Bug 26632

Summary:	mmap runs out of memory on some files
Product:	Sisyphus	Reporter:	Ivan Zakharyaschev <imz>
Component:	hardlink	Assignee:	Alexey Gladkov <legion>
Status:	CLOSED FIXED	QA Contact:	qa-sisyphus
Severity:	major
Priority:	P3	CC:	boyarsh, glebfm, ldv, legion, mike, placeholder, vt
Version:	unstable
Hardware:	all
OS:	Linux

Description Ivan Zakharyaschev 2011-11-26 03:31:46 MSK

hardlink-1.0-alt4

There were some files with films inside the directory I ran hardlink on, and it failed processing them saying that mmap ran out of memory.

And they were not merged with their duplicates (exact copies).

I believe this is a major bug, because these means that the really large files are not treated, and getting rid of their duplicates could save really a lot of space.

Workaround: 

* compare the files on which hardlink fails, and delete/hardlink them manually;
* fdupes-1.40-alt2 and duff-0.4-alt1 were able to process these files, and they did detect the duplicates, although they work much slower. (I liked the interface of fdupes more than that of duff, because it displays the progress.)

Comment 1 Ivan Zakharyaschev 2011-11-26 03:33:54 MSK

hardlink could fallback to more complex or slower comparison methods on large files, rather than just letting mmap to fail.

Comment 2 Repository Robot 2011-11-26 04:57:46 MSK

hardlink-1.0-alt5 -> sisyphus:

* Sat Nov 26 2011 Dmitry V. Levin <ldv@altlinux> 1.0-alt5
- Merged with hardlink-1.0-owl1 (closes: #26632).

Comment 3 Michael Shigorin 2011-11-26 14:34:22 MSK

(In reply to comment #0)
> Workaround: 
* use hardlinkpy?

Comment 4 Ivan Zakharyaschev 2011-12-09 20:52:48 MSK

Thanks, now it works fine.

2mike: I searched for these tools with "apt-cache search duplicate", so I didn't see it (hardlinkpy) as an alternative.