Re: Git bisect run should check for the existence of the script

From: Junio C Hamano <gitster@pobox.com>
Date: Jul 17, 2020, 19:30 UTC
Message-ID: <xmqqa6zydjm4.fsf@gitster.c.googlers.com>
In-Reply-To: <xmqqeepadk8q.fsf@gitster.c.googlers.com>

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

Show 5 quoted lines

> Do you mean a new option must be passed if the end-user expects the
> script to always exist across revisions, or the script is not
> tracked to begin with?  It feels somewhat backwards and the effort
> by the end-user to always type the option is better spent to make
> sure there is no typo on the command line.

I forgot to follow up on this part.

It does not change the conclusion that it needs to be done carefully if we chose to retroactivelyreserve return code 127 for our own use, but such a backward incompatible change can easily be worked around if users relied on the current behaviour that a missing (tracked) script would mark the revision "bad" without being a fatal error in "git bisect run". Instead of

    git bisect run "./tracked-script"

they can just do

    git bisect run "test -f ./tracked-script && ./tracked-script"

and the problem is solved ;-)