Re: daemontools tai64n is unbuffered, s6-tai64n is fully buffered

From: Buck Evan <buck_at_yelp.com>
Date: Tue, 20 Oct 2015 16:59:02 -0700

If the loop is faster than 100ms (I would hope so), then the "read with
chance of blocking" case should happen for each character.

In case it's unclear, the `pv -qL 10` limits the pipe throughput to 10
bytes per second: 1 byte per 100ms.

On Tue, Oct 20, 2015 at 4:05 PM, Buck Evan <buck_at_yelp.com> wrote:

> There's no loop around echo:
>
> $ yes oh, hi! | pv -qL 10 | tai64n | s6-tai64nlocal
> 2015-10-20 09:47:02.681071500 oh, hi!
> 2015-10-20 09:47:03.493098500 oh, hi!
> 2015-10-20 09:47:04.304479500 oh, hi!
> ^C
>
> yes oh, hi! | pv -qL 10 | tai64n | tai64nlocal
> 2015-10-20 09:47:44.813611500 oh, hi!
> 2015-10-20 09:47:45.625627500 oh, hi!
> 2015-10-20 09:47:46.436990500 oh, hi!
> 2015-10-20 09:47:47.248335500 oh^C
>
>
>
> This prints slowly enough that I can *see* that tai64nlocal is printing
> each character separately, but s6-tainlocal is printing per-line.
>
>
>
> On Tue, Oct 20, 2015 at 3:27 PM, Laurent Bercot <ska-skaware_at_skarnet.org>
> wrote:
>
>> On 20/10/2015 23:36, Buck Evan wrote:
>>
>>> Is it expected that it's line-buffered?
>>>
>>
>> It's not line-buffered. It's optimally buffered, i.e. the buffer
>> is flushed whenever it's full (obviously) or whenever the loop
>> goes back to reading with a chance of blocking. When you test
>> with a loop around echo, you send lines one by one, so the
>> behaviour appears to be line buffering, but that's only an
>> artifact of your test.
>>
>> --
>> Laurent
>>
>>
>
Received on Tue Oct 20 2015 - 23:59:02 UTC