Re: [Bug-tar] (forw) tar --listed-incremental bogous?

[Andreas Schuldei]

* Sergey Poznyakoff (address@hidden) [040216 21:44]:
> Notice, that listed-incremental, in its present state, assumes that
> the names given to tar are actually *directory* names (that's because
> it is based on comparing the directory mtime).  So, currently,
> instead of `tar --listed=name ... $DIR/file*' one should use
> `tar --listed=name ... $DIR'.

i would like to do that, but i cant think of a way to do that. i
try to put all my compressable files in one tar archive, and all
the uncompressable (bz2, gz, rar, mp3,...) into an other one.

While it is possible to operate on directories when creating the
compressable archive and --exclude *.bz2 ... all the
uncompressable files, i cant think of a way to 
operate on directories while taring up all the compressable
files. so i settled with creating a file list with find, and tar
all those files. if you tell me a way to operate on directories
(on tar`s level) while still catching all the uncompressable
files i will gladly do so.


> will, in the majority of runs, cause the timestamp saved in
> $SNAPSHOT_FILE to equal the mtime of the file $F1, and therefore
> this file will be considered new when invoking tar second time.
> That is the reason while file1 gets included twice. Simply adding `sleep 1',
> after `touch' (along with the above modification) will make tar behave
> as it should.

this does not work. i added filesizes for the dummyfiles to make
clear that really both files are stored in the tarfile. it seems
to me the snapshot file is not used in this operating mode at
all.

please run my demo script to see for yourself.

tar-demo.sh
Description: Bourne shell script