Re: [PATCH] Reduce performance penalty for turned off traces

From: Jeff King <peff@peff.net>
Date: Nov 17, 2017, 22:16 UTC
Message-ID: <20171117221631.x2xqeeqmf5xob3yv@sigill.intra.peff.net>
In-Reply-To: <CAGZ79kb4Eo0AC4Z42_ks5_+Pcy4DrDCufNb=5J=g8L6pMnHd7A@mail.gmail.com>

On Wed, Nov 15, 2017 at 11:14:20AM -0800, Stefan Beller wrote:

Show 9 quoted lines

> > I did manually disable HAVE_VARIADIC_MACROS and confirmed that the
> > result builds and passes the test suite (though I suspect that GIT_TRACE
> > is not well exercised by the suite).
> 
> GIT_TRACE is exercised in the test suite (though I am not sure if it counts
> as well-exercised) in t7406-submodule-update.sh for example, which uses
> GIT_TRACE to obtain information about thread parallelism used by Git, as
> that is not observable otherwise, if we assume that performance tests in the
> standard test suite are not feasible.

Hmm, yeah, that might cover it. What I'm worried about is that we missed some case where NULL needed to become &trace_default_key. But I did look for that in my review of the patch and didn't notice any spots. And the coverage in t7406 should help.

> > After your patch, the GIT_TRACE=1 time remains the same but GIT_TRACE=0
> > drops to 1ms.
> 
> So does that mean we can use a lot more tracing now?

Yep, that's the intent.

-Peff