Re: not ok 47 - log.decorate configuration

From: Jeff King <peff@peff.net>
Date: Mar 26, 2021, 04:19 UTC
Message-ID: <YF1g5jrCtCSmfwH/@coredump.intra.peff.net>
In-Reply-To: <xmqqy2easpxb.fsf@gitster.g>

On Thu, Mar 25, 2021 at 05:45:52PM -0700, Junio C Hamano wrote:

Show 11 quoted lines

> Jeffrey Walton <noloader@gmail.com> writes:
> 
> > On Thu, Mar 25, 2021 at 3:57 AM Jeffrey Walton <noloader@gmail.com> wrote:
> >> ...
> >
> > Cancel this. It just tested OK a second time around.
> 
> Which may indicate that the test is flakey (e.g. timing dependent,
> only reproducing when two commands in the test sequence are executed
> across a second boundary, or something like that), which may be
> worth fixing.

I can't get it to fail with "./t4202-log.sh --stress", which usually turns up races or other flakes. But perhaps there is something unique about your (Jeff) system that triggers it.

It's hard to tell more without any verbose output from the failing test. In general, failure reports accompanied by the output (or at least the end of the output) of "./t4202-log.sh -v -i -x" will be a lot more useful.

If you're seeing intermittent failures from a larger "make test", consider putting "--verbose-log -x" into GIT_TEST_OPTS (either on the make command-line, or in config.mak). The logs will be in t/test-results/t4202-log.out, etc.

-Peff