[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Bug-tar] bug in filename buffer

From: Christian Wetzel
Subject: [Bug-tar] bug in filename buffer
Date: Thu, 14 Mar 2013 09:14:47 +0100
User-agent: Thunderbird (X11/20080213)

Hi there,

the new tar buffers the filenames read from stdin (cat <some_filenames> | tar c -T- | ...).
There are two problems with that.

1. tar does not check the available ram or address space. On a 32 bit system tar crashes
   when the filelist is too large.
2. On an 64 bit system it uses all available ram resources.

We used tar to copy a lot of files (billions) over the network like thist:

zcat large_filelist.gz | tar c -T- | netcat ip port

on the other side:

netcat -lp port | tar x

It was the fastest way to copy a large amount of files and this is no longer possible. Currently we downgraded to the old tar, but will there be a switch to turn this new behaviour off ? The gnu tools have always been incredible reliable (thanks for that) and now the streaming capability and reliability of tar get's thrown away for a imho unneccessary feature ?

Best regards,
Christian Wetzel

reply via email to

[Prev in Thread] Current Thread [Next in Thread]