Re: [Bug-tar] need help with tar output stability problem

[Paul Eggert]

I cannot reproduce the problem that you observe.
With tar 1.24 and 1.25, I get the
same 60753920-byte file that you generate with tar 1.23.
I also get the same file if I use tar 1.15.1.  I am building
tar on RHEL 5.5 (x86-64) with GCC 4.5.1; except that the 1.15.1
tar is that of RHEL 5.5 itself.

> But between tar 1.23 and 1.24, tar began to
> output smaller tarballs for some inputs than it had before, which 
> while surely a good thing in general unfortunatly breaks pristine-tar.
> 
> -rw-r--r-- 1 joey joey 60753920 Nov 10 03:03 1.23/recreatetarball
> -rw-r--r-- 1 joey joey 60856320 Nov 10 17:38 1.24/recreatetarball

There seems to be some kind of confusion here, as the 1.24
tarball that you generate is larger than than the 1.23 tarball you
generate.  But you write that it is smaller.

You might want to look at your setup to see why your results differ
from mine.  I do notice that the tarballs differ in their representation
of this file:

icedove-l10n-3.1.6/upstream/af/chrome/locale/af/messenger/addressbook/replicationProgress.properties

This file name is exactly 100 bytes long, and if I recall, that used
to be a problem area in GNU tar.  The 1.24 tarball contains a longlink
representation of the file (which isn't right), whereas the 1.23 tarball
is right.

Perhaps you got a prerelease version of 1.24, with bugs?  Did you build
1.24 from sources yourself?  Have you tried with 1.25?  What platform
are you running GNU tar on, and how do you build GNU tar?