P4EDITOR=:" like we do for EDITOR in t/test-lib.sh?
>
> Yes and no - git-p4.py will run $P4EDITOR if it is set, even if it's
> just empty. So it would need a small fix to check for an empty string.
> I can submit a suitable patch.

How does real "p4" run $P4EDITOR?  For example, if you have:

	P4EDITOR='vi -e'

does it start "vi" in its "ex" mode, implying that it behaves as if the invocation were like this in a hypothetical implementation of "p4" as a shell script:

	#!/bin/sh
        ... do some stuff p4 does ...
	$P4EDITOR file-to-be-edited

Whatever it does needs to be emulated by the code in git-p4.py that runs it, as that is the way the end users expect, and if it turns out to be like the above, setting P4EDITOR to ':' just like we do to EDITOR in t/test-lib.sh should be the right thing to do. The always-true command colon will leave the file-to-be-edited unmodified and report success.

Similarly, what happens with "p4" when $P4EDITOR is set to an empty string? If it ignores and falls back to the other usual way to find your preferred editor, you should emulate that as well in git-p4.py.

Thanks.