Re: This should not happen. Contact <address@hidden>. at parallel line 3384.

[Ole Tange]

On Tue, Mar 22, 2011 at 11:45 AM, Sebastian Eiser
<sebastian.eiser@gmail.com> wrote:
> Hi, thanks for the reply!
>
> you're right, --eta is the problem:

Good. That helps a lot in debugging.

> Il giorno 21/mar/2011 23.06, "Ole Tange" <tange@gnu.org> ha scritto:

>> * Can you run a simpler command and still make it fail?
> yes, this one:
> find 2006/ -type f -name *.he5 | parallel -j+0 --eta
> /home/seiser/src/gtproduct-omi-old netcdf --output /dev/shm/asdf {}

Can you run a simpler command that I can run, too?

Maybe you can save this as 'randomsleep':

#!/usr/bin/perl

print "@ARGV\n";
sleep rand()*5;

And run:

find 2006/ -type f -name *.he5 | parallel -j+0 --eta randomsleep {}

>> * Are your files local or on a network file system?
> All files are on a general parallel file system (gpfs).

Does it fail if you copy the files to your local filesystem first?

>> * Can you make a set of dummy files that fails too? Maybe a copy of
>> the original files but with no content: cp -rs /the/dir my_dir; find
>> my_dir -type l | parallel rm {} '&&' touch {}
>
> didn't try, let me know if you want me to test it as I didn't quite
> understand what you meant. Thanks for your help.

If you can make a set of dummy files that fails then you can give me
tar-file of that set: It is way easier for me to fix the bug if I can
reproduce it myself.


/Ole