Date: Fri, 12 Jul 2024 14:18:37 -0500
From: Nathan Bossart <nathandbossart@gmail.com>
To: Sami Imseih <samimseih@gmail.com>
Cc: Bertrand Drouvot <bertranddrouvot.pg@gmail.com>,
	Tom Lane <tgl@sss.pgh.pa.us>, Robert Haas <robertmhaas@gmail.com>,
	"pgsql-hackers@postgresql.org" <pgsql-hackers@postgresql.org>
Subject: Re: Restart pg_usleep when interrupted
Message-ID: <ZpGBjZZtbDGUtvTV@nathan>
References: <AC328D21-21D6-4FFC-8F6B-BFE9AD940524@gmail.com>
 <ZoPC5IeP4k7sZpki@ip-10-97-1-34.eu-west-3.compute.internal>
 <373D199B-2312-431D-8018-F2808913F979@gmail.com>
 <Zo0UdeE3i9d0Wt5E@ip-10-97-1-34.eu-west-3.compute.internal>
 <3712E898-03DA-4E03-AF51-D3E24DE68216@gmail.com>
 <C18017A1-EDFD-4F2F-BCA1-0572D4CCC92B@gmail.com>
 <Zo_7blmO166ZDGbW@nathan>
 <2F781744-69B8-4BC2-A6EF-3998DCCDB634@gmail.com>
 <ZpA4_Pw14FCP6025@nathan>
 <01A15AEA-C35C-41DF-8E81-3B5A0B523939@gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <01A15AEA-C35C-41DF-8E81-3B5A0B523939@gmail.com>
Archived-At: <https://www.postgresql.org/message-id/ZpGBjZZtbDGUtvTV%40nathan>
Precedence: bulk

On Fri, Jul 12, 2024 at 12:14:56PM -0500, Sami Imseih wrote:
> 1/ TimestampDifference has a dependency on gettimeofday, 
> while my proposal utilizes clock_gettime. There are old discussions
> that did not reach a conclusion comparing both mechanisms. 
> My main conclusion from these hacker discussions [1], [2] and other 
> online discussions on the topic is clock_gettime should replace
> getimeofday when possible. Precision is the main reason.
> 
> 2/ It no longer uses the remain time. I think the remain time
> is still required here. I did a unrealistic stress test which shows 
> the original proposal can handle frequent interruptions much better.

My comment was mostly about coding style and not about gettimeofday()
versus clock_gettime().  What does your testing show when you don't have
the extra check, i.e.,

	struct timespec delay;
	struct timespec remain;

	delay.tv_sec = microsec / 1000000L;
	delay.tv_nsec = (microsec % 1000000L) * 1000;

	while (nanosleep(&delay, &remain) == -1 && errno == EINTR)
		delay = remain;

-- 
nathan