Re: GNU Parallel Bug Reports --round and --linebuffer don't like each other

[Kosta Eleftheriou]

With 1dd2d6a, this no longer spawns multiple instances of cat as expected:

seq 10000000000 | parallel --pipe --round --linebuffer cat | tail

In fact, 2 instances are spawned, but no more.



On Thu, Nov 13, 2014 at 11:52 AM, Ole Tange <address@hidden> wrote:
> On Thu, Nov 13, 2014 at 7:28 PM, Kosta Eleftheriou <address@hidden> wrote:
>
>> Where can I find commit [1dd2d6a]? It doesn't seem to be publicly available 
>> yet.
>
> It is pushed now.
>
>> About the default grouping behavior, I am sure there is a healthy
>> debate on it, but I wonder if it would be better for people unfamiliar
>> with parallel to be able to see some output as soon as possible by
>> default, rather than wonder why no output is being produced in some
>> cases.
>
> And the decision has been 'no'. There are two primary reasons:
>
> GNU Parallel has done this since before it was GNU software. Changing
> the default behaviour need extremely good arguments. In the history of
> GNU Parallel a change of the behaviour has happened twice (namely -j9
> => -j100% and the behaviour of having multiple input sources
> --xapply).
>
> The author of GNU Parallel uses it for traceroute to multiple sites:
>
>   parallel traceroute ::: pi.dk freenetproject.org
>
> and here it is important that a complete job stays together.
>
>
> /Ole