Jeff King <peff@xxxxxxxx> writes:
> On Fri, Apr 06, 2018 at 09:31:22PM +0200, Johannes Schindelin wrote:
>
>> In 0b294c0abf0 (make deleting a missing ref more quiet, 2008-07-08), we
>> added a test to verify that deleting an already-deleted ref does not
>> show an error.
>
> Amazing that it took this long to come up.
> ...
>> This patch chooses instead to look for the prefix "error:" at the
>> beginning of the line, so that there can be no ambiguity that any catch
>> was indeed a message generated by Git's `error_builtin()` function.
>
> Yep, this seems obviously correct.
Hits in
$ git grep 'grep ["'\'']*error' t
shows that many checks for errors that are not anchored, but I do
not think any of them are looking for the string in a pathname other
than this instance, so it would be OK.
Thanks, both. Will apply.
