Stripping null byte blocks with a sparse file?

Question

I've created a sparse file with dd. How do I copy contents of another file there, leaving all the zero blocks unallocated?

possible duplicate of Can a file that was originally sparse and expanded be made re-sparse — Gilles 'SO- stop being evil', Jan 09 '13 at 23:39
That question and answer are a bit unclear, but I think they show a fundamental misunderstanding. If you overwrite a file using cp, the overwritten file disappears. It doesn't matter if that file used to be sparse; these things are not contagious. — Tobu, Jul 20 '13 at 13:40
Interesting, but I assume that the source file still exists. The question is if I can preserve this feature. — d33tah, Jul 20 '13 at 14:01

d33tah · Answer 1 · 2013-01-09T22:04:21.847

3

The solution is apparently to use cp --sparse=always. My first attempt was with writing some Python code, but unfortunately the MD5 sums didn't match (could anyone tell me why? the code's in edit history).

edited Jan 09 '13 at 22:04

answered Jan 09 '13 at 21:27

d33tah

1,371
14
28

1

The filesystem block size usually isn't 512B - more often it is 4KB. – peterph Jan 09 '13 at 22:05
You say that might be the reason that MD5 sums didn't match? – d33tah Jan 09 '13 at 22:10
No, filesystem gives you zeros where it finds an unallocated block, so the error had to be in your script. The block size is important for other reason: if you seek by 6 empty 512B sectors, then one with some data an then again seek say 4 empty sectors, you save nothing, because the data in the one sector causes the fs to allocate whole 4K block. – peterph Jan 09 '13 at 22:35
3

GNU tar also has the -S, --sparse option to handle sparse files effectively, btw. – peterph Jan 09 '13 at 22:36
@d33tah your python code is losing a byte every time it seeks because of the -1 in the seek target, which isn't necessary. You should seek ahead a full 512 bytes each time. – Jul 20 '13 at 17:37

Stripping null byte blocks with a sparse file?

1 Answers1