From: Junio C Hamano <gitster@pobox.com>
Date: Fri, 03 Jan 2014 18:00:39 GMT
Subject: Re: [PATCH v2 3/4] Speed up is_git_command() by checking early for internal commands
Message-ID: <xmqqob3tlzco.fsf@gitster.dls.corp.google.com>
In-Reply-To: <20140103154426.GA23534@sigill.intra.peff.net>

Jeff King <peff@peff.net> writes:

> On Thu, Jan 02, 2014 at 11:41:05AM -0800, Junio C Hamano wrote:
>
>>  - builtin/merge.c is the same, but it is conceptually even worse.
>>    It has the end-user supplied string and wants to see if it is a
>>    valid strategy.  If the user wants to use a custom strategy, a
>>    single stat() to make sure if it exists should suffice, and the
>>    error codepath should load the command list to present the names
>>    of available ones in the error message.
>
> Is it a single stat()? I think we would need to check each element of
> $PATH. 

Yeah, load_command_list() iterates over the members of env_path.

> Though in practice, the exec-dir would be the first thing we
> check, and where we would find the majority of hits. So it would still
> be a win, as we would avoid touching anything but the exec-dir in the
> common case.

Exactly.