Re: [RFC] On watchman support

From: Junio C Hamano <gitster@pobox.com>
Date: Nov 18, 2014, 21:26 UTC
Message-ID: <xmqqwq6std27.fsf@gitster.dls.corp.google.com>
In-Reply-To: <1416345123.27401.11.camel@leckie>

David Turner <dturner@twopensource.com> writes:

Show 8 quoted lines

> On Tue, 2014-11-18 at 12:55 -0800, Junio C Hamano wrote:
>> I vaguely recall that the reason why we dropped it was because it
>> was too much code churn in an area that was being worked on in
>> parallel, but you may need to go back to the list archive for
>> details.
>
> OK, in that case I'll try to remember to reroll it once the rest of the
> refs stuff lands.

Sure. But I would much prefer to see us explore an arch independent optimisation of the caller before starting to micro-optimize a leaf function.

It is not check_refname_format() that is the real problem. It's the fact that we do O(# of refs) work whenever we have to access the packed-refs file. check_refname_format() is part of that, surely, but so is reading the file, creating all of the refname structs in memory, etc. (credit to peff@).