From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <vdavydov.dev@gmail.com>
Date: Mon, 13 Aug 2018 11:34:30 +0300
From: Vladimir Davydov <vdavydov.dev@gmail.com>
Subject: Re: [PATCH 4/8] box: remove orphan check from box_is_ro()
Message-ID: <20180813083430.ofl75nzppuixwwpa@esperanza>
References: <cover.1533679264.git.v.shpilevoy@tarantool.org>
 <07b102e6b0abd42b550e718fd366927144b0c527.1533679264.git.v.shpilevoy@tarantool.org>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <07b102e6b0abd42b550e718fd366927144b0c527.1533679264.git.v.shpilevoy@tarantool.org>
To: Vladislav Shpilevoy <v.shpilevoy@tarantool.org>
Cc: tarantool-patches@freelists.org
List-ID: <tarantool-patches.dev.tarantool.org>

On Wed, Aug 08, 2018 at 01:03:47AM +0300, Vladislav Shpilevoy wrote:
> Box_is_ro now checks both for 'read_only' and 'orphan' modes, but
> in promotion only 'read_only' is needed. And now there is no a
> method to get the current 'read_only' value. After replacing
> box_is_ro with box_is_writable it is possible to reimplement
> box_is_ro as a getter for 'read_only' option.
> ---
>  src/box/box.cc     | 10 ++++++++--
>  src/box/box.h      |  3 +++
>  src/box/lua/info.c |  2 +-
>  3 files changed, 12 insertions(+), 3 deletions(-)
> 
> diff --git a/src/box/box.cc b/src/box/box.cc
> index 6eb358442..d8fbc6252 100644
> --- a/src/box/box.cc
> +++ b/src/box/box.cc
> @@ -216,17 +216,23 @@ box_set_ro(bool ro)
>  	fiber_cond_broadcast(&ro_cond);
>  }
>  
> +bool
> +box_is_writable(void)
> +{
> +	return !is_ro && !is_orphan;
> +}
> +
>  bool
>  box_is_ro(void)
>  {
> -	return is_ro || is_orphan;
> +	return is_ro;
>  }
>  
>  int
>  box_wait_ro(bool ro, double timeout)
>  {
>  	double deadline = ev_monotonic_now(loop()) + timeout;
> -	while (box_is_ro() != ro) {
> +	while (!box_is_writable() != ro) {
>  		if (fiber_cond_wait_deadline(&ro_cond, deadline) != 0)
>  			return -1;
>  		if (fiber_is_cancelled()) {

So now we have box_wait_ro() that checks box_is_writable() !=
box.cfg.read_only and we have box_is_ro() that returns the value
of box.cfg.read_only. Looks ugly.

I think that the promotion algorithm shouldn't flip box.cfg.read_only.
Instead it should use its own flag and box_is_ro() should be defined as

  bool box_is_ro() { return is_ro || is_oprhan || is_slave; }

That would be consistent with the orphan state.